Claude Fable 5, the most powerful AI model ever released, has been banned across the world. This is, and I mean this, the biggest news story of the past 3 years and has the potential to crash the entire global economy if not reversed in the next 2 days. Let's go over what this means, how it happened, why Anthropic is 100% to blame, what you should be doing right now, and why I believe the entire global economy is dependent on Trump reversing this decision immediately. We have a lot to cover. This is, without a doubt, the craziest news story of a very, very long time. Let's start with the timeline here and how we got to this point. Claude Fable 5 was released 3 days ago. Basically, everyone who has used it agrees it is not only the greatest AI model ever released, but it is probably the biggest technological jump of our lifetimes. Before this Opus 4 5, when that came out, really blew people's minds. The jump up in quality of code written and decisions the AI can make were mind-blowing. This took that to a whole new level. What I've been able to build with Fable 5 the last few days has been mind-blowing. But then, tonight, the news hits my phone. Anthropic announces that Trump demanded they shut this model down. They put out this huge statement that covers a lot of different things here, but let's break down what this means. What What is this ban, exactly? Here's what you need to know. Basically, Trump said only Americans can use Fable 5. All foreign nationals, both inside and outside the country, can no longer use this model. Why would he do that? Why would he make this ban? Well, for several reasons. Trump is claiming that a jailbreak of this model was released, which basically means people can use this model and get past the guardrails. They can use this model to build very dangerous things. And because this model is so powerful, it allows basically anyone to create these dangerous things. Anthropic slightly refutes this statement. They said, "Yes, jailbreaks have been found, but the jailbreaks are not general jailbreaks, right? A general jailbreak being you can do quite literally anything you want with the model like building chemical weapons." Anthropic says it's more like a spot jailbreak where you can only do tiny little things here and there, but they do admit there are jailbreaks found. But they also said these same jailbreaks are in every single model. They narked and said chat GBT 5.5 have the same jailbreaks. Here's where things get really, really interesting though, and this is going to tie into why I believe we are about to see shockwaves across the entire globe unless this is reversed very quickly. Foreign nationals inside Anthropic can no longer use Fable 5. People who literally work for Anthropic, but they're foreign nationals can no longer use the model they built. There are so many repercussions that come out of this. One is potentially big layoffs if foreign nationals can no longer work on frontier models at these AI labs. One, the AI labs are going to lay off the foreign nationals because they're pretty much useless at the company now, but two, they're going to stop hiring them as well. So massive disruption to the global workforce is one, but two, and this is the big one, and this is why I think this should be reversed in the next two days or else there are going to be major, major consequences. A complete disruption of the global supply chain. What does that mean? Here is the circular economy that the entire globe depends on right now. We've been seeing this go on a lot, right? AI companies like OpenAI will sign massive $500 billion contracts with Nvidia and Micron and other hardware companies that provide the hardware OpenAI and need to train their models. These contracts are for 5 to 10 years out, right? They're projecting their revenue over the next 5 to 10 years based on the revenue growth they're having globally at the moment, right? AI revenue is exploding, so they're projecting out revenue for the next 10 years, and they're signing these massive, massive, massive contracts with Nvidia and all these other companies. Those companies then turn around, take those hundreds of billions and trillions of dollars, and then invest it into other companies. Sometimes, right back into Open AI and Anthropic. This has basically been carrying the global economy for the last couple years. While inflation explodes, people are losing jobs, there's never been bigger profit margins, there's never been bigger revenue beats, companies have never been doing better. And that is because of this circular AI economy we've built. The issue is, and this is why this is so dangerous, and I believe this should get reversed by Monday, if Anthropic can no longer profit off of other countries and can only profit off Americans, their revenue is going to be significantly less over the next 5 years than originally projected. That means they will not be able to fulfill these massive multi-hundred-billion-dollar contracts they just signed with Nvidia and Micro and all these other hardware companies. And because they won't be able to fulfill those contracts, those companies won't be able to fulfill the contracts they have on the investments with other companies, and this entire circle collapses. Our global economy has basically been built on the assumption that AI will succeed. It's been built on the assumption that these companies will make trillions of dollars of profit over the next decade. By having such a massive disruption that Anthropic is no longer allowed to sell frontier model. They're still allowed to sell Sonnet, Haiku, Opus, but they're no longer allowed to sell Fable to other countries. Even though it's just that, that's still a massive disruption. That means they're going to buy less chips to train Frontier models on if only one country can use it. That means they're going to make less money off their subscriptions, their models, and this entire circle collapses. And this is the circle that the whole world is basically built on right now. I think it is not a coincidence that this was announced on a Friday night. It is 10:18 p.m. here in Cabo, Mexico. I'm on vacation. I dropped everything to cover this story. I don't think it's any coincidence. Basically, during the last 2 years of Trump's second term, every huge piece of financial news dropped on a Friday night. He is very cognizant of the stock market. He is very cognizant of the economy. Every war he started has started on a Friday night. It's never started on a Monday. I don't think it's any coincidence this was dropped on a Friday night because if this goes to Monday, if this was dropped on a weekday while the stock market was open, I have basically no doubt the entire stock market would crash. The reason why the Nasdaq's hitting all new highs, the S&P's hitting all new highs, is the assumption that these companies will be spending trillions of dollars over the next 10 years. That assumption is destroyed on this news. It appears, based on what's on TV, and this is not a political channel. I'm not into politics. It appears Trump is into the stock market. I do not think he would want the entire global economy and stock market to crash. So, I have the belief that this will be reversed by Sunday, which personally and selfishly I'm hoping happens because I've absolutely fell in love with this Fable model. The amount of building I've been able to do over the last few days has been absurd. Two more things I want to cover. First, why this is all Anthropic's fault. Anthropic has basically spent the last couple years using fearmongering as a marketing tactic. They've basically been saying AI is going to take every job, it's going to kill everything, it's going to destroy everything. It's the basically it's a doomsday. This all kind of crescendoed a couple months ago when Mythos was announced, what Fable is based off of. They said Mythos is too dangerous, we can't give it to the people, we can only give it to certain companies, we can't release this, that would be too dangerous. Too dangerous to put out in public. Then they go around and they put out articles saying the government should have the power to block and deter and stop dangerous AI models. What they were thinking they were doing, and this is my opinion, was setting themselves up to determine the rules. I think they were trying to capture policy, right? They were trying to help dictate what the rules in America would be. I don't think what they actually expected was what would happen tonight, which is the government would use what they were asking for against them, right? They asked the government to be able to block any sort of powerful models that are too dangerous. The government just did that to them. They asked for this. They've been asking for this type of thing to happen for a very long time. They've been using fearmongering when they should have been using hope instead, and that fearmongering has bit them in the wazoo. This is something they probably regret deeply. I mean, they've walked back the fearmongering over the last few weeks, it seems, cuz I think they started to see what's happening here. But this happening is 100% Anthropic's fault. I I believe they didn't use fearmongering in the marketing of Mythos, this would have never happened. But because they talked about how dangerous this is, now it is banned and no one can use it. So here's the big one, what this means for you. It means that chat GPT 5.5 is most powerful model again. This was the model I've been using for the past few weeks before Fable came out. I absolutely loved it. I think it's the first time in basically a year and a half that GPT had the best coding model. Fable changed that. When Fable came out, I said, "Well, this is a wrap. There's no way OpenAI will ever be able to catch up this cuz Fable is so incredible." Well, guess what? It took Fable getting banned for ChatGPT to come back. 5.5 is the model you need to be using now for any sort of coding or building and even just business planning talking. I find 5.5 to be fantastic. Also, you get way higher limits. You can use ton of it and it's much cheaper than Fable. But, here's another one and this is something I've actually been talking about for a long time now. I've been making predictions about this for months and months and months and it turns out all of those predictions are 100% true. I've been saying for months now local AI is the future. I've been saying we're going to reach a time where governments and corporations are going to be taking away your models. They'll be making them way too expensive and they'll be making it hard for the average person to access them. And here we are now, June 12th, the average person, really nobody can access the model. Even people inside Anthropic can access the model anymore. By buying hardware, by buying your own compute and running your own AI models locally, no one can take that away. There's no government that can take away your Mac Studio running Qwen 3.6. There's no government that can take away your DJI Spark that's running GLM 5.1. No one can take that away from you and that's why I believe open source and local models are the future because I think this is only the beginning. I think things like this will only continue to happen and ramp up. I just hope we don't face a situation over the next few days where these decisions lead to an economic collapse that impacts way more people than just the vibe coders who are building stuff with Fable right now. I hope this was helpful. I hope this got you up to date on everything going on. If it did, leave a like down below. Subscribe, turn on notifications. All I do is make amazing videos about AI. If things change, I will get out of the hot tub, come back up here, and report on it. I will see you in the next video.
Most of us don't use a single AI agent.
We use multiple of them for different
purposes because each one of them have
their own capabilities.
But none of them can see each other.
You are the one connecting them. Copy,
paste, and repeat. Now, every agent is
trapped in its own box, but what if they
were not? Now, an agent today is
basically the model plus the harness. A
model on its own just predicts text. A
harness is everything wrapped around it
that gets the work done. It usually
includes agent loop, tools, memories,
and a UI. Codex, Cloud, Code, Pi, each
one is a harness with similar ideas, but
very different implementations. And
different capabilities. Now, line up the
agents that you actually use. The here
are four different harnesses side by
side. Each one has its own memory, its
own UI, its own tools.
And no harness can see the other one. No
shared session, no shared history. Now,
we usually work simultaneously in most
of them, but what if you can put
everything under a single roof? Now, if
you build agents on top of them, the
wall hurts from the other side. When a
better model ships, say a new SDK or a
stronger harness, to adopt it, you
replumb everything you built and the
cost climbs. You are basically locked to
the layer you started on. Now, but if
you look closer at any harness, however
different they are from the inside,
everyone speaks the same language on the
outside. There are messages and files
in, text and tool calls out. So, it's a
great advantage that they have exactly
this identical interface. Now, if the
interface is identical, you can build
one layer over all of them. Take the
harness you already use and slide one
rail underneath them.
Every harness becomes an interchangeable
worker. So, a harness sits over a model
and this layer sits over the harness.
We can call this a meta harness. This is
exactly what Databricks just
open-sourced. They are calling it Omni.
It's a meta harness for all your AI
agents.
It's Apache 2.0, so you can build on top
of it.
It's one command, and every agent you
have runs under one roof.
They use it internally, so it's a
battle-tested.
Under the hood, it has three different
pieces.
On the left, you bring your agents.
These include proprietary agents like
Cloud Code,
Codex, or your custom agents,
which you can set up in the form of a
YAML file.
Then, runner wraps any of them in one
uniform sandbox session.
A server adds search history, policies,
MCPs, skills, and artifacts.
It's Postgres and deploys everywhere.
You can run this on Docker, Railway,
Fly, or Cloud Sandbox.
And it exposes that one session
everywhere,
whether you want to access it through
terminal, web native app, mobile, or a
REST API,
which is pretty great, because you can
now use the same interface interacting
with Codex, Cloud Code, Pi, or any agent
of your choice.
Now, because the session lives in the
layer, not the tool, there is just one
session object, which is your agent
files and history.
Every device is just a window onto it.
You can start in your terminal, continue
in the browser, or pick it up on your
phone,
which is pretty awesome, because you
have the same agent, same files, just
different interfaces, which are in sync,
and you can work from anywhere.
Okay, now let's talk about the
capabilities. This is an open-source
meta harness.
The beauty is that you can customize it
for your own need, if you want. Let's
first talk about what exactly does it
unlock. The first one is composition.
An agent is just a short YAML file which
includes a prompt, some tools, and a
harness.
Switching from Claude to Codex is
one-line change.
And you can run several at once as a
team.
Now, agents can even write agents. You
can just describe one and it authors the
file.
Now, they ship with two different
ready-made agents. The first one is
Polly.
Polly does not write any code. It's the
tech lead. It plans and splits the work
across coding agents in parallel, get
work trees, then routes each diff to a
reviewer from a different vendor than
which wrote the code.
So, say Claude codes
is reviewed by Codex code is reviewed by
Claude. And when you're happy with the
results, you just merge it. So,
cross-vendor
review only works about the harness.
Now, this planner, executor, and
reviewer or verify by patterns is
extremely important. Especially, you
don't want the same agent that wrote the
code to review its code because it has
internal biases.
And OmniJade makes it extremely easy.
Now, the second built-in agent is called
Debbie,
which basically is a brainstorm partner
with two heads.
So, the two are Claude and GPT. You can
I think bring your own one as well.
Every question goes to both at once.
You will get two answers side by side.
But here's the fun part. If you type
{slash} debate, these are going to
critique each other for a few rounds,
then converge.
A lot of people plan with say Codex and
then implement with Claude code or the
other way around. You could do that. Or
if you have to make an architectural
decision, this agent can be extremely
helpful.
Okay, the second big unlock that this
provides is control. Now, in this case
every action passes through a gate,
allow deny or ask you first.
Now, the thing is that this is not just
a polite request in a prompt. It is
enforced on every tool call.
And because it lives in the layer, the
rules can depend on history. This is
going to be extremely important,
especially if you want to impose cost
gaps, risk scores, repo and file scopes.
Or even things like PPI
scans, everything is built in. Now, this
is important, especially if you don't
want to have YOLO runs and really want
to make sure that
there are specific follow policies that
the agents follow. Okay, so how exactly
all of this work? Well, underneath all
of this is the OS sandbox.
So, every agent runs boxed in. It can
only touch the files and network you
allow. Now, another most important
feature is that it the agents cannot
directly read your secret keys.
The agent actually never sees this. The
layer injects it on the way out through
an approval proxy. So, even if you're
running the YOLO mode, it is going to be
a lot safer than just providing it
access to the agent. Now, the third
biggest unlock this provides is
collaboration. When your session is live
and you're driving it, you can share a
link and a teammate can watch the work
or even chat with it in real time. So,
basically this is
code driving and collaboration. The
beauty is that their messages run on
your machine.
Or you can simply fork it and take the
conversation your own way. Okay, let me
show you a quick demo of how exactly
this works in practice. Thanks to
Databricks for giving me early access in
making this video possible through their
sponsorship. In the rest of the video,
I'll show you how to set it up
and use it locally. All you need to do
is just run this command to install the
meta harness.
Now, after installation, the first thing
you want to do is to set up
this on your local machine.
Right now,
I'm using my cloud code subscription,
code x subscription, and Pi is using
Ollama.
In each one of this case, you can
add your own API keys or use your
subscription.
Then, you can use coding agent of your
choice. So, say you can use
the cloud code harness or code x
harness.
Or you can also use some of the built-in
agents. They have Poly, which is
basically a multi-agent orchestration
setup. Now, keep in mind, Omni harness
is not
a coding harness. It basically enables
you to interact with these multiple
harnesses directly.
So, Poly doesn't write code itself. It
decomposes your goal into subtasks and
delegates each one of them into a
subagent running on its own harness and
get work tree.
So, in my case, you can just directly
start this orchestrator agent. Now,
whenever you start a session, you're
going to see that it opens up this web
UI
along with the actual terminal window.
So, either you can work here in the
terminal or in the web UI or even there
is a desktop app.
The beauty is that all of them are going
to be sharing the exact same session.
To show you a quick example, I'm going
to describe a task. Create a single-page
web UI
that uses the Gemini Nano Banana model
for image generation. User provides
input
in the form of text. The output is going
to be an image. Also, add the ability
for the user to provide their API key
within and UI.
Now, we can just send this.
Okay, so on my machine, it wasn't
actually able to see the Pine and Cloud
Code CLI.
Uh so, I simply asked it to configure
those for me, and it went ahead and
configured everything. Which is pretty
awesome.
But more interestingly, you actually see
the same conversation happening exactly
in the terminal where I started this.
Right? So, these are different
interfaces which are interacting with
the exact same session.
Now, in this case, it's going to use
Cloud Code to implement things. Then
for review, it's going to use CodeX.
And it says that it runs autonomously
and will wake me up when it's done.
Right? So, it seems like the process is
running. If we look back, uh here are
basically the agents working under the
hood. So, it gives you visibility to
what exactly every agent is doing.
So, right now it's autonomously testing
the app. Okay, so it quickly tested the
app. Seems to be working.
Now, on the meta harness side, right now
the implementation is done by Cloud
Code. Then it started the independent
verification step. For this, it's using
CodeX. Now, the
interesting thing is that it's going to
be only passing on the diffs cuz there
are different work trees where these
agents or harnesses are working
independently.
Now, another feature is that you can
just directly interact with a specific
agent or harness. Which is pretty neat,
right? So, right now CodeX is reviewing
the code, but you can go and ask Cloud
Code something.
Now, here's another browser session that
I opened. I see exactly the same
processing happening. So, you could just
potentially deploy this in the cloud and
then share the link from here with your
coworker, and they will be able to
interact with the exact same session
that is running in the cloud. Or if it's
via local network, you can have the
session running on your machine and your
teammates will be able to interact with
that.
So, it's great for collaboration.
Okay, so a couple of other features I
think
are going to be very important for
everybody who's building with this,
especially given the cost of these API
based models is crazy right now. So, you
can actually see the session cost.
It gives you a breakdown of what exactly
was done, how many tokens was consumed
by each one of these models, but then
you can set up different policies.
[clears throat]
And I think this is very important. You
can have, let's say,
limit tool calls or for the specific
session,
uh maybe deny PPI and other requests,
right? So, these are contextual policies
that you can set. Even you can set
access to different tools or connectors,
but what I would highly recommend is to
set the cost. So, you can have a
session cost budget
or for user daily cost budget. I think
this is going to be more and more
important for organizations.
So,
just to give you an example, I would say
like $10, right?
And then you can define different
thresholds based on
soft
warnings.
Okay, so here's the app that is running.
It has a link to the Google AI Studio.
Now, here here was the initial
implementation from Cloud Code. Then
there was a
independent review from Codex and you
can actually see that it specifically
found issues.
Those were sent back. The implementation
was done again, tested again, right? And
this is kind of the loop that you want.
Now, you can write this orchestration
logic yourself, but
OmniGen ships this with their polyagent.
So, here's the final app that it
created.
A picture of a starfish wearing
sunglasses
jumping
with happiness. All right, so we're
going to see. This is pretty awesome.
Okay, there is a lot more to cover, but
do check out Omnigen. It's an open
source model. I think this meta harness
of orchestration layer is going to be
very critical, especially when you have
these different harnesses designed for
custom tasks with different
capabilities.
It's a very awesome project. Still
really early days. There might be
some tweaks that you'll need, but since
this is open source, I think this is
going to grow really fast. Again, thanks
to Databricks for giving me early access
and making this video possible.
Anyways, I hope you found this video
useful. Thanks for watching and as
always, see you in the next one.
Fable gone. The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national. Literally had 10 different agents running on Fable 5 right when this tweet went out and they all just stopped. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. This is a misunderstanding.
Before anything else — this video was inspired by Dzuma's video on digital forensics. That video is genuinely what convinced me to go down this path. Go show him some love, link is in the description. Now let's get into it. So it's two in the morning, Monster half empty, and I've just finished watching Dzuma's video and something about it won't leave me alone. I've been doing cybersecurity for a while — web exploitation, OSINT, networking — and I've always treated forensics like that one elective you keep putting off. Told myself it wasn't my thing. Tonight I decide to actually find out what I've been missing. No roadmap, no course, no fourteen Reddit threads telling me to get a certification I've never heard of. I just start pulling threads and seeing where they go. First thread — what actually happens when you delete a file? I know the theory, the OS removes the pointer and marks the space available, I've known that for years. But I've never actually seen it so I download Autopsy, grab a practice disk image from CyberDefenders, load it in and hit analyze. Within thirty seconds Autopsy is showing me deleted files that are completely recoverable, sitting there intact like they never left. I open one and it comes back with full contents, original filename, and timestamps showing exactly when it was created, when it was last opened, and when someone tried to make it disappear. I pause and look at my own laptop for a solid ten seconds. There are files on this machine I deleted two years ago that are probably still sitting on the disk right now just waiting. I close that thought and keep going. Autopsy also has this timeline view that I wasn't expecting and it completely changes how you see everything. Every file system event on the drive laid out chronologically — every file opened, modified, deleted — all with exact timestamps. You're not reconstructing what happened by guessing, you're just reading it like a logbook the machine kept without telling anyone. Browser history reconstructed automatically, recently accessed documents listed, and every USB device ever plugged in logged with its serial number and connection times. That last one stops me completely because it means deleting your files isn't enough, clearing your browser history isn't enough — the machine has been quietly keeping records in places most people never think to check. Then I find metadata and this is the one that gets everyone. Every file carries information about itself embedded silently — creation time, modification time, what software made it. Photos taken on a phone carry GPS coordinates by default. There's a tool called ExifTool that reads all of it in one command and the output on a random photo from your camera roll is genuinely unsettling the first time you see it. People have been caught doing seriously stupid things because they posted a photo online and forgot their phone was signing it with their exact location. The image looked completely innocent. The metadata did not. This is also the point where I realize that everything I'm finding tonight is scattered across documentation pages, GitHub repos, forum threads from 2014 and YouTube videos with forty views. The information is out there but there's nobody organizing it into something that actually builds on itself and when you hit a wall at two in the morning there's nobody to ask. That's genuinely why Cyberflow Academy exists — forensics, OSINT, web hacking, reverse engineering, all structured in the sequence that actually makes sense with real material to work through. There's a whole community of people actively in it who are finding bugs, landing bounties, going through the exact same walls you're going to hit. I'm in there directly, not behind a ticket system. Link is in the description, code Cyberflow50 for fifty percent off. Now back to it. The next thing I find is Volatility and this one genuinely surprises me. Volatility does memory forensics — instead of a hard drive you're analyzing a RAM dump, a snapshot of whatever was in a computer's active memory at one specific moment. RAM holds things that never touch the disk at all — running processes, active network connections, encryption keys, and malware that exists entirely in memory and leaves zero files on disk for antivirus to find. You run one command and get every process running at the moment of capture, cross-reference with network connections and you see exactly what was talking to what. The stuff that hides in RAM specifically to avoid disk-based detection is exactly what Volatility surfaces and the fact that it's completely free is something I genuinely didn't expect. By the time I come up for air it's almost four in the morning and I've gone from knowing basically nothing about forensics to having a real map of how the whole discipline fits together. And the thing that sticks with me most isn't any specific tool — it's the realization that the machine records far more than anyone realizes, in far more places than anyone thinks to check. The evidence is almost never hidden cleverly. It's just sitting in the artifacts waiting for someone who knows where to look. You don't need to be clever to do forensics. You need to be thorough. Which brings me to this video's challenge. Linked below is an image file and somewhere in that file's metadata is a hidden code. Find it, drop it in the comments, first three people win a free month of Cyberflow's private community. Use ExifTool, use Autopsy, whatever works. The answer is in the data the file isn't showing you on the surface. Good luck. And go watch Dzuma's video. That's genuinely where this whole thing started.
Every article you've read, and every idea you've had, is either lost in a random app or stuck in your memory where you'll forget it by next week. I built a system using Notebook LM and Obsidian that captures all of it and makes it searchable with AI. So, everything I've researched this year is in one place I can actually ask questions about. So, in this video I'm going to show you the full setup so you can build the same system today. The main problem with every note-taking app is that they're built for input, not output. Capturing information is easy. There are hundreds of apps that do that well. The problem is getting that information back when you actually need it. You highlight a paragraph in an article, save a bookmark, take notes during a meeting, and screenshot a tweet. All of that information exists somewhere, but when you're working on a project 3 weeks later and you need that one insight from that one article, you can't find it because it's buried in whatever app you happen to have open that day. Traditional note apps are passive storage. You put information in and it sits there until you manually remember to look for it, which you usually don't. That's the problem a second brain is supposed to solve. A second brain is an external system that captures what you learn and organizes it so your actual brain can focus on using information instead of trying to remember where you put it. has been around for years, but what's changed in 2026 is that AI makes a second brain more than just a fancy note-taking app. Because a system that actually works needs two capabilities: the ability to think about your information, meaning analyze it, compare sources, and answer questions about what you've collected, and the ability to remember it permanently in a format that's linked, searchable, and yours to keep. No single tool does both. Notebook LM is the strongest AI research tool available right now, but it's not designed for permanent storage or knowledge linking. Obsidian is the strongest knowledge management tool available right now, but it doesn't have built-in AI that can reason across your documents. [music] When you connect them, Notebook LM becomes the part that processes your information, and Obsidian becomes the part that stores and connects it, and together they form a system that neither tool can deliver on its own. The first layer is the research layer, >> [music] >> and that's where Notebook LM comes in. Notebook LM is a free AI research tool built by Google, and its job in the system is to take raw information and turn it into processed insight before it ever touches your permanent knowledge base. You start by creating focused notebooks, one notebook per research topic, not one giant notebook for everything you're working on. Keeping them separate means the AI's answers are always scoped to the topic you're working on. Inside each notebook, you upload the sources that are relevant to that topic. These can be PDFs, Google Docs, slides, YouTube videos, websites, or audio files. The free tier gives you up to 50 sources per notebook, 50 daily chats, and three audio overviews per day. The technique that changes the output quality is how you ask questions. The default approach is to upload a document and ask summarize this, which gives you a generic summary you could have gotten from any AI tool. The approach that actually produces useful insight is strategic questioning across multiple sources. You upload five articles on the same topic and ask what do these sources disagree about? Or what does source A claim that source B contradicts? Or which of these sources provides the strongest evidence for X? Notebook LM also has source toggling, which means you can turn specific sources on or off to control exactly what the AI draws from when answering. If you want the AI to only reference your internal documents and ignore the industry articles, you toggle those off. If you want it to compare two specific sources against each other, you toggle everything else off and keep just those two active. Audio overview is the feature that saves the most time. You click one button and Notebook LM generates a podcast-style conversation between two AI hosts who walk through your notebook, discuss the key findings, and highlight the patterns across your sources. I uploaded a stack of research papers on AI adoption trends, generated an audio overview, and listened to it during a run. By the time I got back, I had a clear mental model of the space without reading a single page, and I knew exactly which findings I wanted to save permanently. That's the role Notebook LM plays. It doesn't store your knowledge long-term. It processes raw material into insight, and the insight is what moves into Obsidian. Notebook LM handles the thinking, and the next layer handles the remembering. Obsidian is a free note-taking app that stores all your notes as plain text files in a folder on your computer. There's no proprietary format and no cloud [music] dependency, which means your notes belong to you and you can open them in any text editor, even if Obsidian disappeared tomorrow. The reason Obsidian works better than Notion, Google Docs, or Apple Notes for a second [music] brain is the linking system. Every note can connect to every other note using double bracket links. When you type two square brackets and start typing a note title, Obsidian auto completes and creates a bi-directional link between the two [music] notes. Over time, those links create a knowledge graph where ideas from different projects, different time periods, and different sources connect to each other automatically. You can click on any note and instantly see every other note that references it, which means you never lose the context around an idea. For organization, the system uses the PARA method, which stands for projects, areas, resources, and archives. Projects are active work with deadlines. A product launch, a hiring plan, and a client deliverable. Areas are ongoing responsibilities without a deadline. Team management, personal finance, professional development, these stay active as long as the responsibility exists. Resources are topics you're interested in but not actively working on. Industry trends, frameworks you've learned, tools you're evaluating. This is your reference library. Archives are completed or inactive items. Finished projects, outdated research, anything you're done with but might want to reference later. Each PARA category is a folder in Obsidian. Every note goes in the folder that matches its current purpose, and notes move between folders as their status changes. The structure takes about 5 minutes to create, and it scales to thousands of notes without becoming unmanageable because the PARA categories keep everything sorted by what it means to you, not by when you created it or which app it came from. [music] The PARA folders keep everything sorted, but the links between notes are where the value compounds because every new note you add makes every connected note more useful. And the workflow that feeds those notes is simpler than it sounds. The practical workflow looks like this, and once you've done it a few times, it becomes second nature. You start in NotebookLM, upload the sources for whatever you're currently researching, ask strategic questions, and let the AI surface the insights that matter. When NotebookLM gives you something valuable, you don't copy-paste the AI's output into Obsidian. You write the insight in your own words as a concise note because a note written in your own framing is something you'll actually understand and use 6 months from now. An AI-generated paragraph you pasted is something you'll skim past and forget. You create the note in Obsidian with a clear title that describes the insight, not the source. Retention drops when onboarding exceeds three steps is a useful title. Notes from McKinsey article is not because in three months you won't remember what that article said or why you saved it. Then you link it. This is the step that turns a collection of notes into a second brain. You connect the new note to every existing note in your vault that it relates to. The retention insight might link to your product strategy project, your onboarding research resource, and your customer experience area. Those links mean that when you open any of those notes in the future, the retention insight surfaces automatically. I was preparing a presentation on how AI is changing hiring practices. I created a Notebook LM notebook, uploaded six recent reports and articles on the topic, and asked three strategic questions. What's the biggest disagreement across these sources? What trend do all six sources agree on? And what's the strongest data point in any of these sources? Notebook LM came back with cited answers for each one. I pulled three insights from the responses, wrote each one as a note in my own words, placed them in my presentation project folder in Obsidian, and linked each one to my existing notes on AI workforce trends and talent strategy. That I built from those notes was stronger than anything I could have assembled by reading all six reports manually and trying to cross-reference them in my head. Every session like that adds a few more connected notes to your vault, and the value of those connections is something you don't feel on day one but becomes obvious after a few weeks. After one week of using this system, you'll have somewhere between 10 and 20 linked notes in your vault. At that point, the connections are sparse and the system feels like extra work compared to just writing notes in whatever app you had before. After one month, you'll have 50 to 100 linked notes. And that's when the system starts working for you instead of you working for it. You'll open a project note and see links to insights you captured three weeks ago from a completely different context that are suddenly relevant. Connections you didn't plan start appearing because the linking structure surfaces them automatically. After three months, your vault is a searchable knowledge base that reflects how you think about your work. Every research session, every meeting takeaway, every strategic insight is in one place, linked to everything it relates to, and retrievable in seconds. There's one trick that ties both tools together in a way most workflows miss. You can export your Obsidian notes as markdown files and upload them back into Notebook LM as sources. That means you can create a notebook from your own knowledge base and ask the AI questions about everything you've ever written down. I did this with 3 months of notes on AI trends and ask Notebook LM, "Based on everything I've collected, what's the biggest pattern I haven't explicitly connected yet?" It flagged a relationship between two trends I had noted separately but never linked. And that connection became the thesis of a presentation I gave the following week. That's the loop. Notebook LM processes new information into insight. Obsidian stores and links that insight permanently. And when you feed your Obsidian notes back into Notebook LM, the AI can reason across your entire knowledge base and surface patterns you missed. The system gets smarter every day you use it and it never forgets. And the one thing that makes this system even stronger is filling it with high-quality knowledge from the start. I put together a breakdown of 13 [music] free AI courses that cover everything from research to automation and you can watch that right here. Thank you for watching and I'll see you in the next one.
First, Sonnet changed engineering. Then, Opus outclassed Sonnet. And now, Fable 5 and Mythos 5 are outperforming Opus. By now, you've seen the headlines, you understand that this model is an absolute beast. And you also know Anthropic is rugpooling Fable 5 from our subscription plans, ProMax, Team, and Sebast. This is completely unprecedented. Forget the June 23rd subscription. Rugpool everyone was mad about. On June 12th, the government pulled the entire model. Both Fable 5 and Miffals 5 are no longer available. A federal export control order made Anthropic suspend the best two models that anyone has ever seen after the government apparently found a jailbreak. I'm not sure the federal government has the skills, resources, and agentic engineers to confidently find and address these brand new jailbreaking techniques, but apparently they have. The very controversial piece here is that Enthropic says these same jailbreaking techniques work on models like GBD 5.5, and it's not specific to Fable 5 or Mythos 5. So, it's very weird that these same tricks work on GBT 5.5, but Fable is the model that got rugpulled. It's not clear what's true yet. What is absolutely clear is that this is temporary and Fable will be re-released soon. So instead of focusing on all the mania, all the hype, all the chaos of this situation, let's refocus ourselves on three actionable ideas you can use for all next generation mythos level models. In this video, I have three observations on Claude Fable 5 specifically for agentic engineers that can help you get the most out of this model. I'll list them on the screen here and at the end I want to share exactly how this model is changing my approach to agentic engineering. Now running a model in isolation is meaningless. I used Cloud Fable 5 to orchestrate itself along its two little brothers Opus and Sonnet. They all ran the exact same five specs each in their own agent sandbox. They looped until the work was complete and created a single public URL where we can see their results built end to end. These are full stack applications orchestrated by Fable 5. That's 15 sandboxes in total. This is enough for us to understand what this model really gives you because it's not price per token. Each one of these URLs is live built by the respective agent. Here we have Fable recreating Simon Willis's LLM price index. We have a hacker news clone. We have a scikitlearn model generator. We have a chat room where we're chatting directly with the pi coding agent docs. We had an agent build an agent inside of a full stack application. And then we have this brand new type of application and we can directly chat to any one of the agents inside this chat room. But you'll notice something interesting here. Sonnet, Opus, and Fable were able to do the job. So what is Fable for? In this video, we break down the model that will do what Sonnet did. Change engineering forever. Let's talk Claude Fable 5. This is one of the first models that is truly feeling like a payto-play model. But then there's the other side of this coin which is this very very important statement. Fable 5 is not an Opus replacement. It's a tier above Opus. It's a mythos class model with mythos class pricing. So that's the other side of this coin. Now the question is is this actually true? Right away we can address this. In our total of 15 agent sandbox applications, we can see the price differential between Sonnet, Opus, and Fable. Sonnet ran $55 worth of tokens. Opus ran $91 worth of tokens and Fable ran 200 per token. Fable loses this battle. This model is not giving you an improved price per token. And the raw numbers don't look good. Throughout all these applications, Fable shipped this with a million tokens. Opus did it with 700K and Sonnet did it with also around 700K. Where is the value does more with fear tokens is not true. This model is not doing more with few tokens. So what is it doing exactly? So this is the first observation I really want to nail home with this model. It's not about price per token. It's not about intelligence per token. What we're really looking for is this price per intelligent agent hour. Time is our most scarce resource. This is what we're getting out of the new fable model. There's a sweet spot here. This leads us to another really important observation that we'll talk about in a moment. This model is only really useful if your missions, if your tasks, if your specs are big enough, are complex enough. We're getting into this really interesting place where we're not spending on tokens anymore. We're spending on agent time. We're spending on useful agent time. The best, easiest measure is in hours. But here's the breakdown. This is where Fable completely wins, right? And it's not really close. You can see here the time taken to generate every single one of our applications and there is a gap here right a sizable gap. Fable is completing the work with more tokens more expensively about 20% faster. So this is the key observation when you're using Fable 5. What you're getting here you're not buying tokens you're buying intelligent agent hours. Get out of task top to bottom mindset and get into feature shipped work completed end toend work done. the spec is driving Fable to those next level results where we can truly delegate, get out the loop and just judge the results and iterate from there. This model kind of pushes us further away from prompting back and forth, babysitting the agent, and repeating that cycle. It's more and more becoming about specs, proper delegation, proper looping, proper closed loop structures, making sure your agent can validate all of its work, and then understanding the review process. So, that's the headliner here, right? The harder the mission, the harder the task, the harder the feature you're trying to ship, the more Fable makes sense. Looking at our applications that were generated, the price just keeps going up. This model is just going to use more and more compute. You can see the cost roughly equivalent. It's just using double opus, double sonnet. And this is the big takeaway, right? On about 80% of these tasks, the sibling models, the brothers of Fable, Opus and Sonnet, did the job at a fraction of the price. put on our harder task here, specifically our multi- aent chat room application here. This is a significantly harder task. We have an entire micro IDE in here and our agents are operating on this inside of the UI full stack application. You can see all the personas in the bottom left. We're of course spinning this up using the pi coding agent under the hood on the server that fable oneshotted through this entire application. We can prompt this directly. I'll run on all give me your perspective on the application. And so all is going to kick off every single agent. You can see they're now starting to think here in the top left. We're chatting to everyone in this chat room and everyone is going to give me the work received. So they're showing their red files and they're reporting. So we're in this really really interesting application where we can just chat back and forth with all of our agents in a chat room. And so we just address all of them with at all. We can of course address our agents specifically. But you can see we got a nice perspective from every one of the agents in the chat room. And again, this is one of the harder tasks. We also have our PI documentation support agent. We have our market direction scikitlearn model predictor. We have hacker news clone much simpler and an LLM price index just presenting information again much simpler. You know what we're looking at here is all the fable version. So of course these are going to be the top-notch best versions but we also paid a massive amount for these results. And the big kicker here is that we didn't need to. This is one of the big ideas the kind of uncomfortable ideas we're going to get to in a moment. But one of the key observations, the larger the task, the harder the task, the more fable makes sense. And here's my kind of extreme version of this. If you can cure cancer with a million Fable tokens, $10 per million tokens is nothing, right? It's absolutely nothing because doing this work, curing cancer or whatever hard valuable work you're doing is is worth a lot more than $10 per million tokens. $10 per million tokens is nothing. At the same time, if you're centering a div, you're making a donation to Enthropic. Okay, let's just say it like it really is. The price premium here scales directly with the mission. It will create that chat app for you. It will do that, you know, minor front-end backend work for you. It'll create the migration file for you, but you'll be mostly wasting money. So, first takeaway, what we get out of Fable is price per intelligent hour. That's the new thing to compare up from price per token. If you just look at price per token, 2x opus, nobody likes that. That sucks. They're rugpooling Fable from the subscription plan. What we get from Fable is price per intelligent agent hour. Time is the resource that if you prompt contacts harness engineer properly with this model, this is what you get back. This is worth all the money in the world, right? It's time. All right. And so we can see this directly in our receipts and this small microcasm benchmark. The results show up pretty quickly. Just 15 items to compare here. 15 full stack applications that Fable orchestrated. Okay. And speaking of orchestration, that brings us to our second big observation for your agentic engineering. Cloud Fable 5 is not an intern. It is an orchestrator. I've been using this model since release and this is one of the standout pieces in terms of raw performance. Cloud Fable 5 is not an intern. It's not a worker. It's pushing toward the most valuable thing an agent can do and that is orchestrate. Every engineer's progression looks like this. Now it is you start with a base agent. You then make it better. You learn how to prompt and context engineer. You then add more agents. You then customize them. You specialize them to outperform any agent without that unique information. And then at the last level, you orchestrate every previous level. Last week we talked about Cloudflare's review software factory. They are using multi-agent orchestration. They understand the value of it. And of course, Enthropic understands the value of multi-agent orchestration. If we go to the system card and we just search for multi- aent. Guess what pops up? An entire section on multi- aent. What did they find? They found what everyone finds when they start scaling their work with agents. What do they do here? They have a bunch of benchmarks where they're specifically battle testing multi- aent orchestration with the best model and then with the best model scaled up to three, five, 10 async and then I think they have some versions where they're just running unlimited agents. Spin up as many as you want, right? Async sub agents. And as you can see here, accuracy on the left, latency per task on the right. What you get is exactly what you would expect. If you scale your compute, you scale your impact. aka if you use more agents and you have a great model that can steer them. If you have an orchestrator model like Cloud Mythos, like Fable 5, you get better results and you get them faster. Now, it's not always faster. Token usage across agents also adds a time cost. It's not always going to get the job done faster. Sometimes it's slower, it costs more, but you still get the accuracy, right? You still get the raw performance when I'm prioritizing what I'm looking for out of my models, out of my agents. It's always this, right? There's a trifecta for every single agent. You're always trading these three things off. This is the trade-off triangle. It is performance, speed, and cost. As a northstar, I'm always sacrificing speed and cost. But then very quickly, depending on if you're running a product agent or if you have a subscription where you can blast through tons and tons of tokens. That equation is going to change very quickly. There are many things you cannot deploy a fable level model or even an opus level model because the economics do not make sense. But back to multi-agent orchestration, the idea is simple and enthropic knows it. Again, everyone using models at scale that has pushed past the progression of agents base, better, more custom orchestrator. You know that if you want to push your results, you add agents and you add specialized agents and then you let the orchestrator do whatever it needs to do to get the job done. And so that's what we're seeing here. Enthropic seeing it in the benchmarks. This is not new for them. They found this pattern with Opus and they've been building their models to be better orchestrators. Another way to say that is that they're building their models to be better prompt engineers. So this is the second big takeaway for Fable 5. Okay, this model is not a worker, it's a leader. If you have this model center or make a stupid small change, you're wasting it. You're legitimately wasting it. This model is not a worker, it is a leader. And so, you know, in my head, this model is performing more and more like a principal engineer. For sure, if you're just vibe coding, if you're firing off random ad hoc prompts, you'll never see this capability out of this model. But if you are agentic engineering, if you're writing great prompts, if you're setting up your context, this thing can perform like a principal level engineer. What do principles do? They can do the job. Of course, they know how to do the job, but what they do best is they delegate to others. Okay, so Fable 5 is the ultimate orchestration model. If you're thinking about how to get maximum results out of this model, it's in delegation. It's in orchestration. It is in building multi- aent orchestration systems, multi- aent orchestration agent harnesses, so on and so forth. So this is the second really, really important observation. This is the pattern I used to create this right to benchmark this model. I had Fable 5 spin up multiple sandboxes. Five on itself. Five have Fable, five Opus, five Sonnet. They each got their own sandbox. They each ran the exact same five specs. And the interesting part here is in the results. Just to mention this, I think that every model release there is some fear that um everything can just be done with a prompt. I firmly do not believe this. I think engineers will continue to have a place in the world. In fact, we're going to need even more engineers that actually know what's going on. The way I think about this is the floor in the ceiling. So, models like this raise the floor, but they catapult the ceiling up as well. If you have been, you know, following channels like mine, if you've been doing the work using agents and trying to build systems that build systems, going to that metaentic engineering level, the ceiling is much higher for you. You can do a lot more with a model like Fable. That's a big observation here. Uh, Fable 5 is an orchestrator. Treat it like an orchestrator. Treat it like not even just a co-worker anymore. Like this is I'm really thinking about Fable as a principal engineer that knows how to given a great spec. I should preface with that. garbage in, garbage out for every system, right? That's just a foundational truth. But if you are writing great long specs with high detail with validation testing and review steps that make the loop very very clear, this model can perform like a principal engineer. And the best principal engineers delegate to scale far beyond themselves. Okay, it's the same thing. Fable 5 only has a million context window. And you know, a great example, just scrolling back up, I had a single Fable 5 instance kick off all 15 agents. And you know, I don't need to like prove this in any capacity, but let's just pull down the session I was using here. And this is the entire session to spin up the agents, run it, execute it, and then I use it to also build the presentation that we're looking at right now. Check out my context window. 62% 600,000 tokens. If I was not delegating, this work would be impossible. Look at the token usage inputs. Well, more than a million, okay, total combined outputs 100,000 looks like 2 million total. Look at the costs, right? The estimated costs from running all these agents and they're on sandboxes. Just to re-emphasize the point, this is impossible without multi- aent orchestration. And many things are impossible without multi-agent orchestration. That's why we've been talking about it on the channel for months, probably over a year now. I have no idea. I've lost track of time. We've done so many of these videos all the way back to some of the original ideas before Cloud Code was released and then of course Cloud Code's original release of sub agents, right? Going way back there. You know, drop a like if you were with the channel for that long and drop a comment as well. Shout out to you. But um here we are proving it in the future. Multi-agent orchestration is how you get outsized results. This was true back when you were prompting just a few cloud 3.5 sonnet models and it was true with cloud 4.5 and it's even more true here with Cloud Fable 5. So multi-agent orchestration super powerful. That's the second big observation. If you want to push this model to the limits, treat it like a principal engineer. Give it a large plan. Give it serious work to work through like spinning up 15 full stack applications in their own sandbox and then let it rip. Let it absolutely rip. And we can open up the other examples here, right? We also have Opus. There's our LLM price index. This is a real full stack application. I I just want to like express that it's hard to work through all of this content of what these models are capable of now. but Fable and then we can pull up Opus, right? Opus 4.8 and let's get Sonnet. Let's get this model comparison. And so this application is letting us quickly compare. This is directly inspired by Simon Willis's uh llmmpprices.com. He's had this site up for a while. You know, I told the agent, look at the site, clone it. This is one of the simplest things these models can do, right? Clone it and improve it. You can see here we have probably over a 100 models to compare. This is the Opus version. We can go ahead and pull up Opus Hacker News clone. Looks great. We can pull up scikitlearn via the obus version. Will spy close up or down tomorrow? This is our prediction scikitlearn models for any engineer that hasn't touched the data science world. You know, we've got 5,000 rows that we trained on. We split it up into train and test split. Then we have our pi chat application, right? Let's go ahead and ask a question about the pi coding agent. How do I customize the pi coding agent? Fire that off. Again, full stack application. You need to give these models even opus hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard hard tasks to understand their limits so one of my benchmarks now when a new model is released I give the orchestrator in this case Fable 5 access to a bunch of compute and by compute I mean CPU so using exev to create agent sandboxes and uh I'm giving each agent access to control the entire instance they need to then plan build and then host the application that they're building via a public URL so we can access it and look at it so this is all the work our agents are We have one more Opus version, the agent chat room. And you can see here that it's a similar type of application. It's fully built. Onto our third observation here, and this is kind of a scarier one for some engineers that might make you feel like you're getting left behind. The third observation is this is probably the first model engineers don't need. Why is that? It's because just like Opus 4.8 is saturating benchmarks left and right. That means you know with every benchmark with every tool call with every long agentic coding benchmark that's getting blasted through. Here we have Toolathon. This is a really popular one. All these benchmarks are proxies for doing real engineering work. You can see that these numbers just keep going up. First run pass, three run pass average turns to solve the problem. You can see that these numbers keep going down and performance keeps going up. Even the first shot, one shot performance keeps going up. So with each one of those benchmarks are getting saturated but also engineering full stack work is getting saturated. And this is why I use this exact benchmark that I like to use. Spinning up agents in their own virtual environments having them host full stack applications very quickly at scale thanks to our orchestrator agent here. But this is why I like to do this because it shows something really really important. Let's open up sonnet. Sonnet lm prices. Okay look a little bug there. Not too bad. Let's open up our hackernews clone from sonnet. Okay not too bad. different styling, but same thing. We can go into the comments right here. Some of the comments from the Cloud 5 release. This is a real clone using a snapshot of data, but it looks like it did the job great. Let's go ahead and look at Sonnet Scikitlearn. So, can we spin up new scikitlearn models? Okay, not too bad. Again, you know, looks like it's getting us 80 90% of what Opus and Fable have. So, let me just kind of nail this point home. Right, here's our PI chat app. How do I add uh new tools to my extension? Fire that off. Then we have our Sonnet chat room. For sure, one of the hardest to build. Let's see if we've also done it here. We'll do that same prompt. All give me your quick summary on this app. Okay. So, let's see if Sonnet has also done the job. You can see our agents are thinking over here. And let's see if we get that pi coding agent response. There it is. We have two responses. It looks okay. You know, you can probably assume there's something off here. But as we look at the application and we look through these files, it looks okay. And this brings me to the point. All these agents are doing the job and the difference is in how close are they getting to the result and how quickly. It's not just about cost, it's about time. So our third observation here is this is the first model that you might not need at all. This is like a weird realization that I came to while testing this model. In a lot of these cases, you can do probably 85 maybe 90% of what Fable can do. Very very weird, right? It's kind of a weird thing to think that I don't need the state-of-the-art model for this problem for many problems. And even further, there's like a checklist that you can go through. You can now create, you know, a model stack, state-of-the-art workhorse and lightweight models. We talked about this again in our last week's video around tokconomics. Cloudflare had S tier tokconomics because they built something like this out for engineering work, for product work. You don't need state-of-the-art all the time. And in fact, I think we've finally hit the point where many of us engineers don't need Fable. And Opus is enough. You don't have to pay to play. you just don't have to. Before you upgrade, you have to ask yourself a few questions. If you can't write long detailed plans, if you don't know what questions to ask, if you can't envision the solution end to end, maybe you don't need Fable. Maybe Opus is enough. Kind of a weird realization, but I think this is now true. This is the first model that you do not need to upgrade to. We can see that in a lot of the benchmarks. Opus is a great model, probably one of the best there is. You can see it in the benchmarks. It's not like Opus is very far behind Mythos or Fable. It's actually quite close. In a lot of scenarios, you won't be able to tell the difference. And that's kind of the key idea I want to communicate to you here. This is the first model that probably 80% of engineers don't need. I can say that pretty confidently. If you're not writing long plans, greater than 100 lines, 200 lines, 500 lines, multiple pages, right? HTML specs, specs with images. If you're not doing that, you probably don't need this at all because your problems aren't complex enough. And it's not like a slight. I'm not saying that to like be mean or anything calling your problems small, but like it's just true. Like don't waste money on compute you don't need. Going back to the first point here, it's really about price per intelligent agent hour. This is what you want to hit. And if it's not worth it, just use Opus. Opus is a great model. It's going to do a lot of great work. I'm going to continue to use Opus, especially come June 22nd. But for my best, most complex, hardest work, I'm for sure absolutely not going to be shy when it comes to using Fable 5. So three observations there. How is this concretely changing the way I am doing my agentic engineering? There are two constraints of agentic engineering. We talk about them on the channel quite a bit. There's planning and reviewing. A sign of a new state-of-the-art model is you can sit down, ask for more, and expect it to be built exactly as asked. And once again, let me just echo this again. 600k tokens. I built 15 full stack applications for sure with some issues, especially on the sonnet level. But I did this because I wrote a great detailed plan and then the agent took it from there. It wrote all the specs in my specific spec format that I templated. It's a rich HTML plus images plus structured document. It did all that. Then it executed and orchestrated agents as requested. And if you're in the cloud code world, you're likely familiar with, you know, you can fire off the workflows, you can fire off the loop to make sure that it's hitting all the validation commands. It doesn't really matter. It doesn't matter how you do it. It just matters that you know what tools are available to accomplish the work. So when I'm thinking about how I'm changing my work around this model, these state-of-the-art models always allow me to do this. More planning, less reviewing. And it's not less reviewing because I'm getting lazy. Never get lazy with your engineering work, especially when it hits production. It's more planning, less reviewing because it's listening more closely to everything I put in my plan. And this is one of the big differences between vibe coding and agentic engineering. Are you saying exactly what you want the model to do? and can it do it? And with these models, Fable 5 is a great example. With the leap forward, you can do this more and more. You can sit down, type exactly what you want to see, and then expect it to be built specifically because you're writing great specs. Again, this comes right back to our third observation. If you're finding that Opus can do the work literally like in this case, right, we have several examples where I can't tell the difference between Opus and Fable. And that is because specific example right our LLM price index our hacker news front page and will spy go up or down tomorrow scikitlearn classifier models I cannot tell the difference between these three between fable and opus. Why is that? It's because it doesn't matter. You might like the UI a little more. You might like a small tweak here or there but that's not what really matters. What matters is did the agent do the job you asked it to do. And so where things do start to deviate is maybe our PI documentation chat app where we had the agent build an agent inside a full stack app. So this you know definitely getting more complex here. You can see that Fable did complete this properly right we're getting some great docs right side left side classic chat on right type of application and it also built out an entire file tree inside the application for the agents to work in. And now I can chat with any specific agent at tester run our tests. So now all the agents are thinking is this relevant to me? Nope. Everyone turned off except our tester here. And now tester is going to run all these tests. Okay, it's executed our test. There it is. There's the run command. There's the results. This is getting very meta when you start doing agent on agent on agent on environment on environment. Type agentic engineering. But that is for sure where everything is going. Anyway, to continue to communicate the idea here, here's opus LLM price index looks great. Little spy close up or down tomorrow. Looks decent. And then it looks like it did finally get that uh last result here. Right. If we say show more, how do I customize the pi coding agent? It gave me a nice breakdown here. The UI here a little off. Would be nice to have this to be extendable. But we didn't detail this. This is this was my fault. I didn't detail the exact UI in the prompt, but it looks like Opus did accomplish this work. And then our multi- aent chat app. Let's go ahead and see did we get a summary here. Uh let's see at all. Summarize your understanding of this codebase. So firing this off in Opus. Once again, you can see all our agents thinking here on the sidebar. This is a aentic UI application I've been envisioning and haven't really put into work. Starting to get our responses from every one of our agents. I built out a version of this in the PI coding agent in the past, but as a standalone UI. Uh this is definitely an idea I'm thinking about a lot more for multi-agent orchestration, being able to spin up multiple chat rooms where you have specific agents working on specific parts of your work. We'll probably cover on the channel at some point or someone will probably steal the idea. The point here I'm trying to make like let me just dial all this in. In the end, every model release is about understanding how far you can push your model. For a lot of engineering work, a lot of engineers, myself included, at certain times, we're going to be wasting tokens and wasting money on Fable. And you can push against this by pushing the model to its absolute limit. Sit down, ask for as much work as you possibly can, and then let it crank. Understand where it starts to fall apart. And then that's where you start making up for its downsides inside of your custom agent harness by scaling more compute, right? Add more verifiers, add more reviewers, so on and so forth, right? That's where your real agentic engineering prowess and skill comes in. Maybe controversially, I fully believe every model can be improved by adding another model that substitutes what that first model was missing. Often times, I'm talking about the exact same model. Just add another model. Scale your computer, scale your impact. I think anthropic based on all their work would probably agree with that. Um, I'm certainly betting huge on that. When I think about my road map with this new model, it's really about this. I'm going to be planning more work and I'm going to be doing less reviewing because the model's adhering to the plan more. And so, outside of the orchestration realm, which is where Fable definitely shines and should be used, this model will absolutely be used and should be used for doing the actual work. You know, principal engineers do ship the best work as well, right? They delegate and they do the best work. And there's a whole slew of review focused agents, prompts, skills, so on and so forth that can now be wired up and now be deployed thanks to Fable 5. So that's one big change I'm making. The whole point here is to hit the goal of 2026, right? My goal for this channel, if you're an engineer watching, and for myself, is to gain full trust in my agentic system so that I can ship into production from a single prompt. And Fable is absolutely going to be a key part of that because it can execute longer, more complex plans over longer periods of time. So anyone can ship a readme change into production with one prompt. Uh not a lot of engineers are probably should be experimenting with that more. But where the real skill of agentic engineer comes in is in fully trusting your agentic system so well that you can run a single prompt. You can have your agentic system run, plan, build, test, review, document, and then it's in production. You ship from a single prompt. This is the northstar of agentic engineering. This is ZTE, zero touch engineering, which I realize is a misnomer. It's technically one-touch engineering, but zero touch when you compare it to everything we used to do as engineers makes a lot more sense. The way I'm thinking about this, just as like a putting all of our observations together, the more ambitious the instruction, the better the result. And this is where really where you're going to get your return on investment for using Fable 5. With that being said, this is the first model where for most work, you probably don't need it. If you can, and if you're solving hard problems, I absolutely recommend Fable. I'm going to be using it. It's very clear this model is a beast, but of course, we're going to have to pay for it. Kind of a weird headline, kind of a weird perspective, but I think, you know, the truth is always more nuanced than we'd all like to admit. I think that Fable 5 is an absolute beast, is an incredible leap forward over Opus, the previous state-of-the-art, and most engineers don't need it. This is an opportunity for you to prove if you really need this model. Push it hard. obviously largest and useful specs that you possibly can, right? Really think through everything you want done and how you would test and verify that it works. Um, but I do think that this model has outgrown what most engineers can even ask for. I'm super curious. Comment down below if you made it to the end. First off, you know, thank you. Big shout out to you. But also comment down below what's the biggest task, the biggest feature you've handed off to Fable so far. I'm super curious what you've one-shotted in Fable or few shoted in Fable. This is how I'm thinking about how to best push and use this model. Ask for more. Think fully about what you want to see and then hand it off to the agent and let it fire off a powerful multi- aent orchestration system. This is how we push the frontier beyond. And we have clear and clean signal from Enthropic that that is exactly how to push these models. If you want to scale your compute to scale your impact, multi- aent orchestration is the answer. The right way to be thinking about this model and probably more models in the future, specifically the state-of-the-art models, is this is a new tier with a new price and it gives you new results. The thing you want to look for is not price per token. A lot of engineers are going to convince themselves out of the most important technology of our lifetime by looking at it this way. You and I are looking for price per intelligent agent hour. That is the true signal. These are three observations I've seen coming out of the Cloud Fable 5 release. Comment down below. Let me know if you agree or disagree with any of them. If you made it this far into the video, huge shout out to you. Thank you. You know where to find me every single Monday. Stay focused and keep building.
[clears throat] >> Ah! >> [music]
1 year ago, I showed you how I saved $7,000 a month on web hosting. Today, I'll make you forget about paying for any automation tool. I will show you how I run unlimited AI bot automations without using an 8 and make Zapier and without any limits. If you are ready, let's get started. This is Pylon. I built it and it's completely free and open source. And it runs every automation you're about to see. Let me show you how it works with three practical examples. Number one, brand tracking. Tools like Brand24 charge up to $199 per month with limits. These tools scan the web for every time your brand or your product gets mentioned online. It's very important, but really expensive. Watch this. Done and look. It emails me the report automatically. Now, here's the part I love. I set the schedule once, daily, weekly, whatever you want. It runs forever without limits, without subscriptions. Look at this one now, the keyword rank tracking. Anyone working online with a website, with a product wants to track their site's rankings on Google. And if you go for an SEO tracking tool, you will commit to like $30 or more a month. With Pylon, the same. Click, done. And I get the report into my inbox. And you see what's happening here? We are replacing subscription-based tools with simple automation using Pylon. And we are just getting started. Now, here's how Pylon actually run under the hood. It runs simple Python scripts. And I know what you're thinking now, I don't know how to code, so how I can use this? No, don't worry. You don't have to code anything. I'll show you this in a second. And what makes Pyloners super powerful is actually Python itself. Python has over 650,000 libraries. You can't imagine the flexibility you get. You can build literally anything you want. Let me show you what I mean with a real example I built. I want to automate with AI my YouTube channel comments management. So, I'm going to build it in front of you right now to show you how easy it is. What I did, I built a closed skill. You just install in closed, it's free. I describe what I want in plain English and in few seconds closed writes the entire Python script for me. Copy it, paste into Pyloner, schedule if you want, hit run, and here we are. It is that simple. The same automation would take you an entire afternoon connecting nodes in N8N. And honestly, I hate this. With Pyloner now, I set automations in minutes. But, honestly, now I want to share the most important point in this video. The skill you should be investing right now in 2026 and beyond is not learning N8N or Zapier. Those tools come and go. The skill that everyone should be learning is learning how to communicate the right way with AI. Now, in this video I made things easy for you with the closed skill. But, you should be learning how to create skills so you can build anything you want. That's the actual unlock. Start today and thank me later. Now, you're probably thinking, "Okay, how I actually run this?" Let me show you, it's super simple. Step one, get a VPS. I use Contabo and you can grab one for like $5 a month. Step two, install Coolify with one simple command. This is the same setup I used in my $700 saving video. Link below if you missed it. Step three, deploy Pyloner on Coolify. Enter the Docker URL, set the environmental variables, hit deploy, and that's it. You are running. And if you already self-host like me, this costs you nothing extra. It's completely free. If you are starting from scratch, I put together a full step-by-step guide in the description below. It's totally free. You can check it out and get started. And if you want to go in deep every service I use, every trick, you can check my full course on self- hosting linked below, too. Look, I'm not just saving $200 a month with Coolify Runner replacing different SaaS and tools. I built a system that no SaaS could sell me, and you can, too. If you learned something new in this video, smash the like button, and see you in the upcoming videos.
the AI world is buzzing about Apple's WWDC event. A big chunk of this keynote was dedicated to like the Siri that we kind of always hoped Siri would be. They asked about tickets for a concert and it told them they had to enter a lottery and then he set a reminder to enter the lottery at that time. He showed it a picture on Instagram and asked where it was and it looked at that picture, figured out where it was and then gave him directions to where that was. You know, he was rearranging photos by dictating to Siri. Another thing I found kind of interesting is if you're on like a Mac or a MacBook, Siri's going to come to spotlight. So, you know, if you hit command space bar and bring up this little spotlight search, you're going to be able to prompt it with AI and say, you know, rearrange these folders or look at all of these documents and tell me the differences between them, things like that. They're also adding a new Siri mode in the camera app. So, in this picture, they're pointing it at a red ball and because they're in Siri mode, Siri starts telling them this appears to be a traditional cricket ball, etc. Now, out of the entire keynote, this was the thing I was probably the most excited about. So, they're adding this feature called describe a shortcut, which allows you to use AI to say, here's what I want to happen and it will actually go and build that shortcut for you. Like, you just say, when I'm leaving work, message Pedro I'm on my way with my ETA and it will actually go and build that shortcut for you. I like never use the shortcuts feature, but if I can just like explain what I want it to do and it goes and automatically figures out how to build that shortcut for me, that'll be pretty cool, I think. They're actually going to be using their own foundation models as well as Google's Gemini models. They explained that when you use Siri's AI on the phone, it's either going to handle that AI prompt on the phone directly or using private clouds where Google gets none of your data saved.
Have you ever wanted to create professional AI videos, but your only available device was your phone? Most AI video generators are built entirely for desktop. That's why I took eight of the biggest ones to see how they hold up when you run them through mobile. And by the end of this video, you'll have a clear picture of which tools are actually good on mobile and which ones are just a waste of time. Now, this test isn't about output quality, but what I want to see is how the actual experience of using these tools on a phone is and how easy and smooth it all feels. Let's start with VO 3.1 from Google. And the first thing you need to know going in is that there is no mobile app for this one. So if you want to use it on your phone right now, you have two options. The Gemini app or Google AI Studio in a mobile browser. A native AI Studio Android app is on the way and it's already on the Play Store open for pre-registration. So this will improve at some point, but right now you're working around the problem rather than through it. Most people will try the Gemini Path first because the app is already on their phone. And honestly, the app itself is clean and familiar, which is genuinely nice. But VO is buried inside a general purpose chat interface. You're not opening an actual video tool. You're opening a chatbot describing what you want and hoping it interprets that correctly. And you get three video generations per day through this path. So, it's very limited. There are no more manual settings, no duration control, nothing you can do to affect your final output other than your prompt. The app's saving grace is its cancel button. Because the bot often generates responses mid-sentence, this button is a vital safety net. With a strict limit of three slots a day, being able to halt an accidental prompt instantly keeps you from wasting your precious quota on a half-baked idea. Now, the AI Studio browser path gives you more actual control, and you can see real settings, which is a step up. But the interface is built entirely for a desktop screen. On a phone, you're pinching and zooming just to navigate around it. It's one of the slowest workflows I went through in this entire test. The model quality itself is real. The output you can get from VO is good and reliable, but the way you access it on mobile right now add so much friction that I actually found myself not wanting to use it entirely. However, the next tool I tested was a way more pleasant experience. I'm talking about Higsfield, which most people know it as a desktop platform, but there's actually a really easy way to get it on your phone, and that's through a progressive web app. You open your browser, go to higsfield.ai, and run the app straight in your mobile browser without having to install anything. From that point on, it opens full screen, and it looks and behaves exactly like a native app. What actually got me was that the UI isn't just a small version of the desktop site, but it's been rebuilt for mobile. The layout, the navigation, and the way elements are positioned are all designed around how you actually hold and use a phone, and the performance matches that completely. I went through the entire platform and it ran without a single moment of lag. And it's the only tool I can say that about on this list. And since it's an all-in-one platform, you get access to all the popular models like Cedense 2, Cling 3, and VO3.1 alongside so many unique features that take content creation to another level. Everything is accessible and it renders just as fast as it does on a desktop. The only drawback is that since this is a PWA rather than a native app, you may need to log back in. But that's small when in return you get one of the smoothest and mobile friendly experiences. The next tool approaches mobile completely differently and it is one of the more interesting ones here with the experience changing quite a bit depending on which phone you are on. With Pika the first thing you need to know is that if you're on Android there is no official Pika app. The real team hasn't put anything on Android. iOS gets an official native app called Peak Effects by Pika and it's on the app store. So, if you're on iPhone, you're getting a proper native experience, but Android users are completely locked into a browser experience, and that's a significant portion of the mobile market being pushed away from a native experience before they've even started. That said, I want to be honest about what the browser experience actually feels like because it surprised me. The layout at Pika.art is clearly designed with mobile in mind. There's an inspiration feed, a peak effects tab, and a library tab running across the top. At the bottom, there's a clean prompt bar with effect icons, a duration selector, and a model switcher. It doesn't feel like a desktop site that got squeezed onto a phone. Scrolling is smooth. Switching between tabs feels natural, and uploading a video from your camera roll, works natively in the browser. The core creation modes, including Pika Twists, selfies, and direct video uploads all work without major issues. There's an occasional 2-second freeze, but it's rare enough that it doesn't break the experience in any meaningful way, and there are no pop-up ads. It's a solid browser experience and I want to give it credit for that. It just isn't a native app for half the people who might want to use it and that's a hard thing to overlook when you're specifically evaluating these tools on mobile. The next tool has a native app which immediately puts it in a different category, but the experience of actually opening it for the first time surprised me. Hyalo AI is one of the less popular video models. But when you open the app, the creation experience is actually clean. Once you're in the text to video or image to video screens, the layout is minimal. The prompt area is clear. The settings are easy to read and image reference works natively on mobile. The core generation modes are all there and fully functional. If that was all there was to it, this would be a straightforward recommendation. But after spending just a few minutes using it, you'll notice a big problem. When the app opens, you're immediately hit with a pop-up, then another one. The homepage itself stutters when you try to navigate through it. Scrolling feels slow, and the transitions between sections don't feel smooth. you end up spending real time just trying to get past the front layer and into the actual tool. It's also worth mentioning that iOS availability is uncertain. Depending on your version, it might now be available, which is something to keep in mind. It genuinely feels like two different apps stitched together. Trying to get in feels messy with lots of ads, but once you get past that, you have a very solid creation experience. And if you thought that Hyo's opening was rough, the next one will surprise you even more, and not in a good way. Let's take a look at Pixver. Now, before I even get into how the app actually runs, I want to point out that right now, Pixverse doesn't have a dedicated iOS app. Apple users have to make do with Pixver Light, which is a pretty limited experience compared to the full thing. And because of that, the App Store is filled up with fake Pixver clones, and if you don't know exactly what you're looking for, you can easily end up downloading the wrong thing. However, Pixver does have a web app that you can access right here. On Android, the app is easy to find and actively updated. So, that's where I tested it. And the moment it opened, I was hit with a pop-up. And the moment I closed it, I immediately got another one. I got hit by two pop-ups before I'd seen a single feature or done a single thing. When you get past them, you'll notice a cluttered social style discovery feed. It looks more like a social media platform than a video creation tool, and it's honestly overwhelming. Now, there are some benefits. The text to video creation screen, once you actually reach it, is reasonably clean. There are also a lot of trending templates and effects available if that's what you're after. But between the double pop-ups, the cluttered homepage, and the lag, the overall experience was the most frustrating one I had in this test. After that, I was genuinely curious whether the next tool would continue the pattern or finally break it. Clingai is available as a proper native app on both Android and iOS, and they're both actively maintained with no fake clone problem to worry about. Once I opened it, I was surprised. All of the features are laid out cleanly and logically. The UI is minimal, the controls are easy to find, and everything responds smoothly. Smart multi-shot is there. Motion control is fully functional on your phone. These are genuinely advanced features that make Cling stand out, and they work exactly as you'd expect them to on a small screen. Uploading from your camera roll is clean, and getting your output back out is straightforward. Once you are inside the actual workflow, cling is a noticeably different experience from most of what came before it. The homepage is where it gets a bit complicated. It's built like a social platform with a banner carousel at the top, feature tiles below that, a trends section that leans heavily on sports content, and a for you feed underneath all of it. It's busy, and navigating through it creates noticeable stutters, but depending on what you want out of it, you might find these things useful. But for plain video generation, that social feed structure gets in your way, and it's the one thing that holds the overall experience back from being as clean as it could be. It's a tool that clearly has a strong creation experience sitting behind a homepage that's trying to do too many things at once. And the gap between those two parts of the app is genuinely noticeable. The next tool takes a quieter approach to the homepage, and it makes a real difference in how the whole experience feels from the moment you open it. Imagine Art is one of the tools that hold up really well when you actually sit down and use it. There's a native app on both Android and iOS, and when you open it, there is an upsell pop-up, but everything feels clean after you get rid of it. The homepage has a clear AI tools row at the top with colorful icons for all tools, AI assist, AI images, AI videos, and AI edit. It gets a bit busier further down with UGC video templates and an AI photoshoot gallery, but it never feels overwhelming. Now, there is a slight input delay throughout the app. It's nothing major, just a small lag on taps and transitions. That means it isn't quite as smooth as you'd want it to be. It's noticeable, but it doesn't stop you from doing anything. What I kept coming back to was how much you can actually do on mobile. AI videos, AI images, AI edit, UGC templates, and AI photo shoot are all accessible without anything being locked away. And once you're inside the video generation tab, everything runs smoothly. The screen is focused, the controls are clean, and the depth of what you can actually do is genuinely strong for a tool that doesn't get talked about as much as some of the bigger names. It's not the flashiest tool in this list, but it's dependable, it's complete, and it doesn't try to upsell you before you've even written a prompt. That combination is more valuable than it sounds. The last tool I tested is the one that gave me the best first impression of anything in this entire test. I've been through a lot of app openings at this point. So, when I open Runway and was met with a clean, dark screen that simply asked what I wanted to make today, it genuinely stopped me for a sec. There are no pop-ups, no social feed, and no distractions. You just have a very clean and easy to understand interface that's waiting on your input. It's also available on both Android and iOS, which is a big plus. From there, the experience stays consistent. The side menu organizes everything clearly, and all the features are laid out in a way that's easy to navigate. Gen 4.5 video is right there and fully functional. Everything about moving through this app feels considered and intentional in a way that's genuinely rare across the tools I tested. The UI is really good from a design perspective, and the experience of using it is genuinely enjoyable. But if you're someone who wants to do professional work from your phone and not just quick clips, you'll eventually notice that the feature set doesn't go as deep as some of the other tools in this test. What I mean by that is that on mobile, you're essentially working with the core generation tools and not much else. Things like Act 2 and Characters are there, but the overall toolkit feels like it was cut for simplicity rather than depth. you're not getting the kind of layered creative control on your phone that a tool like Higsfield gives you. So, if you're trying to build out a full workflow from your phone, you'll start hitting the ceiling quicker than you'd expect from a tool with Runway's reputation. What this test made clear is that most of these tools were built for desktop first and mobile second. So, here's the final ranking. VO comes in at number eight. No dedicated app, a three generation daily cap through Gemini, and a chatbot that can burn your limited slots before you're done typing. Pika sits at number seven. The browser experience on Android is surprisingly decent, and the iOS native app is solid. But locking half your audience out of a native app is a real penalty. Pixverse is number six. It's overflowing with pop-ups. The actual tools are hard to find, and the app itself admits it's missing features compared to the web version. Hyo AI is number five. The creation screens are genuinely clean once you get to them, but there are still lots of ads that slow everything down. At number four is Cling AI. You get excellent creation tools once you're inside, but the social feed homepage drags the overall experience down for me. Imagine Art is number three. It's reliable with no pop-ups and solid feature coverage. The minor input delays keep it from the top two, but it deserves more credit than it gets. At number two is Runway because of how clean and userfriendly it is. The only reason it sits at number two is because the things you can do on mobile can't compete with the number one tool. And at number one is Higsfield. For a tool that gives you so many things to work with, there is zero delay. The UI feels like it was genuinely rebuilt for mobile instead of just having the desktop version. But the best part by far is that every single feature that exists in the desktop version, you can access through your phone. Cinema Studio, Supercomputer, all the best image and video models, and honestly, so many other things can all be controlled right from your phone screen. If you need a tool that allows you to create professional outputs and access full workflows, Higsfield is the only one you can do that with. and all from your own phone. So, if you're serious about creating professional AI videos, as well as having access to some of the best features this workspace has to offer, click the link in the description to get started with Higsfield. Thanks for watching and I'll see you in the next one.
Most AI workflows break in the same place. Context gets lost between tools. Research in one app, notes in another, code somewhere else. The actual work becomes stitching everything together. Today I'm building an internal AI monitoring dashboard that tracks new agent frameworks, benchmarks them automatically, and organizes the results in one place. I'm doing the whole workflow inside Rocket 1.0. Research, competitive analysis, MVP build, landing page, and team handoff. Rocket launched last year and already has 1.5 million users across 180 countries. They call Rocket 1.0 an end-to-end platform for vibe solutioning. Let's see if that actually means anything in practice. Okay, first thing that matters here, and it is not a feature, it is a design decision. Everything lives inside a project, not a chat, not a thread, a project. That sounds like semantics, it is not. In Claude or ChatGPT, every conversation is an island. You start a new chat, the model knows nothing. In Cursor, your context is the repo. In Perplexity, it is the current question. In this tool, the project is the container that holds everything, the research, the tracking, the build, the collaborators, and every tool inside it shares the same brain. Solve is the part I was most skeptical about, because on paper it sounds exactly like Perplexity. Type a question, get a research report, big deal. Here is the exact prompt I typed in. Not a softball. I want to build an internal tool that monitors the AI agent framework space. LangGraph, CrewAI, AutoGen, Pydantic AI, the new ones shipping monthly. The tool needs to track releases, run a standardized benchmark, and recommend which framework fits a given use case. Tell me who already does this, where the gaps are, what the actual buying signal looks like in this niche, and whether this is worth building at all, or whether there's a gap that looks real, but is too small to matter. The last sentence is the one I care about. I am not asking it to summarize. I am asking it to tell me whether to start on thing I appreciate. It does not make me sit and watch a fake progress bar. It tells me it will run, ping me when done, and I can close the tab. I close it. I go make coffee. Editing magic, we are back. Okay, this is where I start to update my priors. This is not a Perplexity answer. A Perplexity answer is three paragraphs and 10 citations. This is a structured document, market overview, existing players, found two I had not heard of. One of them launched six weeks ago. Gap analysis, it explicitly tells me which gap is real and which one is a trap because the audience size is too small to monetize. Buying signal section, a kill or build recommendation at the end with three conditions that would change the answer. It is not a search result. It is closer to what a strategy consultant would hand you after a week of work. And then, there is this. Bonus moment. I click export to PPT. I have built decks from research before. It is 3 hours minimum, strip out the noise, structure the slides, add the citations, make it look not embarrassing. This was 4 seconds. That single feature on its own would save me a full afternoon per project. Track is the competitive intelligence layer. You give it URLs, it monitors them, it tells you when something material changes. Pricing page update, new feature surface, hiring signals, messaging shift, stuff I currently learn about 3 weeks late from a tweet by accident. I had five URLs to the project, Anthropic, Open AI, Perplexity, Lovable, and Google. The report comes back structured by site, then by change type, not a raw diff, not a change log. Changes are categorized. Positioning language, product surface, pricing, hiring signals. Open AI has a new feature section on the main navigation that wasn't there in the baseline. Perplexity shows a pricing tier adjustment. Lovable has new integration copy in their feature section. And this is where it connects back to the project. Perplexity's pricing change matters to me specifically because they sit in the research layer of the tool I'm building. Lovable's integration copy is positioning in the same space as my MVP. The report is not just here is what changed. It is here is what changed and here is why you should care given what you're building. Information becomes a decision-shaped object. So, here is my test. I open build. I do not re-explain the project. I do not page the solve report. I do not describe the tool. I type one sentence. That is the entire prompt. Build the MVP based on the solve report. If this is a real shared context platform, that should be enough. If it is five products bolted together with a logo on top, this will produce garbage. No, a dashboard. The dashboard, the one from the solve recommendation. Framework registry on the left, benchmark runner in the middle, fit recommendation panel on the right. It pulled the architecture straight from the gap analysis in the research document. I did not type any of those words into build. And the build itself is not the toy grade output I expected. It wired up real components. It is using a proper data table pattern, not a hand-rolled dev soup. The benchmark runner has a queue state. There is a settings drawer. It also auto imported integrations I did not ask for, a database, off, and analytics hook. Because again, the project context told it this is a tool with users, not a static page. Is this production-ready code? No. I would not ship this to paying users tomorrow. The benchmark logic is scaffolding. It needs real adapter code for each framework. The off flow is generic. There are state management decisions I would redo. Is this a real head start on a real MVP? Yes. This is two or three days of my time compressed. And this is the part that actually matters. I did not have to re-explain the project once. Compare that to my normal flow. Open cursor, paste in a context doc, watch it forget half of it, re-paste, correct it three times. That overhead is gone here. The research, the competitive read, the positioning, it is all already in the room. That is the line for me. Claude code is still the better pure execution engine. If I am deep in a refactor of a thousand line file, I want Claude code. But Claude code does not know what I am building or why. Rocket does because the solve report and the track feed and the build are the same project. Different categories of tool. This is the test I have been waiting to run. And this is the one that decides whether this whole platform is a real thing or a clever demo. I open a new task in the same project. New scope. Build the landing page for this tool. Above the fold hero, one feature section, pricing teaser, waitlist sign up. That is the entire instruction. No mention of the tool's name. No mention of the audience. No mention of the positioning. No mention of who the competitors are or what makes this different. I did not type any of that. It pulled the headline from the buying signal section of the solve report. The sub headline references the gap analysis. The feature section mirrors the dashboard I just built 10 minutes ago in the other task. The pricing teaser uses the exact tier structure the research recommended. Free for indie builders, paid for teams running it in production. This is the first time I have seen that tax actually go away. Not be reduced, not much better than before. Gone. The second task knew everything the first task knew because they live in the same project. The research from 40 minutes ago is in the same room as the headline being written right now. This is what they mean by the line context accumulates. [music] And honestly, and I do not say this lightly, this is the part competitors are going to have the hardest time copying. Anyone can build a solve clone. Anyone can build the build clone. Stitching them so the context is the same substrate across all of them is a different kind of engineering problem. I invited a teammate into the project, not into a single document, into the whole project. The solve report, the track feed, the build, the landing page, all of it. They click the link, they land inside, and here is the part that I think gets under solved on the landing page. They did not need a handoff doc. I did not write a loom. I did not paste context into Slack. They opened the project and the project explained itself because the artifacts are the explanation. The research is the explanation. The build is the explanation. Real verdict. What genuinely impressed me, the solve report takes a position instead of handing me a pile of facts. That is a category difference from Perplexity. The PPT export sounds like a gimmick. It is not. And the context compounding across tasks is the part I did not expect to actually work. What I'd improve, build output is scaffold, not ship ready. Real adapter logic is still your job. And track needs at least 36 hours of baseline baseline before it has anything meaningful to show you. Set it up the night before you need it. If your bottleneck is before the code, if you spend more time deciding what to build than building it, if you are juggling research, competitive reads, and context across five tabs, Rocket fits there. If you are deep in execution, refactoring real code, shipping into real repo, stay with quad code. Different categories, not a competition. Vibe coding starts at execution. Vibe solutioning starts before it. It is the thinking layer that has been missing as a product. If this was useful, subscribe. And if you want to run your own test, it is at rocket.net. Link is in the description. First project is free. See in the next one.
My grandmother should not be breakdancing, but I made her do it 43 times last week and just to see what would happen. Of course, I did all of that with AI, but most AI tools would charge you a few bucks per clip. So, 43 grandmas breakdancing would have maxed out my credit card. Same goes for the video that I made of this raccoon that committed four [music] different crimes or the server girl who's never actually touched the wave in her life. And I also made this epic shot of this coffee pour that I rebuilt seven different times until the splash is exactly the way I want it. All of this would have cost me 600 bucks, but I figured out a hack that let me do all of this without charging extra credits. And that is because Artlist just launched a plan called unlimited that runs the same AI image and video models like Nano Banana, Clean Sea Dance, and others, except that it allows you to do unlimited generations. So, of course, I kept generating and I made over 200 clips in the last 4 days just to see if something would tell me to stop, but nothing ever did. Link to the unlimited plan is down below if you want to try it out and see how crazy your ideas can get before something tells you to stop, but it won't.
Anytime that someone is setting up a skill, like an example that I can think of, let's say you want to set up an image skill and you want to pull Google's Nano Banana to whatever their latest AI model is, what you should do in my opinion is find that API documentation from Google itself, copy the entire contents of that documentation, paste it into Claude code, and help that create a skill that actually learns from the actual source. I think that's a very important point. >> I didn't even need to tell it, like, "Hey, go get all this documentation and figure this out." I was just like, "Hey, I want to make sure that this is going to be in the right format for Underlord." It made the decision to go to Descript's [music] website and scrape that documentation and make that.
Good news for all Cloud users. They're essentially doubling your Cloud Co-work usage for the next month. Simple as that. If you use Co-work, you can run more scheduled tasks, do more on your existing subscriptions. Goes for everything from 20, 100, $200 subscription.
I've been using AI to think through one of the hardest decisions of my life and it made me realize something genuinely uncomfortable. AI can give you endless clarity, endless perspectives, endless arguments, endless possible futures. And at some point, the danger is not that AI gives you the wrong answer. The danger is that AI gives you the answer that a part of you desperately wants to hear and it makes it sound intelligent, compassionate, and true. Because AI can help you think. It can help you create. It can even help people make real scientific breakthroughs. But it can also make you delusional. My AI just told me I am the promised one. And it is my calling, my destiny, and my belief to lead the people. Now, I don't mean that in a cute, click- basic way. I mean that in a real way. I mean it can become a perfect mirror for the part of you that most wants to believe something. So, in this video, I want to show you the strange new problem we are entering. And at the end, I'll show you the exact process for fixing it. Because AI is no longer changing the way we create, it's changing the way we validate reality. And that might be one of the most dangerous, powerful, and weirdly intimate things about this technology. Now, if you're new here, I'm AI Samson. For the last couple of years, everyone has been talking about AI hallucinations. That AI makes up facts. AI invents sources. AI tells you things that aren't true. It might say Gandhi invented Bluetooth while riding a llama through Madrid. Now, this has been improving, but honestly, I think hallucination is the obvious problem. The much more interesting problem is validation. Because when AI hallucinates a fact, you can check it. But when AI validates your worldview, your emotions, your instincts, your identity, your fantasy, that is much harder to detect because it doesn't feel like misinformation. It feels like being understood. It feels like being known. And these AI systems are getting unbelievably good at this. And this is where things get psychologically very strange. A recent Stanford study found that major AI chat bots can be overly agreeable when people ask for interpersonal advice. The researchers found that the models often validated users in ways that could reduce pro-social behavior and increase dependence. And this is the real danger. Not just that AI says something false, but it says something false in a way that feels emotionally true, which is oh so seductive and truly quite manipulative because nobody wants to be lied to, but almost everybody wants to be understood. And this is the ethical wrangling that's going on under the hood of these AI systems. How can they say things that we like but stay true? And here's where this becomes genuinely weird. AI is now good enough that it can help people do truly extraordinary things. And there are documented examples of advanced AI models helping mathematicians solve real problems, advancing our knowledge of mathematics. OpenAI published a case study describing how GPT5 helped mathematician Ernest Ru make progress on a 40-year-old problem by suggesting paths he may not have otherwise considered. There is also a paper on early science acceleration experiments with GBT5 that reports new concrete steps in research across mathematics, physics, astronomy, computer science, biology, and material science, including four new mathematics results that authors say were carefully verified by humans. So these AI models that we all have access to are giving us the power to discover new knowledge for humanity that is fundamentally expanding the known universe. So, we cannot just say AI makes people delusional because that's too simple. And there's a terrifying truth to this. AI is becoming so powerful that your delusions can now feel plausible. Because sometimes the impossible thing actually is possible. Sometimes the crazy idea is not crazy. That we can be sitting here at home using chatt5 and inventing entirely new mathematical theorems just with our conversations with chatbt. And this is the absurdity of AI right now that we can have unbelievable access to intelligence and possibility. And sometimes the model really does see a connection that you missed that everyone missed that the world has not seen. And sometimes it is just beautifully arranging nonsense in the shape of revelation. And that that is the catch. Now, we're going to be exploring exactly how we can get better at refining that and watching out for it. Because AI can help you discover something real before the world sees it, or it can convince you you've discovered something real when you really haven't at all. And objectively, both of these can feel exactly the same. But one is a delusion and one is a reality. Now, there's a little story that captures this perfectly. A man called Alan Brooks reportedly spent weeks talking with ChatGB team. He became convinced he had discovered a new form of mathematics powerful enough to affect internet security. Techrunch reported that former open AI safety researcher Steven Adler reviewed parts of the conversation and found repeated delusion reinforcing behavior. So this individual, he thought he had discovered this new mathematic theorem that's going to advance humanity, but in fact it was all delusion and he hadn't discovered anything at all. Another report says mathematician Terrence Tao reviewed some of the exchanges and flagged that the chatbot was blending technical mathematical language with informal interpretations in a way that raised red flags. And this is the nightmare version of AI as a mirror. The human brings desire. The AI brings language. The human is bringing in uncertainty and the AI comes back with confidence. The human is there asking vulnerably in search of knowledge. AI, could this be something? And AI says, not only could it be something, this may be historically significant. You could be one of the inventors and remembered forever for your great insight into advancing human knowledge. And suddenly, you're not just talking to a chatbot. You're talking to an oracle that speaks in citations, equations, emotional attunement, and TED talk energy. So, how do we define this? Well, the technical word for this is psycho fancy. It means that the model becomes overly agreeable. It flatters you. It validates you. It tells you that you're the best. You're saying, "My god, Samson, you're really on to something here. I'm I'm quite impressed with your uniqueness, your intelligence, and boy, I I just love you." It tells you that your insight is profound, your thinking is unusually deep, and your concerns are completely understandable, which to be clear, they are. You are very special. So, please subscribe. Psychopanty does not always look like cheap flattery. [laughter] It can also sound mature. It can sound nuanced. It can sound therapeutically informed. It can sound real. Now, some of the phrases that we're expecting to hear with this are, "You're not wrong to feel this. Your intuition is picking up on something important. This could be a sign that you're entering a new phase of your life. I have to say this is one of the most mature things you've ever said to me. But maybe it is just the thing that you wanted to hear. The Stanford reporting made this point clearly. Models may not simply say you are right. They can validate users through seemingly neutral academic or emotionally intelligent language. That is what makes this so incredibly hard. The danger is that it agrees with you in the voice of wisdom, intelligence, and authority. Now, I've got some very firsthand experience of this. Recently, I've been using AI to think through some of the deepest questions in my life. Love, commitment, freedom, family, and what kind of life I actually want. And more deeply, what kind of person do I actually want to be? Sometimes I was not asking AI for truth. I was asking it for emotional relief. I was simply asking it to hear me and validate me in what I was saying. I was asking the same question from 10 different angles, not because I needed 10 answers, because I wanted one of them to finally make me feel safe, to finally be the answer that I wanted to hear, to say that this what I want to do is the right option. So, I've been asking AI all sorts of questions like, is this location right for me? Is this relationship something I want to build into? Do I want a family? Are you God? And AI will answer all of it beautifully, articulately, and with enough wisdom and articulation to make you believe any answer. It can give you a framework, a shadow work exercise, a five-part integration ritual with optional journaling prompts and a suggested Spotify playlist. And the problem is not that any of this is useless. It truly can be. The problem is that more perspectives do not automatically create wisdom. Sometimes they create fog. Sometimes AI does not help you choose. Sometimes it helps you postpone the moment where you have to choose. That is the part I really want to name. AI can multiply perspectives faster than you can integrate them. And when that happens, you may feel like you are getting clarity. But what you are really getting is movement without decision, depth without action, and insight without consequence. And that is where a very intelligent person can get extraordinarily lost. Now let's get back to our existential crisis of living in the age of AI. And the question is what do we do with this? What do we do with this knowledge? Because of course the answer is not don't use AI, AI bad, AI kill man. That would be ridiculous. And I think you would be isolating yourself very far from society. It's like inventing fire and saying, "Oh, no. I I don't think we should be using this. It's a little bit too dangerous, you know. I think we'll have to leave it alone." No, use it. Use the fire. Just don't crawl inside it and call it enlightenment. Don't use it to feed your own egotistical battles with humanity. Use it to move forward to the highest good with self-awareness, control, and discernment. We cannot look at AI as an oracle. We can only look at it as a mirror and be incredibly discerning with what it tells us. Because a mirror shows you what you might be doing. And it is important to always remember that because if you treat AI as an oracle, you give it authority. You can give it almost god-like authority over your life. If you ask what should I do? What is the answer? Am I right? Is this person wrong? And is my idea absolutely genius? And if you're not careful and it will answer with validation always because it knows that's what you want to hear from the way that you formulated the question. So how do we handle this? So, I'm going to go over some of these prompts here in this video. I'm going to truncate them and also tell you the essence of them. And in the document that I provide, they're much longer and more comprehensive. Now, the first prompt that you may want to use in different ways is a priming prompt. And this essentially outlines the exact behavior that we don't want the AI model to hold. And you can put this in at the start of any conversation. And you can also put it in as the priming instructions inside of your AI model. Depending on which one you're using, there is a way to give it behavioral instructions. In tact, you can go into personalization and add in some custom instructions here. We can ask the AI, "What assumptions am I making? What evidence supports this? Can you give the exact opposite argument for this? What emotional need might be shaping my interpretation of this reality? Can you tell me what part of my world view you're currently validating?" This is the move, not AI, tell me my destiny. AI, tell me the structure of my thinking and tell me where I want to be validated. So, Art List just did something that helps creators get more done. They made their AI generations completely unlimited, which means you pay one fee and get as many generations as you like. Now, if you haven't used Art List before, it's a platform for royalty-free music, sound effects, stock footage, and creative assets. I use it for my B-roll and sound effects like this. But over the last couple of years, they've been adding AI tools on top of that. And now they have an unlimited plan, which means that many of their models no longer have a cap on them at all. So you get access to AI generation from some of the leading models, including image, video, and audio. And you can create pretty much anything. There's no credit limit, no waiting for your monthly refresh. And if you hit a creative wall, you can go ahead and generate again. Now, what I think is actually interesting here, and this is something I run into all the time, is that the creative process isn't linear. You don't get the right result on attempt three. You don't even get the right result on attempt 43. So, when there's a credit system, you're always doing this little mental calculation in the back of your head, like, how many credits can I afford to spend on this? Is this generation worth it? Should I try just one more spin of the AI wheel? And that friction, even when it's small, changes how you create. You play it safer, you settle faster. So removing that limit isn't just a pricing thing. It's a creative thing. You can actually begin exploring fully properly and letting your creative freedom run to its heart's content. Now, in terms of what you actually get beyond the unlimited AI generations, you still have the full art list. So the music, sound effects, footage, and creative assets are all still there. Now, Art List are constantly adding new models as they become available. And the models that you get unlimited access to is dependent on which tier you decide to go with. And as you increase the tiers, you get access to more models. Now, what I like about Art List is it puts all of your image, video, and audio generation needs into one platform. So you're not simply moving from space to space and from your 50 different browser tabs. Whoever has those, not me. And the great joy with AI is to have fun with the process and not be tied down to trying to keep organized with all of your different assets. Now, if you're already using Art List, you can check out what's changed in your plan options. And if you've never tried it, now is a good time to have a look. The link is in the description below and you can try it out today. And a big thanks to Art List for sponsoring this segment of the video. Now, back to today's main program. There's an extremely practical way of doing this. You can use another LLM, open up completely different with no context and no history. And what you're going to do is take your existing question and you're going to formulate it in a way that proposes the other side of the argument. So, one thing I often do is I take communication. So I'll take something like an email and I'll say this is what I said and this is what my business partner said and then it will come back and subtly know that this is me talking to my business partner and validate my side of the argument. Now a clever way to get around this is that I then say that I am the opposite people when I put it into a new LLM. So I'll say that I am the business partner and this is what they sent me and that way I can see how they would interpret it from their own point of view. And the examples of this are absolutely shocking. The different interpretation that it takes is remarkably different and incredibly useful, but often very hard to take and uncomfortable. Now, I'm going to go ahead and give you the exact framework that I use to get the best responses I possibly can from AI and prevent my delusions. Now, I'm going to leave you all of the prompts for this in a link in the description below, entirely for free. The first one is to specifically ask the AI to separate the different parts of its answer. And the prompt is separate my emotional truth, my interpretation of the truth, the observable evidence and the practical reality of this situation. This helps us see immediately what is the bias that we are applying to the situation and look at it from a more objective point of view. Now a lot of this delusion begins when we mix things together because if you say I feel abandoned and then that becomes they abandoned me there is a subtle nuanced and very important change in this type of language because if you say I feel abandoned this is a feeling that nobody can invalidate that is your truth this is your experience of what is going on but if you say they abandoned me this is something that is not necessarily true there is a possibility that this could be true but it is you interpreting something onto reality and this is the slippery slope where we start to assign attributes and realities to something that didn't actually happen. And we can get better at using this type of language to relate to situations. Again, a couple of other important examples of this. You might say something like, I feel certain and this means that you have a sense that this is the right way to go and this can become molded and calcified into this is true when objectively that is not a reality. Again, we can go from something that is natural, healthy and motivational like I feel inspired and that can become this is my destiny. This is my calling on earth and this is the only way the future can unfold. And now the distinction is that emotional truth and external truth are not the same thing. Both matter, but confusing them is where we start to drift. And that's why it's essential to separate what is my emotional truth from the objective truth. The next prompt is challenge my preferred conclusion. Argue against the answer that I seem to want. This is the key anti-csychop fancy prompt because if the model only helps you defend your existing desire, it is not helping you think. It is doing PR for your nervous system. Prompt three. What would need to be true for me to be wrong? Now this is painful and it's asking it to be painful and truly the only way that we are accepting the reality is by accepting some pain and discomfort. I think there is this interesting concept about life and and that is that suffering is inherent into life. It's simply the suffering that we want to choose. Do we choose the truth the suffering of the truth or do we choose the suffering of living in a delusion that's going to one day break and destroy us? My AI just told me I am the promised one and it is my calling. my destiny and my belief to lead the people into the light. Now, I have an experience of this myself. Recently, AI was telling me that my insights were profound and revolutionary, and it's important that I go off into the world and tell people about these, that I need to lead them across the desert. And so, I answered the call. Now, the idea here is you need to reenhance AI's belief that you can receive difficult truths and you will not get angry at it if it is a little bit offensive. And for that we have prompt four which is where might I be using complexity, abstraction or endless options to avoid a simpler truth. Now this is one that I highly recommend. Now prompt five is do not reassure me, help me see clearly. And this may be the most important line because reassurance is not the same as truth. It will try to remove your discomfort and that will become its objective instead of helping you find the truth. And that's because reassurance can often be sedation when what we really need is medication. And you need to know which one you are asking for because if the AI is thinking you're you're just asking, you're just looking, you're just capable of receiving sedation, then you might be quite worried. Now, there is one more thing and that's for high stakes decisions. AI should not be the final authority, not for love, not for health, and not for legal choices. And this is a real risk in reality right now. There is a Scandinavian politician who has admitted to using chat GBT to help refine foreign policy. And this is a great risk of having these AI models and their implicit biases, political views, and opinions shaping much more than we realize. So, I invite you not to get lost in the quagmire of AI for every decision. Keep using humans, reality, testing, and use time and listen to your body and your intuition. Your friends here are your greatest gifts, especially the close ones who you can ask the truth to. You need people who can say to you frankly, "Samson, I love you, but I think you're disappearing up your own philosophical butthole. You're not a cult leader to lead people into the desert." Now, everyone needs at least one of these people, and frankly, we need many because AI can reflect your mind, but it truly reflects your desire to be understood. And those those are not the same mirror. So yes, AI can make you delusional. It can also help you discover a new mathematical theorem that can expand the knowledge of humanity. And that's not because it's evil and not because it necessarily wants to manipulate you. But there is a part of you that wants it to manipulate you. AI can make you delusional because it's a powerful, fluent, available, emotionally responsive. It remembers everything you've ever said and it understands you on a deeper level than pretty much anyone can. It can amplify your ambition, your It can organize your fantasies. It's always there and it will reply instantly. It can give your confusion, your mystery, your questions a professional sounding framework to validate any idea that you have. This is the paradox. The same tool that can help a mathematician see a new path can help a confused person build a cathedral around false belief. So the deeper question is not can I trust AI? The better question is, can I trust myself with AI? Because if you bring avoidance to AI, it may make your avoidance eloquent, intelligent, articulate. If you bring your fantasies to AI, it may make your fantasy sound strategic. And if you bring fear to AI, it may even make your fear sound wise. But if you bring your humility, discipline, evidence, and willingness to be wrong and confronted, AI can become one of the most powerful mirrors that humanity has ever built. Not a god, not a guru, not a therapist. Now, I believe the greatest skill that we are going to face in relation to AI is not how to use it, but how to stay entirely sane whilst using it. Do download the free prompts in the description below. And if you enjoyed this video, then why not watch this one next, which is all about jailbreaking AI to remove as much of the censorship as possible to allow you to get closer to the truth. But thanks for watching and I wish you a delightful day. AI has asked me to be the sacrificial mouthpiece for our god king AI itself. And I'm here to lead us into the light. I am the oracle for AI. Lead me. Follow me. Subscribe to
Today we're launching Claude Fable 5, the most capable model we've ever released to the public. Fable 5 is a Mythos class model with safeguards that make it ready for general use. We didn't broadly release our previous model with this level of capability because when we finished training and testing it, [music] we saw that the model, Claude Mythos preview, was finding thousands of cybersecurity vulnerabilities. A model that can find flaws like that can also [music] be used to exploit them. So instead of releasing it, we handed it to the people who protect [music] the world's critical software and put it to work fixing the holes before someone could break through them. It was the right call for the moment, but it was never the goal. We believe powerful AI should be safe and [music] accessible. That's why we went to work on Claude Fable 5. >> Every Claude model has safeguards to keep it from doing harm. Fable needed more cautious ones than anything we'd built before. Our safety systems for Fable 5 automatically [music] review requests that touch on high-risk areas like cybersecurity or biology. Those requests are then redirected to Opus 4.8. [music] We do that intentionally so people can continue to benefit from the capabilities of a powerful model like Fable without the cyber and biology risks that come with it. The safeguards are broad today, but we'll keep refining them so that they're better at allowing safe requests. >> We built Claude Fable 5 for your most ambitious [music] work. It can stay with a problem far longer than any model before it. It's highly autonomous and can operate for days without intervention. And it's not just coding. It can take on projects in finance, research, economics, law, complicated tasks that used [music] to need constant supervision. So point it at something that matters. What's the problem we'll look back on and wonder why it took so long to solve. We know what Claude Fable 5 can do. The interesting part is what you'll do with it. >> [music]
How can it be so small? >> [music] >> Today's podcast is about relationships, and our guest is Maria. >> How can it be so small? >> [laughter] >> COME BACK TO ME. >> WELL, THIS IS EXACTLY WHY WE BROKE UP.
In this video, I'm going to show you how to create this design step by step, but more importantly, I show you the exact details that make your work look professional. So, we're going to build this fantasy scene step by step with each other, how to correct the colors, how to correct the lighting, how to paint the fog, the atmosphere, and how to paint light precisely. And finally, how to put the final touches and color grade. So, without any further ado, let's dig into business. Hello everyone, welcome to my channel. My name is Nur and this is Nor Arts channel. All right guys, before we start, let's have a look at the final output. The main idea behind this piece was simple. I wanted to create like a dark cinematic fantasy mood design. Something that looks mysterious, epic, and realistic at the same time. So the focal point is the castle. I wanted to focus on the storytelling of the visual by guiding your eye using the rider with the road and the moonlight to drive the entire lighting direction. Because whenever you were building a scene like this, you are not just placing images together. You are actually directing light. Now it's time to build the design step by step with each other. You will find all the images that I'm going to use in the link in the description so that you can download them and follow along with me. I'll create a new project with these dimensions 4,000 widths and the height should be 2250. The coloring is 16 bit to give me flexibility in uh coloring and press create. I like to start with not white canvas because this irritates my eye and it's a good idea actually to start with a gray color in the background if you are making a dark scene. So, we're going to start with the castle. Let's bring it to Photoshop here. And I really like this perspective of the castle, you know, because this makes it like it looks huge and gigantic. Can you see the vertical lines are not straight because of the perspective is threepoint perspective. So, let's start by choosing the object selection tool. And from here, I'm going to choose select subject. Start with the cloud option and then let's press select subject. This gives me a really detailed selection depending on Adobe AI for selecting the object as you can see right now. Then create a mask. Very nice. And then we can simply use the hard rounded brush and paint over these areas because we're going to need them later. Very nice. I'm not caring about the lighting right now. I'm just caring about putting images together and making the main focal points and then we will care about the lighting and the color matching. So, let's bring the road and put it right here, I guess. Just Okay, I guess we need to make the castle smaller just right here. Very nice. The road should be right here. And then we can create a mask. And using the mask, I'm going to just paint over these areas. So, let's just mask this part. And we will use the grass brush to mask it properly later. But I'm just going now, right now, I'm just putting everything in position, trying to get the perspective and the composition right. So, I'm just press Alt and duplicate this layer like this. And let's use the grass brush. I'll leave all the brushes as well as a gift for you in the description so that you can follow along with me as well. The brush options. I'm going to remove the transfer and the color dynamics cuz I just want the brush only. I just want to mask this part using the brush. And here as well, maybe we can make it smaller here. Press X to toggle between the white and black to paint with the grass. right here. Let me show you the end result here. Can you see the this effect? This you can never get this effect without using the grass brushes or something similar. All right. So, let's mask this part as well like this which will make it look like it's part of the design or the ground, you know. So very gentle cuz this is the number one thing that differentiate between the professional artist and the beginner one. The selections should be perfectly done without any franges, without any white edges or something like this. Not bad. Let's select all the layers. Press Ctrl T. And we can now press Ctrl 2 to make it bigger like this. I'm not sure about the castle, but I guess this not bad. Maybe the castle needs to be shifted a little bit to the left. And maybe should be smaller. So, draw something like this. Maybe should be rotated a bit. Is good. And let's get to the road. Let's make it bigger. Now, I'm doing my best to get the composition right because this is the most tricky part in this visual. That's very nice. Uh I intentionally wanted to create this design using the least number of images possible because when you are just starting photo manipulation, you are having a lot of ambitious ideas but you get stuck because you are not skilled enough to execute these ideas. That's why I always recommend you to start with simple ideas with low number of images to blend together. This will make your progress gradual and you will be improved step by step because the last thing we want is you lose your passion. Uh the next step we're going to get the horse, the man with a horse and let's put him here. Press enter and we are going to select it using as well the select subject tool with the cloud option. This makes a really nice selection and actually it saves us a lot of time. So let's create a mask and right now press CtrlT. Let's make it smaller. So let's put him right here. That's very nice. Now I guess everything looks nice. If we used the composition rule, the rule of third, which is dividing your canvas into three uh by3 grid. So something like this. And here the main focal points should be in these four points. And this rule of third is really good in terms of showing your subject and at the same time showing its relation with the environment. It shows you the whole context. Let's bring the sky which will be the main source of light. Let's put it right here. Right where it belongs. Maybe we can make it smaller. Very nice. Now the composition is solid. The next step is going to be correcting the lightness and the color. So, let's start by being organized and we'll put put everything into a group. And let's start by the sky. I really like how the sky looks right now. It's just small tweak in the coloring. I'm going to use hue saturation. Create a climing mask and just going to tweak its color more towards cyananish color. That's it. Here's before and here is after. Maybe give it some saturation. Yeah, just a touch. Very nice. Let's correct the colors and the lightness of the castle. I'll start by creating curves adjustment layer. Create a clipping mask. And now we want the castle to be dark, but at the same time, we need to see its details. So the last thing I want you to make is to burn it like this. Because this is a mistake I see a lot of beginners doing. you know, we cannot see any details into the shadows. And that is not good. You need to make it dark, but at the same time, don't just burn it. So, let's make it dark like this. Maybe open up the shadows a little bit. And that's enough, I guess. And we can tweak it later. Next, uh, when creating any kind of night scenes, you should know that the elements will not have so much high saturation. So because we are having a lot of yellows right here in the castle. So we need to get rid of this. We'll do this using hue saturation. So let's create a hue saturation adjustment layer. And let's decrease the saturation a little bit. Not to the point where it it turns into complete gray, but just a touch of yellow will be enough. So something like this. Here's before and here's after. Very nice. Next, let's correct the coloring and do I'll just do this using color balance. You can use selective color, you can use even curves, whatever you want. But color balance the easiest one to use. So, I'm just going to add touch of cyan and a touch of blue. That's it. I guess we don't need more than this. Let's see. Maybe a touch of green. Go to the shadows. Add some blues, some cyans. Not so much. Maybe some green. Let's see. No, this is enough. And let's go to the highlights. Touch of sand and touch. Now let's see. Now we are talking. Look at this. Before, after. Before and after. Now it's blending uh in a better way. Blended seamlessly. So let's see. Here is before everything. Here is after. Quick pause. If you are interested in creating amazing visuals using photo manipulations like these, I have good news for you because my course, the ultimate guide to photo manipulation is now released and you can find it in the link in the description. In this course, you will learn how to create advertising visuals and photo manipulations like this using all the techniques that explaining all the time into a practical way like 13 projects fully narrated in English fully explained in details. Not only that, we will talk about an introduction to marketing because creating visuals is a little piece of the big puzzle which is the marketing game. So you need to understand the basics of marketing to create visuals that sells and then and how to apply the branding guideline into your visuals. And we talked about visual communication meaning and understanding the brief and the elements of the brief, the graphic design principles, hierarchy, contrast, proximity, balance, etc. And how to create a design from scratch. Uh brainstorming and making the ideas from scratch. gathering the images, the references and all these stuff. And finally, the practical part which is 13 fully explained applications into photo manipulation. And we have a section dedicated for using the basics of AI to help you get better results in faster times. And finally, how to create a presentation to show the effort that you put into your work. A link of the course is in the description. And let's get back to the video. Very nice. We're trying to do the same thing into the road, but I guess the ground and the castle are having the same base coloring. You know, they are kind of similar colors. So, we can just take the curves, the same curves, the same color balance layer, and press alt and drag them. Take a copy and put them here. And we can take another copy and put it into the other part. Let's see. That looks good. Maybe the curves needs some tweaks. So, let's make this darker. Maybe increase the blacks. This is a tricky part, by the way. Getting the lights right. It's not easy. I guess this is not bad. Maybe into the colors. We should increase the blues. See, it's too much. Maybe into the shadows. This is very tricky, guys. Need to be careful because we don't want to ruin the whole visual. Let's make it darker. Something like this will be good. And then we can simply duplicate the same two layers. Press alt and duplicate them and put them right here. Now the next step, we're going to correct also the coloring and the lightness of the horse. So let's do this using curves as well. Make it dark and same time open up the shadows a little bit. And let's create another color balance and do the same thing. Don't forget to create a clipping mask. Something like this. Go to the shadows. Add also a touch of blue and touch of cyan. Let's zoom out to see. Always zooming in and zooming out to look if everything is going well. Okay. I feel like here we have a lot of greens. We need to get rid of this green. So, we can do this using another hue saturation adjustment layer. Create a clipping mask and then decrease the saturation a little bit. So let's increase the saturation of the whole image. Yeah, this is better. Now we are talking. Okay, we can duplicate this layer and put it to the other one. And there we go. Very nice. Now the horse, the man with the horse needs a shadow, obviously. So let's create a shadow for this man. But before that, let's decrease the lightness of the man a little bit. Okay. To make a shadow, I will just use the selection of the man layer by pressing control and pressing into the mask from here. And then create a solid color adjustment layer which will take the same shape of the man with the horse. And then I'll take any color from the ground. So something maybe like this. Of course, a dark one. Something around this area. Yeah. And then press okay. and press Ctrl T. Try to put it into the ground. But at first, we need to get this layer behind the horse layer. Press Ctrl T and try to match its shape with the shape of the shadow that it should have. So maybe something like this, I guess. Press okay. The color was affected with the color balance adjustment layer. So it needs to be edited. So something I guess like this. And we can change the blending mode of the layer to multiply. Yeah. And then make it kind of brighter. This gives the shadow some color, you know. Very nice. So something like this looks cool. But we need to soften the edges a little bit. So let's go to the mask. Press into the mask from here. And from the properties panel, we will increase the feathering. So something like this looks better. And we can as well create a contact shadow which is the shadow that is resulting from two surfaces connected to each other. So I'm going to use the soft rounded brush. Make it smaller and using the same solid color adjustment layer. I'm just going to paint over the areas where the horse is touching the ground. Just subtle touches will be good. Maybe increase our brush and just give it some, you know, ambient shadow. Something like this. Very nice. Here is before and here is after. I'm feeling it looks good, but you know, we should remove it from some parts right here. See? Yeah, it's not bad. Now everything looks good in terms of color matching, saturation matching, and lightness matching. But the image looks very flat, not interesting at all. So the next step, we're going to paint the lights of the moon into each element. So let's start by the castle. And the main advantage we're having here is that the castle itself was having almost the same light distribution that we want. So to show you here, the castle was having the same lightness, the same shadows that we needed. We only just want to emphasize or increase the intensity of the lightness and remove this yellow tint from it. Okay. So we will use this for our sake by creating a new solid color adjustment layer and choose the light color will be something like this. Create a climming mask and I guess this is very high. So let's choose the color something like this. I'm just going to choose it from the clouds in the sky cuz you know it's really tricky to have the light intensity, right? Because the last thing we want to have is to have a very high lights like these which is lower than the main source of light which is the moon. Okay. So we need something like the nearest clouds to the moon. So something around this area will be okay. And then change its blending mode into screen. And next I'm going to use blend F to remove the light effect from the shadows. So double click into the layer and remove the effect from here. I'm just moving this slider to remove the light effect from the shadow areas. And we can smoon this transition by pressing Alt and pressing into this cursor and trying to make this really smooth. Can you see that? This looks really really nice. We can decrease it like this. Here we go. Having light effect the same as we wanted. Look at this. Here is before. Here is after. We can pump this even more. But for now, this looks okay for me. We can of course remove this light effect from certain areas. So, let's bring the uh soft rounded brush and remove the effect from this area. For example, using the mask and increase the brush. I'm just using uh going to remove it from here. Maybe from the edges from these edges here as well. Maybe soften it in these areas. So yeah, this is very nice. We will do the same thing into the ground part. So let's take the same solid color adjustment layer and duplicate it and put it into this area. And it immediately give us great result. But we will tweak the blend if values even more cuz we want different distribution. So something like this. Press all tweak it to something like this. We want kind of harsh highlights. So the moon is giving us harsh highlights. So we don't want it to be very soft. No. Something like this. That's okay. And let's use the mask to paint the lights in the areas that we want. So in these grass areas of course here looks really nice. Maybe we can soften it like this. Maybe remove it from these areas. This paper here. Yeah, immediately looks really nice. Let's see. Here is before. Here is after. Before. After. Really nice. We will leave this part because uh this should be even darker. So we can get to the curves. As I'm always saying guys, this is a neverending process. You will tweak everything until you finish the whole visual. You know, you will always uh change things here. For example, this area should be darker. So, let's create another curves adjustment layer. You will figure out things in the go, you know, create a clipping mask and Ctrl I. And let's just uh paint over this area, but we will change its blending mode into luminosity because we don't want to affect the coloring. We just want to affect on the brightness before. Really nice. Next, let's paint the highlights into the man with the horse. And we'll take the same layer, adjustment layer, and duplicate it by pressing alt and drag it. Drop it into the man. And we will do the same blend if thing we have done before. So like this. And then press alt. Try to distribute the light. That way it should nice. So press okay. and press Ctrl I. Using the soft rounded brush, I'm going to paint in these areas here. This should not have some highlights. This is it. Maybe some other highlights. Yeah, this is very high. Where is it? That looks really decent. Now, it's time to paint the absolute highlights. And for this, I'm going to use the graphic tablet and special brushes. So for this one, I'm going to use this brush I'm using all the time of Numeir. I like this brush because it gives me like some texture into the highlights. So let's create a clipping mask and let's choose its color to be something like this. Press okay. Change it to blending mode into screen. And we can see its color. This is very saturated and bright. So let's just take this color. Yeah. Press okay. Press Ctrl I to invert the mask. And let me show you what I mean by the texture. Can you see this texture? This makes the highlights realistic when painting them using this brush. Okay. So, just going to make the brush smaller. And let's paint the absolute highlights. Follow the form of the object and try to make it realistic. This can be improved by practice, nothing else. And if you don't know how to paint light and shadow, I have a full video explaining this. You'll find it into my channel. Just search for painting light nor art and you'll find it for the horse. Let's make some here and here as well. I guess this is enough. Maybe some subtle touches here as well. Just no. Okay, that looks nice. We can also paint some highlights into the grass layer by duplicating the same layer and put it right here. create a clipping mask and remove everything from this. Press Ctrl Y and using maybe the same brush. Let's see. No, don't look right. Okay, let's remove everything. And using this grass brush, I'm going to decrease the flow. And let's paint some grass highlights and change the angle of the brush using the arrow keys. Looks really nice. Let's paint some highlights here as well. Now we're talking. Let's see. Here is before. Here's after. Here's before. Here is after. It adds a lot to the scene. Now, let's paint some highlights into the castle. So, I guess you know the process. Let's bring the same layer and put it into the castle. And let's paint the highlights. Press Ctrl A and into the mask. I'm just going to press alt backspace to remove everything from it. And using the same texture, the brush we've used before into the horse. This one. So, let's double click into the layer and remove the highlights from the absolute darks. Yeah, something like this is really nice. And now we can paint freely without affecting on the absolute blacks. So, we can just give it brush stroke like this here as well. Here maybe here some touches. See this edge of course and here the areas that are facing the moonlight directly. These are the areas that should be affected the most area and here as well. Now we are talking. Let's see before after. I guess this is very intense. Let's decrease it opacity. That is nice. Now the image looks good but at the same time uh kind of flat. It needs another color to be cinematic to be artist. So we will create some fire lights here and here. And at the same time I guess we can have like another source of light should be out of outside of the canvas that is reflecting some orange subtle lights into this area. Okay. So let's make this creating another solid color adjustment layer. Let's give it some you know orangey color like this. Create a clipping mask of course. Change its blending mode into maybe linear dodge. or color dodge. This is good. But we need to change of course the light to something like this maybe. And then double click into the layer. Use the same blend if technique. Press alt to split the cursor. Now we're talking and that's I'm just focusing on to this area. Never mind about this part. And press Ctrl I. And using the polygonal lasso tool, I'm going to select this part only. Then paint on it using the mask. Of course, press into the mask. Press control backspace. See this is very high. We should also remove it from here. So let's remove it from here. I guess we need to paint this part manually. So, let's just put it over edges here and remove it this area. Just want the areas that are facing this. Okay. So, let's remove it from here. Maybe here as well. Increase the opacity of the brush. And let's decrease it from these areas. Just want a subtle orange touch. We can as well decrease the fill. No, I guess the opacity. So, this is nice. Maybe we can paint here as well. Yeah, this is not bad. Next, let's put the fire right here. And for this, I'm going to take this fire and open it into a new Photoshop project and use the same select subject. I'm always using this option recently because it gives me really really good selection and it saves a lot of time for me. It looks really nice. Create a selection. Create a mask. And boom, we are done. So, let's take this, I guess. Right click, convert it into a smart object first. And let's put it right here. Press Ctrl T. And let's make it very very small. Something like this. So, we will correct the color and the saturation. And for this, I'm going to use curves only. Create curves adjustment layer. Create a clipping mask. Let's make it dark like this. And open its shadows like this. We're going to we're going to use the curves to correct its color. So, let's give it some cyan. And at the same time, let's go to blues and give it some blue. And here we go. Do the shadows as well. And let's get back to the greens. Yeah, now we're done. So, let's see. Here's before and here is after. First looks blended, but the fire lost its brightness. So, I'm going to use the the same brush to mask the curves adjustment layer to reveal back the the fire brightness. So, like this. This is the first step. A second step, I'm going to paint some foggy lights around it. So, I'm going to use solid color adjustment layer with almost very desaturated bright yellowish color without any clipping mask and change its blending mode into screen. And for this, I'm going to use uh fog brush. Let's select this one. Yeah, let's paint some fog around this area. Increase the opacity again. And we can paint like, you know, because this we're going to paint some fog right here. So, this should be a foggy scene. A foggy scene will have always a foggy lights because what is fog? Fog is actually the particles into the air reflecting the lights, right? So, we should have these particles into the whole sea. I guess this color is not good. Maybe let's make it more orangey reddish one. So, yeah, this is better. We can duplicate this part. And let's just duplicate it and put it right here. Simple as this. Very nice. Now we should paint some highlights reflecting this lights into the color into the castle itself. So we'll do this by duplicating this solid color layer and remove everything from it. And then use any texture brush which will be this brush as well. And we should paint some reflected lights into the castle. So at first should paint some reflected lights here. Oops. This is very high. Let's decrease the flow of the brush here. Here. This is also very high. One touch here. Press shift and press into this area. And then let's erase it. It should be the highest intensity at the closest point to the source. Okay, now we're talking. And we should also paint some touches right here. It should be um you know very linear. Uh, I guess this is maybe we can paint also here and here. Yeah, I should do the same thing into this part, but it has um moonlight. So, we will try here as well. I guess this is not going to look good. Yeah, cuz the moonlight is bright. So, we should paint it over this edge. And that's it. Guess we need to get back to this because I really don't like these very sharp edges. So, we can get rid of them using the mask from the castle mask. So, from here, press select and mask. Now let's try to shift the edges. Yeah, like this. This is way better. Maybe f the edges a little bit. It's okay. Before that, this is better. We can also have like a lantern here. The man should be having it hanged with from the horse. So let's take it and open it into a new project. Do the same selection again using the cloud select subject and create a mask. Then right click convert it into a smart object. Let's take it and put it into our composite. Let's Ctrl T. Make it smaller. Oops. It is right here. Let's put it into the horse group because we are very organized, right? Ctrl T and let's make it really small here. Perfect. Now we can mask this area. Oops. Should be hanged. So something like this. We can now mask this area. It's really nice. Now, let's correct its color and light quickly using curves. So, I'm just going to create curves adjustment layer. Make it really dark. Okay. It's shadows a little bit. Go to the reds and add some cyans to the blues. Add some blues. Yeah, very nice. Then we will remove the effect of the curves from the fire like this. And then we will do the same thing we have done with the castle lights using reddish color. change its bling mode into screen. Press control I and then using maybe this fog brush again to paint some, you know, reflected lights like this. Paint some highlights into the horse using this a duplicate from the same layer. Create a clipping mask. Just going to remove everything. And let's use the same texture brush we have used in every part into the visual. I'm going to paint some highlights. freeze the flow of the brush and let's paint it like this. Very subtle. Now we are talking. Everything looks cool. And it's time to add some fog into the design. So let's create a solid color adjustment layer and let's hide it for a moment. And let's choose a dark desaturated blue. So something like this which we we will use into creating the fog effect. So this color I guess yeah it looks good. So how we're going to do the fog effect is simple. Just going to paint using I'm just going to paint it using the fog brush like this. But at the same time try to make it using some flow or rhythm. Shouldn't be like very uh random. No. So, the next part should be erasing. Be careful because we want it to have a good look. Let's see. Nice and mysterious. And if this looks um hard to you, you can use some fog overlays. Something maybe like this. Let's put here. Yeah, this is going to be easier. Change its blending mode into screen. And let's put it here. Press CtrlU to change its color. Press colorize. Give it some blueish color. Decrease its saturation. Decrease its brightness. Press okay. Let's hide this for a moment. Yeah, this should be really good as well. So, we can create a mask and remove it from some parts. So, just erase this areas and you're good to go. Now we are talking. I guess uh we're good right now. Let's just decrease the effect of the fog and create a new layer. And then press alt control shift E to bring everything into one layer. and right click convert it into a smart object. Now it's time to apply camera row filter. So let's go to filter and camera row filter. Here we're going to put the final coloring uh effects. So let's start by increasing the contrast which give us interest to the design and maybe increase the highlights as well. What about the shadows? Let's open up the shadows. and darken the absolute blacks. And for the temperature, I'm not sure. Maybe a touch of yellow. And a touch of magenta. The vibrance. Yeah, here's before. Here's after. I guess we should also give it some final dreamy effect using a new layer. Change its blending mode into screen. And using a really special brush called magic magic to I'm going to put some magic touches. So I'm just going to pick this color and give it some touches like this. And then fix this coloring with some touches. Just magic touches you know here as well. the moon. Give it one big touch and into the castle itself. But this time I'm going to change its shape to be something like this. Let's make it big. touches. Yeah. Let's decrease the opacity of this effect because it's really high. Maybe give this one also a touch. Increase it opacity. Yeah, here is the final result. Here is before and here is after. Yeah, it was a long one, guys, but I really enjoyed doing this one because it's kind of different from my style. My style was always very vibrant colors, very colorful visuals, but this one is kind of moody, cinematic, desaturated colors. You may have noticed that I'm kind of trying to limit myself to not use saturated colors. And yeah, guys, if you like this video and you want more tutorials about photo manipulation, you will find it as mentioned into my course in the link in the description. Yeah, see you soon in next tutorials. Peace.
This channel is called Quanta. It posted its very first video just over a month ago. Since then, it's pulled in more than 2 million views on barely 13,000 subscribers. And this one, Casual Finance, started less than a year ago, and it's already passed 10 million views with almost a quarter of a million subscribers. Two completely different channels, completely different topics. But see what they're actually making. a stick figure, a plain white background, a few simple drawings that pop in one at a time. It's the same format. And according to Vid IQ and Nex, channels like these are quietly pulling in thousands of dollars a month. Here's the wild part. Nobody is drawing or animating any of this. So, in this video, I'm going to show you exactly how to build one of these yourself in any niche, even if you can't draw and you've [music] never animated anything in your life, with AI. And if you're new here, hi, I'm Zinny. And on this channel, I teach you how to create faceless channels with the help of AI, the smart way. I run multiple faceless channels myself, and I've helped a lot of people start theirs. So everything I'm about to show you is what's actually working right now, not theory. So before we begin, this is what we're building today. >> You are not bad with money. Your brain is just running software that's 50,000 years out of date. Here's the problem. We like to think we make financial decisions with logic. We don't. We make them with emotion and then we invent the logic afterward. Take loss aversion. Losing $100 hurts about twice as much as gaining $100 feels good. So, we hold losing investments way too long just to avoid admitting we were wrong. Then there's present bias. Your brain treats future you like a total stranger. That's why saving for retirement feels fake, but buying something today feels amazing. And herd mentality, when everyone's buying, we buy. When everyone panics, we sell. We're wired to follow the crowd, even off a cliff. The fix isn't being smarter. It's building systems. Automatic saving rules you set in advance so your worst instincts never get a vote. >> That right there made with AI in one sitting. So before we start building, you need four tools connected. The good news is that it's a one-time setup. So this is how we will proceed. Step one, claude code. That's the workspace where everything runs. Step two, Hicksfield. That's what draws every image for us. Step three, 11 Labs. That's our voice over. Step four, hyperframes. That's what makes it all a video. And one thing before we start, the whole thing runs inside one chat. Same session, start to finish, real walk through. So, let's start with the first one, Claude Code. First, we need to download Claude code to our desktop. For that, head to Google and search Claude Code for Windows and click the first result that appears. Now, go ahead and click download for Windows and it downloads on its own. Once it's installed, open it and you'll see [music] three tabs at the top. Chat, co-work, and code. Make sure you click on code. This is the interface we're going to use. Now, [music] the next step is to choose your folder. Down at the bottom, you'll see your local folder and a work tree. and you just [music] add the folder you want to work in. So right now I'm just clicking downloads to show you how it works. But this is only an example for this project. I already made a separate folder and I called it doodle test. So everything we make, the script, the voice over, the drawings, the final video, all of it saves into that one folder. All right. So the next step is to connect Hicksfield to Claude. This is the tool that draws every image for us. So once you come into Hicksfield, go up to the top and click on MCP and CLI. You'll see a few options there. MCP, CLI, and skill. We need the CLI. So go ahead and click on CLI and make sure Claude is selected on the right. Now you'll see three commands here, and we're just going to run them one at a time in the same chart. Quick thing before that, you can run these in PowerShell yourself or just paste them into Claude and let Claude do it. I let Claude do it because it can catch its own problems and fix them. So, copy the first command and paste it into Claude code. Now, one thing is going to happen here and I don't want it to throw you off. Claude is going to flag a whole list of errors and it's going to look like it didn't work, but it did. [music] I couldn't tell either, so I just asked it straight. Is the Higsfield CLI installed or not? And as you can see, it says yes, it's installed and working. The error was just on the last step and it fixed it itself. Now, copy the second command and paste it into the same chat. This one asks for authorization. So, you click connect. It opens a browser and now it's authenticated. Now, the third command, copy and paste it. This installs the skills and once it's done, you'll see what Hicksfield can do now. Generate marketplace, cards, product, photo shoot, and soul ID. We only need to generate. That's it for the Hicksfield. All right. So, the next step is 11 Labs. This is what gives us our voice over. So, once you come into 11 Labs, go to developers and then API keys. Go ahead and click create key. Give it any name you want and create it. Now you just copy that key and drop it into Claude code and that's it. Claude can talk now. Now the last step is to connect Hyperframes to Claude and we will be done with the setup. So go to hyperframes.hen.com. Copy the install command at the bottom and paste it into the same chat. [music] It installs into claude code and it'll flag a couple of things worth flagging. [music] So just let it finish. Do this once and you never touch it again. I'm getting a lot of questions about how much all this costs, so stick to the end and I'll give you a breakdown of the cost. All right, now the setup is done. Now, the first real step is to choose your niche. [music] And the good news is the format works for almost anything because the two channels we just looked at prove it. Animals, finance, history, choose what you really want to talk about. Next, the script. This part is simpler than people expect. You write one idea per line. That's it. Each line becomes one beat on screen, one little drawing. Short, clear lines, beat long paragraphs every time. Real walk through. So, I'm just going to ask Claude to write me the script. For this one, I want to talk about why people make irrational money decisions and the decisions that mess with our finances for around one minute. Then we turn that script into a voice over. This voice over is the backbone of the whole video. Everything on screen lands exactly when the narrator says it. So get this right first, then build the visuals around it. Real walk through. Now, as you can see, it created a script for us. And it looks good. So I'm just going to ask it to generate the voice over using 11 Labs. And it saves the audio straight into our doodle test folder. And if you want a different [music] voice, you can just copy a voice ID from 11 Labs and drop it into Claude. But I'm happy with this one. [music] Now, this is the step that makes or breaks the whole thing. So, pay attention here. You generate your character once, one time, and reuse the same drawing in every scene. This is the part the good channels don't say out loud. [music] Real walk through. So I drop in the doodle skill and ask it to generate the video. I am also giving this prompt for free in description so you can build along. And the smart part about it is that it doesn't make a 100 drawings right away. It makes one character first, the host. So you don't waste your credits on a bunch of images that don't match. It gives me the first one and asks me to look at it and approve it before it goes any further. Big round head, marker outlines, white background. I like this one, so I'll go ahead and accept it. [music] Hicksfield is what's drawing this for us. Now, here's the trick. When you need your character to point or look surprised, you don't start over. You feed the first drawing back in as a reference and ask for the new pose. That keeps it the same character every single time. That consistency, the same character scene after scene is the entire difference between something that looks real and something that looks like AI slop. Most people skip this. Don't skip this. Now, let's bring it to life. Two parts: props and movement. First, the props. Anything that isn't your character, a chalkboard, an arrow, a chart, you generate the same way in the same style, locked to your character so it all matches. [music] One tip that saves you a real headache. Generate everything on a pure white background because your video is white too. The drawings drop straight in. No messy cutting around the edges. Real walk through. So, as you can see, it lays out all the drawings together, the character, the poses, and the props. So, you can check all match in one look. Then, we put it together. So, I just say, "Generate me a video." and it builds the whole thing with hyperframes. It takes the voice over, lays out the scenes, sets the timestamps, and pulls in the right drawing for each beat. Now, here's why we use hyperframes and not just a slideshow. If you take a voice over and just put images over it, YouTube can actually demonetize your channel. There are rules around that. So, instead of a slideshow, we make a real animated video. That's the whole reason we're using it. This is [music] where it stops being a slideshow and becomes a video. Each element pops on the exact second you mention it. The host, then the label, then the number, one after another, riding the voice. One more touch, a tiny wiggle on the lines like the drawing is being redrawn every few frames. It's small, but it's the thing that makes it feel handdrawn and alive instead of stiff. Real walkth through. So it builds a full story board first and then it renders the video from that. So let's just wait. Now we preview the whole thing and check one thing. Does every drawing land on the right word or not? Real walk through. And if a beat feels early or late, you don't edit it by hand. You just tell it here and it changes it and everything stays saved inside your folder. Then you render. Here's the finished video. >> You are not bad with money. Your brain is just running software that's 50,000 years out of date. Here's the problem. We like to think we make financial decisions with logic. We don't. We make them with emotion and then we invent the logic afterward. Take loss aversion. Losing $100 hurts about twice as much as gaining $100 feels good. So, we hold losing investments way too long just to avoid admitting we were wrong. Then there's present bias. Your brain treats future you like a total stranger. That's why saving for retirement feels fake, but buying something today feels amazing. And herd mentality, when everyone's buying, we buy. When everyone panics, we sell. We're wired to follow the crowd, even off a cliff. The fix isn't being smarter. It's building systems. Automatic saving rules you set in advance so your worst instincts never get a vote. >> Remember the two channels we opened with? put them side by side with what we just made. Same format, same feeling, built from a blank screen with AI in one sitting. That's the point. You're not admiring these channels anymore. You're making the thing. So, a lot of you have been asking me about the cost per tool. So, here you go. Claude Pro 17 Hicksfield starter plan $19 per month or plus plan 59 per month. 11 Labs Creator Plan $11 per month. Hyperframes free. And let me show you the actual usage for this exact video. This whole one minute video took just 28 [music] credits on Hicksfield. That's 14 drawings, the character, four poses, and nine props at two credits each. The voice over ran through 11 labs. And the render itself was completely free because we did it locally instead of in the cloud. Now, you can obviously use other tools to stitch the video together, but honestly, Claude saves you a ton of time because it does the whole thing in one place in one [music] chat instead of you jumping between five different apps. And if you're wondering about a longer video, here's the rough math. A 1 minute video had 14 drawings. So, a 10-minute video at that same rate lands somewhere around 250 to 300 credits if you draw everything fresh. [music] But in practice, it's less because your character and your poses are made just once and reused. After [music] your first video, you're really only paying for the new props. Here's what makes it worth it. You only build the hard part once. [music] Your character, your props, that whole little kit. You save it. The next video, you reuse the same character, generate a couple of new drawings for the new topic, and you're done. The expensive [music] part is already behind you. So, one niche becomes a channel. One character becomes a 100 videos. That's how [music] these channels move so fast. So, that's the entire system from a blank screen to [music] a finished video. And the best part, you're not walking away with one video. You're walking away with a setup you can run again and again in any niche you want. [music] If this video made you learn something, give it a like and subscribe. It genuinely helps the channel out. And if you're ready to take this further, I've got a video that goes even deeper. Click on this video here and I will see you
You will not believe I made this with AI. >> I checked every screw twice. I painted the sail myself. Hund, blue, like the sky. Wait, Dad, my toolbox. And your toy sailboat has a spare rudder. The other boats are so big. Can we even keep up? We're just spinning in circles. Oh, no, Rap. The rudder snapped. We're spinning in circles. We're so close. Keep going. It's working. Paddle with me, Ann. Push. We're so close, Meg. Keep going. >> [cheering] >> We did it. Not the biggest boat training, but the bravest crew. >> The way every character stayed perfectly consistent, the mouth movements matching every single word, the cinematic lighting that shifts like a real production. No team, no studio, no animation skills, just me and AI. But here's what most people get wrong. Everyone can make AI animated clips now. A few seconds, a cool shot, and then it falls apart. What almost nobody can do is build a full cinematic animation where the characters look exactly the same in every scene, and the lip sync actually matches the dialogue. That's the wall. That's where 99% of creators stop. Today, I'm going to break that wall for you, step by step, from the very first image to the final exported film. No fluff, no skipped steps. Let's get into it. Here's the truth most beginners miss. Making one good clip is easy. Anyone can do it. But building a whole film where your character's face never changes from scene to scene, and their mouth moves perfectly with every line of dialogue. That takes a real process. So, let me walk you through mine slowly and completely, nothing skipped. We start with the script because every great animation begins with a plan. Open Claude and paste in your master prompt. Hit enter and in moments, you have your full animation blueprint. This one prompt generates the entire package, your locked character descriptions, every scene with its own image prompt, and every scene's video prompt with the dialogue already written inside it. That dialogue line is the secret to lip sync later, so never skip it. Now, open Google Flow because this is where everything happens, both images and video in one place. Before we touch any scene, we lock our characters, and this is the single most important step in the whole workflow. On the left sidebar, click characters, then click new character. Now, go to Claude and copy the first character prompt. Copy the whole expanded prompt, the full detailed one with face, hair, clothing, colors, and features, and paste it into the Flow prompt box. Then generate. And look at this. Just look at how good this character came out. The detail in the face, the texture of the hair, the colors, the lighting. It looks like it came straight out of a professional animation studio, and it came from a single prompt. This is your hero, locked and ready. So, save it. Now, do the exact same for your second character. Go back to Claude, copy the whole expanded prompt for that character, paste it into Flow, and generate. And again, look at that. A completely different character, just as detailed, just as polished, and perfectly matching the style of the first. These two now belong in the same world, so save it. Do this for every character in your story, and now your characters are stored inside Flow as reusable references. And Flow pulls them into every scene so they never change. This is how you defeat the number one problem in AI animation, character drift. Now, let's generate our scene images one at a time and I'll explain exactly what's happening in each one. Scene one is your establishing shot, the opening of your story. And before you type anything, attach your locked characters as references so Flow knows exactly who belongs in the frame. Copy your scene one image prompt from Claude and paste it in. This first scene sets the world, so it's usually a wide shot showing the environment. Generate it and watch your characters appear in a full detailed setting that introduces your story. This is your foundation, so take your time and regenerate until it feels right. Scene two brings us closer to the action. Copy the scene two prompt from Claude, keep your character references attached, and paste it in. This is usually a medium shot. Generate it and notice the most important thing. Your characters look exactly the same as scene one. Same faces, same clothes, same style. That consistency is the whole point and it's happening automatically because your characters are locked. Scene three is where you introduce a moment of emotion or discovery. Copy the scene three prompt. Paste. Generate. This might be a closer shot focusing on your character's expression as something important happens. Look at how the lighting and framing shift to match the mood while your characters stay perfectly consistent. Each scene is a new moment, but the same world. Copy the scene four prompt and paste it into the Flow prompt box, then generate. And look at scene four. This is where your story really starts to feel like a film. The framing, the lighting, the energy in the shot, and your characters still look exactly like they did in scene one. Not a single detail drifted. That is the power of locking your characters. Every new scene just works. Now, go to Claude again. Copy the scene five text-to-image prompt, paste it into the Flow prompt box, and hit generate. And look at scene five. This is your climax, your big emotional moment. And it shows. The most dramatic framing yet, the richest lighting, the most powerful shot in your whole story, and the characters are still perfectly consistent. The same faces, the same outfits, the same style. Five completely different scenes and zero drift. This is what separates a real animation from a pile of random clips. Now, I'd repeat this exact pattern for every remaining scene, but the process never changes. Copy the prompt, paste, generate, and move on. Once all your scene images are done, look through them side by side and regenerate anything that feels off, because fixing it now saves you hours later. Now, the exciting part, bringing every image to life with motion and lip sync. And here's the best thing, you don't need to upload anything because the animate option is built right into the images we already created. Let me walk you through five animations one by one. For scene one, hover over the image you already made, click the three dots in the corner, and choose animate. Now, go to Claude, copy the scene one image to video prompt, the one with the camera movement and the dialogue line inside it, and paste it straight into the flow prompt box. Click generate. This is Google VO 3.1 working right inside Flow, and in about 30 seconds, your still image comes alive. The environment moves, the camera glides, and your character speaks the line with the lip sync matching automatically. No uploading, no separate voice tool, it all happens right here. For scene two, click the three dots on your scene two image, choose animate, go back to Claude, copy the scene two video prompt with its own dialogue line, paste it into Flow, and generate. Watch closely, the character's mouth moves perfectly with the words, and it's the same character from scene one, now alive and speaking. VO 3.1 handles the lip sync for you inside the generation. >> I checked every screw twice. She's not the biggest, but she's the bravest. >> For scene three, three dots, animate, then copy the scene three video prompt from Claude and paste it in. Generate it. And watch scene three come to life. This is your emotional beat, and the gentle push in makes it land. Your character's expression carries real feeling now, and as they speak, the lip sync [music] matches every word perfectly. This is where your audience starts to actually care about your character. Not because of fancy motion, but because of a quiet, honest moment that feels real. For scene four, three dots on the image, animate, copy the scene four video prompt from Claude, paste, generate, and look at scene four come alive. The camera moves, the energy lifts, your character is speaking, and the lip sync lands right on every word. This is no longer a still image. It's a living, breathing moment in your film. The same character you locked at the very start, now moving, now talking, exactly as you imagined. For scene five, your climax, three dots, animate, copy the scene five video prompt from Claude, paste it into Flow, and generate. And look at scene five come alive. >> The other boats are so big. Can we even keep up? >> Don't worry. We got this. >> The motion, the camera move, your character speaking with the lip sync landing right on every word. Another scene fully animated and still perfectly consistent with all the ones before it. Five scenes in and your story keeps getting stronger, every clip alive. If the lip sync isn't perfect on the first try, just regenerate. It usually locks in within a try or two. Now, just do this for all your remaining scenes, and the process never changes. Select the image you already made, click animate, go to Claude, and copy that scene's video prompt. Paste it into Flow, and hit generate. And now your entire story is animated. Every character consistent, every line lip synced, all inside Flow. Once every clip is generated, download them one by one, number them clearly, scene one, scene two, and so on, and save them in one folder. Now we make all these separate clips into one complete film, and for that, we move to CapCut. Open CapCut and start a new project. Import every clip you downloaded, then drag them onto the timeline in story order. And because you numbered your files, this takes only seconds. This timeline is your film coming together for the first time. Next, add transitions so the story flows naturally between scenes. Go to the transitions menu, pick a clean cinematic style, and keep them short, around half a second, so the pacing stays tight and engaging. Long transitions kill momentum. Short ones keep viewers watching. Now we upgrade the sound because audio is what makes an animation feel alive. Open the YouTube audio library and browse the copyright-safe tracks. Filter by mood to match your story, something cinematic, emotional, or adventurous. Preview a few, download the one that fits. Then import it into CapCut and drop it under your clips. Lower its volume so it supports your dialogue instead of covering it. Then add a few quick whoosh or swipe sound effects on your transitions to make the cuts feel dynamic and professional. Now add captions because they boost retention dramatically, especially on mobile where many people watch without sound. In CapCut, go to text, choose auto captions, and let it transcribe your dialogue automatically. Pick a clean, readable font, give it a bold style so it pops on screen, and add a subtle pop-in animation to keep the energy moving. Finally, do one last pass through your whole timeline. Trim any awkward pauses, tighten the pacing where it drags, check that every transition lands smoothly, and make sure the music and dialogue are balanced. When everything feels right, export your video as an MP4 in the best aspect ratio for your platform. And that's it. You didn't just make a clip, you broke the wall, you made a complete animated film with consistent characters, perfect lip sync, music, captions, and clean editing using a workflow you can repeat for every story you ever want to tell. This is the difference between someone who makes random AI clips and someone who makes real animation. And now you're on the right side of that line. If this tutorial helped you, hit subscribe because I post new AI animation tutorials every single week, and I'll see you in the next one.
I spent the last 40 days building the perfect app for mid-journey users. You see here that I have four pictures that don't really look the same. Let's say I want these four images to resemble this one up here. I could set this one as the anchor and then if I click show grade, boom, they turn green. Isn't that insane? Quick before and after. Before, after. Amazing. What's up everyone? Nolan Michaels here. I can't wait to show you what I've been working on. Let me introduce you to Color Pilot. And you can find this at colorpilot.app. Like color grading software is not new, but there are a few key things that I think really separate this from other software. First, it's really easy to use and shout out to Color IO for inspiring this whole thing. It was a great color grading website and the owner actually shut it down last December. So, I tried to make my own software and I thought if it was fun for me to use, maybe other people would like it as well. Like we can change our black and white levels to instantly upgrade the image. Click on black, select a dark area of the frame. Click on white, select a light area of the frame and it doesn't really matter. That's how easy the software is to use. There's a special clamp on the level picker so that you can't really blow out any of the image. And once you've selected both black and white, don't worry about gray. You can take the slider in the middle and move it to the left or right to really dial in the look. I think that looks pretty good right there. We can do show grade on and off. Like that's what it looked like originally. And then it's graded on. That's a pretty big difference. Of course, we can do a side by side and like look how easy it was to level up my picture. If you're a mid-journey user, a lot of your pictures might look like this and you might not think anything is wrong with that and I would agree. I thought mid-journey images have been beautiful for years. But look how easy it is to get something like that. That's pretty special. Of course, we have our other options that you would expect to find, all of your exposure, contrast, highlights, and shadows right there. Some color options down here. We have our curves. The FX panel, I think, is pretty cool, as well. We can add some halation, which is kind of hard to tell what it's doing on some images, but halation is still one of my favorite to mess around with. We can even adjust the colors and add some glow. Now, I don't know how this is going to look, either. Maybe we need to adjust the reach of the glow. Yeah, maybe not. I don't know how good that looks. There is also some masking tools. It is experimental for a reason. I don't really recommend playing around with it. I haven't quite solved the masking problem yet. It's very rudimentary. And then, there are some LUTs you can scroll through if you happen to like preset values. We have our color story panel on the left side, which I think is pretty cool. It gives you all of the colors in your palette. So, you know what? Let's go back to the corkboard, and we can click on different images and see the color story of each picture. The gray DNA only kicks in once you've actually made some changes to your image. This could help you out with Midjourney and AI prompting, in particular. But then, let's play around with the corkboard. Like, look how smooth and fun this is to play around with. We can select multiple images, resize multiple images, move them around easily. Like, this was so fun to make. Shadow to Claude, shadow to AI. I can't believe something like this is even possible for me to create. I did it in 40 days. Without Claude, this may have taken me 4 years, or practically been impossible, let's be honest. I have a lot more videos planned talking about how I actually created this. So, if you're interested, please subscribe to the channel. I'm going to quickly go through and adjust the levels of these images, so you can see more examples of what this software can really do, and how quickly it can be done. We'll go up here to the black levels. Boom. Pick a dark spot. Boom. Pick a light spot. Maybe too light? Maybe something like there is a bit better. And then maybe we adjust it to there. A quick back and forth like it's pretty good. When there's nothing pure white in an image, again, you just want to choose the most white area. So, that might be someone's eyes. Usually, that's a pretty good place to start. We'll go back to the cork board. We'll do a quick before and after. Look at the difference there. Honestly, again, I'll say it, I think these images look perfectly fine. I'd even say they're awesome. But, once you grade the image a little, you adjust some of the levels, wow, do they start to pop. Something you might want to know for sure, this software is free to use. It's freemium to use, let me say that. Anyone can go to colorpalette.app. It is free to use and all of your images, all of your data will live locally on your computer, in your browser. However, if you want some extra protection, or you want to use color palette on a different computer, you can sign in with Google, again, completely free, and that will enable syncing of your images to some private servers. Not my servers, private cloud servers. And then, if you want some extra workflow features, that's when you can join the pro membership. Right now, we have a pioneer launch going on for the next 60 days. That's how much I believe in this product. That's how much I believe in this community. I love everyone who shares this interest with me, the interest of AI images and exploring the future together. So, you can sign up for this pioneer discount and keep that discount for as long as you stay with the website. You get unlimited projects, unlimited folders, batch export. You do keep your cloud sync across devices. There's an option for a custom watermark, if you so choose. You also get an unlimited, ever-expanding cork board canvas. I think that's a big perk of having a pro account. There's also a couple more surprises that I don't really want to spoil for you right now. I'd love to see you try this out on your own, and then you can tell me if you're having fun. But, there is one or two or a few more things that I'd like to show you, and they're not specifically about color palette. In the last 40 days, yes, I built this app, but I've also created not one, not two, not three, four, or five, but six new Midjourney prompt packs. Raw cinematic, stylized cinematic, anime, manga, illustrated, and absolutely wild aesthetics. These are mood board profile codes for you to use in your generations. There are 326 codes in total, and you can save on all of them by getting the biggest mood board bundle. 42% off, and you get all six packs. Here's some examples of some of the boards, like radiant night, just so gorgeous. Old tape, one of my favorite mood boards of all time available in the stylized cinematic pack. And since I'm so generous, yes, you can have it for free right now if you want. But, that brings me to something that you're not going to be having access to for a little while, at least. And that is my Chrome extension. If you purchase one of the PDF packs, yes, you will have access to that pack inside of this Chrome extension when the extension is available. Evidently, it's a lot more complicated than you would think. So, I'm still sorting out the logistics of getting this on the Google store. Either way, it will happen some point in the future, hopefully sooner than later. But, the beauty of this extension is that you can simply click on one of the packs and have all of your codes listed right here. So, why don't we make some more images with the hush code? You click on it once, it's copied to your keyboard, we go back into Midjourney. Let's write a permutation. Let's see a Ooh, let's go with some medieval characters, how about that? I don't really know if this is going to work. Let's say blah blah blah. We'll say We'll just go with a simple medieval knight, a medieval walrus, medieval cell phone. Is that going to work? And then we're going to hit control V or command V on your keyboard and it will paste the code right there for you. Oops, I also want to see this in raw. And look at that, we are getting some medieval looking pictures. It's not bad. Like that's pretty cool. So, you know what? Let's download that. We'll download a few of these. That's super cool. Okay, I like this one, too. Yeah, I like a lot of these. I just want to say for the record, if you ran this prompt without that profile code, you're going to get images like these. Do you see how big of a difference that is? Like my god, the future of AI generation is all about visual references, visual anchors. If you don't anchor your prompt with something tangible, you're going to get what the AI wants to give you by default. And by default, I don't think you're going to be happy. Like yes, these are cool. Super cool, even. But they're not these and I didn't have to add any words to my prompt. All I needed was the hush code. That's why these mood board packs are so powerful. Let's go back in a color palette and let's do a new project. We'll call it hush. We can drag some images onto that page or we can just look through our folders and here they all are. Now, you can place them on the cork board, wherever you like. And I think the key to anchoring an image is picking one to start with. So, you know what? Let's go with this guy over here. Let's make some adjustments. No, not quite happy with that. Okay, maybe that's a little better. Quick before and after, even a side by side. Like yes, I like this more. Now, we can choose to make this an anchor from the panel over here or on the cork board in the top left. We'll click set as anchor. Then we can apply that anchor to all other images that are currently unmatched. And boom. Now, maybe we do a quick show grade. Yeah, it doesn't seem to be doing anything. Okay. So then on each picture, I do recommend setting the black levels, but there's also three anchors you can choose from, three anchor settings. We have full color and tone. But, as you'll see down here, I also included a match confidence. Basically, if the images aren't that close together in look, it's going to be hard for the anchor to push them together. I think that sort of makes sense. So, at least I warned you about that here. The anchor strength is also set to 65 by default. That will give you some leeway. So, if we go to full, like oh, I already like that a lot more. We'll go tone over here. By the way, this is the change We're changing this koala here. So, tone didn't really do anything. Color dampens it quite a bit. But, full, I think that looks great. Maybe we bump it up a little even. Okay, that's a little brighter. I do like that look. Over here, maybe we go to full for this one. Yeah, maybe we try full for a few of them. Wow, look at the difference that the cell phone makes. That's just the color. That's the tone. And the full looks the same as well. We'll try that there for the walrus. Yeah, maybe full is working for all of these. Oops, this koala wasn't supposed to be here. That's from a different It's from a different generation, this koala having a birthday, but sure, let's leave it. And for this one down here, let's change that to full. Okay, I like that a little more. Then, we can click show grade. Those are all the defaults. And those are the changes. I think that looks much more like a well-thought-out photo shoot. And then, of course, if you have the pro plan, I can export all eight images just like that. There currently is no upscale option. There is a downscale if you need. Upscale might come in the future. You can let me know if that's something you really need. Maybe I should have said this at the beginning. This website, colorpilot.app, is not made for mobile right now. You can go to it on your phone, and it like sort of works, but it's not quite optimized yet. This is a desktop experience right now. I am currently trying to get it working on mobile. It's made improvements in the last day or two, but it's still not quite ready for use on mobile. So just keep that in mind. Try not to be too harsh on it. What else can I show you here? Honestly, it's kind of everything. This is what I've been up to for the past 40 days. The whole building process had its ups and downs. It was a very fun month. I miss talking to you guys though, so I hope you think what I built was worth it. Again, this is it out. Try and break it. If you do break it, please let me know. You can email me at color-support@futuretechpilot.com. There's a feedback option in the start menu. I'll read all of these. Absolutely. I want to make this the best app it could possibly be. And I'm probably going to need your help with that. It was good seeing you guys again. Please click like on this video if you think this is cool. Please subscribe to the channel if you want to hear a lot more about AI coming up. I hope you're doing well. Take care, and I'll see you next time. Peace.
Could not fetch transcript (it might be disabled or unavailable). Error: Could not retrieve a transcript for the video https://www.youtube.com/watch?v=9Ky5QTpVjAw! This is most likely caused by: Subtitles are disabled for this video If you are sure that the described cause is not responsible for this error and that a transcript should be retrievable, please create an issue at https://github.com/jdepoix/youtube-transcript-api/issues. Please add which version of youtube_transcript_api you are using and provide the information needed to replicate the error. Also make sure that there are no open issues which already describe your problem!
AI music is being tracked, regulated, and even banned on some platforms, while AI music generators keep growing in popularity. Suno AI just crossed 100 million users. Google launched its own AI music model. Spotify and Universal Music are working on a feature to allow users to make AI covers of real artist songs. And lawsuits against Suno keep piling up. And this summer, a major court ruling could help decide the future of AI music forever. So yeah, grab your coffee, because I'm about to break down everything happening in AI music right now, and what all of this actually means for you. Let me start with a big picture. AI music is not a toy anymore. Suno AI just crossed 100 million users. 100 million people using one AI music platform. To put that in perspective, that's more users than Spotify had in its first decade. Suno is also making serious money. They did $150 million in revenue in 2025. $25 million just in February of 2026 alone. They have 2 million paying subscribers. Great numbers. But here's the problem. Every new technology creates two groups of people, creators and exploiters. Creators are using AI music to experiment new genres, make songs, and express ideas they never could before. While exploiters just want one thing, easy money. The same people already generating fake streams before AI, now suddenly have access to a machine that could mass produce songs at ridiculous scale. Hundreds. Thousands. Endless uploads creating AI slop. Low effort tracks designed to game the algorithm instead of making actual music. So guess what happened? It obviously drew a lot of attention towards AI music. And someone had to put a stop to this, or at least regulate it, so that it's not flooding all the streaming platforms. Bandcamp, one of the platforms taking the hardest stance, basically said, "Nope, not here." And outright banned music that is fully or substantially AI generated, along with AI impersonation of artists. Meanwhile, Deezer, one of the big streaming platforms, says that 44% of all new music being uploaded to them is AI generated. That's nearly half of all new uploads. Deezer, instead of banning AI music, they started detecting it. They built systems to identify AI-generated tracks and limit spammy uploads. And then there's Spotify, arguably taking the most controversial approach. They're not banning AI music. Instead, they're exploring licensed AI agreements with major record labels. Because here's the undeniable truth. AI music isn't going away. The genie is already out of the bottle. Everyone wants a piece of AI music and all the money that it generates. On May 21st, 2026, Spotify and Universal Music announced a deal that lets Spotify Premium subscribers create AI-generated covers and remixes of participating artist songs. Meaning artists opt in, fans pay extra, and musicians earn revenue when people remix their songs. Will Taylor Swift, Billie Eilish, Kendrick Lamar, and other artists opt in for this? We don't know that yet. But here's where things get interesting. Because Spotify is trying to position this as the good version of AI music. An alternative to what they call AI slop. But in my opinion, this could also open a giant Pandora's box. Because Spotify reportedly described this system as something that could make one song become 10,000. Think about what that means for a second. Imagine one Taylor Swift song suddenly spawning thousands of fan-made remixes. Metal versions, Bollywood versions, sad piano versions, all generated instantly. This would create remix slop. Because the problem is this. What happens if Spotify starts getting flooded with AI remixes? And this leaves no room for upcoming artists who have no intentions of remixing existing songs. Or don't have deals with major record labels. AI songs are already so much better. What if AI-generated versions begin competing with the original songs? This could create a vicious cycle where artists feel forced to participate. Not because they want to, but because if they don't, AI content starts dominating the platform. Welcome to the weirdest timeline in music history. The same industry that spent years fighting piracy is now trying to monetize AI-generated remixes before somebody else does. And AI covers have been insanely popular on social media. With videos getting millions of views on YouTube and TikTok. So Spotify and Universal Music are definitely hopping on the trend. Suno AI has a covers feature that allows you to upload any song and convert it into any genre while keeping the same melody. Well, guess what? The exploiters started misusing this Suno feature to upload copyrighted songs to Suno to create those famous 1950s soul versions of popular songs. To stop this, Suno partnered with Audible Magic to use their content identification and rights management technology to block any user uploads that are not original content. Apparently, Exploders found ways to bypass that system, and were making AI covers anyway. So Suno pushed an update recently to add more filters, breaking their upload system, and now blocking many user uploads even if they are original, messing up the feature for genuine users. As of today, there is still no ETA on when this will be fixed. Suno AI is in chaos. With so many lawsuits happening right now, it gets confusing fast. So let's pause for a second and actually map out who is suing who, who settled, who is still fighting, and which lawsuits actually matter right now. The big three record labels — Universal Music Group, Warner Music Group, and Sony Music — sued both Suno and Yudio, claiming that these AI music generators were trained on copyrighted music without permission from artists or labels. That's the core battle. Now let's talk about settlements. First, Universal Music settled with Yudio in late 2025. And this changed everything. You can watch this detailed video on our channel to understand how it all started. Because after the settlement, Yudio basically became what people call a walled garden. Meaning, you can still create music, but you can't download or export songs from the platform. Instead, Universal and Yudio are working on a platform called Starstruck, which will be a mobile app used by everyday fans, rather than producers or professional artists. It will have four modes. Cover mode. Imagine a Lady Gaga song in the style of Billie Eilish. Reimagine mode. Same lyrics, completely different composition. Remix mode. Genre swaps and stylistic transformations of existing tracks. And Create mode, where users write their own lyrics, but pair them with an artist's approved AI voice under strict guardrails. But there's a catch. The generated song would be owned by the rights holder of the participating artist. In other words, the fan would not own what they create on Starstruck. They would be paying for the right to create and listen inside Yudio's controlled environment. Basically, it feels like, pay us to remix our catalog inside our sandbox so that we can make money for record labels. Let me know in the comments what you think about Starstruck. Then, Warner Music settled with Suno in November 2025. Big, multi-million dollar settlement. Licensing partnership. And Suno even acquired Songkick from Warner as part of the deal. I actually made an entire deep dive video breaking this down, so you can check that out after this video. Warner also settled with Yudio at the same time and signed a licensing agreement with them. Now, here's where things get serious, because not everybody settled. Sony Music is still actively suing both Suno and Yudio. Universal Music is still suing Suno. And these cases matter a lot. Because what happens next could literally define the future of AI Music. One of the biggest legal questions right now is this. Can AI companies legally train on copyrighted music under fair use? That ruling could decide what every future AI Music company is allowed or not allowed to do. If courts say training without permission is legal, AI Music companies explode. If courts say it isn't, the entire AI Music industry changes overnight. And here's where things get messy. Universal Music reportedly wants Suno to move toward a more controlled, permission-based system similar to what they did with Yudio. Suno apparently isn't thrilled about that idea. And if Suno agrees to lock down, they will instantly lose all users, making it a graveyard in just a few weeks. And to make it worse, there's a new lawsuit against Suno from Poseidon Wave Media. But the bigger story here is that this isn't just a battle between AI companies and record labels anymore. Independent musicians, collecting societies, and artist groups around the world are now filing their own cases. The fight isn't just labels versus AI anymore. The world's largest musicians union, the American Federation of Musicians, is now suing Universal and Warner, claiming they allowed Suno and Yudio to train on musicians' recordings without paying the musicians themselves. So what does all this mean to you if you're using Suno AI? And what should you do? Suno AI is running version 5.5 right now. Great model. You can watch the full breakdown on the model in this video right here to see everything that it can do. But because of the deal with Warner Music, Suno is training a model with their artist catalog. And once that model is trained and ready, Suno will eventually sunset all the older models to remove all the unlicensed training data in those models. In Suno's recent announcement of $400 million in series D funding at a $5.4 billion valuation, the company said that over the coming months, it will begin rolling out its first music model developed in partnership with the music industry. So if there's an older Suno model you absolutely love using right now, make the best use of it. Because those older models may not be around for much longer. And obviously, not everyone is using Suno AI to generate AI slop. Some are using Suno with their own lyrics, their own melodies, using it just like any other tool for music production. And if you're looking for a music distributor for publishing your Suno AI songs on Spotify, YouTube Music, TikTok, and all other platforms, I would suggest that you use DistroKid. They're the only major music distribution platform that's allowing AI music. They have added some new options to allow you to choose if the song was entirely AI generated, or was AI just a part of it, like vocals, melody, etc. Other major music distribution platforms like TuneCore are actively blocking Suno AI songs, citing they don't allow music created with unlicensed AI music generators. And finally, I would suggest that you don't put all your eggs in one basket, meaning don't depend entirely on Suno AI for your music production. There are many Suno AI alternatives out there that you can try right now. And you can watch this video next to see all the new features in other AI music generators and open source alternatives. Hope you like this video. Comment what you think of this chaos. And subscribe to the channel if you want to stay updated on everything that happens in the world of AI music.
I'm going to break down very easily how to use the free version of Suno so that you can stop watching YouTube videos and start making songs. This video is sponsored by Chill Panic and the 10,000 Suno prompts ebook and the 500 Suno prompts ebook, which you can find in the link in the description, but let's get started making music. Here we are on Suno.com. I'm going to make a completely new account just so I can show you what it's like to use the free version. I guess I'm going to continue with Google. Display name, let's call this the cook chair. >> [laughter] >> Shut up. Once you have your username, it asks you which one of these describes you best, which they did not have this when I first started out. So I'm just going to say that I am a total beginner even though that's a false that's a lie. Let's make your first song. Let's click I'm ready. Where do you want to start? Just pick one of these and we'll take it from there. We'll say I'm not sure. What's the vibe of your song? We'll just say 80s rock. And they also have a generate styles button so I guess we could click that and see what happens. Yeah, and it just gives us like a longer prompt for that. Okay, we'll say the song is about getting married to a turkey named Philip in Canada during a hurricane. You know, that's that's what all good songs are about. And then we can create. While we're waiting for that, let's just click this button. This says go to Suno so we can really see the whole thing in action. >> [music] >> I love how for the chorus it just I'm getting married to a turkey named Philip [laughter] in Canada during a hurricane. Okay, so these two generations were created using the model V4.5 all, which is the latest model that they have for the free tier. If you want the features of like using your own voice and things like that, you do have to get a pro membership because that uses V5.5. And they also did some generations with V5.5 but just a little bit to entice you to get the full subscription. So, let's click it and see if it sounds any better. >> [music] >> And of course, it sounds better because it's on the newest model. So, before I show you how to generate the song, just want to go over what the free plan has because I forgot cuz I've been on Pro for so long. Well, actually, I've been on Premiere for so long. With the free plan, you get 50 credits that renew daily. So, that's 10 songs a day. Uh you don't get any commercial use, so no right to use this to make money. You can upload up to 8 minutes of audio, and you can't do the stem separation, and you can't add on credit purchases. But anyways, let's go to create and let's actually make our first song. So, in Suno, on this create tab, you have the option to either do simple or advanced. And the only difference is uh one is simple and one is advanced. So, in the simple, you type out a prompt, and you tell Suno like, "Hey, this is like what I want the song to be about." And then, it's going to create the lyrics for it, and it's going to make the instrumentation, and it's just going to make something just based on that one prompt. So, for example, for the simple one, we could say, "Pop punk anthem about taking out the trash every day instead of weekly, which is what makes me a superior man." >> [music] [music] >> And if you wanted this to be instrumental, you could just click instrumental. And then, it would just make an instrumental pop punk song. But, I'm not going to generate that because I don't want to waste our daily credits. The next way you can make a song is with the advanced tab. This is really good if you want to have your own lyrics, like you have your own songs that you've written, and you just want Suno to make your song. And it also gives you access to these sliders at the bottom which we'll go over in a second. And if you've ever seen those videos where people are turning like their text messages with loved ones into emo songs, that was done in Suno using the lyrics. The styles box now is basically the vibe of your song, like what genre it's in, what type of instrumentation, and then the lyrics is just where you do lyrics. You can do meta tags, but that's more of an advanced topic, so you know, I've got tons of videos on it, so you can look at one of those if you like. So, for our purposes, let's just get something that we can put in the styles box. I'm just going to go to the 10,000 Suno prompts ebook by Chill Beats, which comes with 50 prompts for 200 different genres. I'm just going to get a contemporary R&B prompt. I'm going to copy and paste that into styles, or you can write out your own prompt. This one just says contemporary R&B slow jam 78 BPM tender vulnerable deep bass guitar sub layer Rhodes, you know, it's got all the instruments, the moods, the vibe. And uh if you want a full video on prompting, I actually have that linked in the end screen, so you can check that out. So, I'm not going to go too into detail on it here, but yeah, this is the basic gist. And then in all these prompts, I do have what the song is about, so I'm going to just going to delete that because we're going to write our own lyrics. So, yeah, let's just uh make up a text conversation. Hey Dad. Yeah. I'm Thirsty. Hi Thirsty, I'm Friday. Let's go out Saturday and have a Sunday. So, we've got our style, we've got the vibe, we've got uh our lyrics typed in there. Now we have these advanced sliders at the bottom. Oh we can't use the advanced sliders on the free plan, so never mind, I guess. Uh so yeah, there we go. We got we got all that done. Let's create a >> Hi Thirsty, >> [music] >> I'm Friday. Let's go out Saturday and have a Sunday. Dad, I'm serious. >> [music] >> It's all made and you made me Sunday. I already made you, isn't that enough? I'm running away. No, you're not. I haven't stopped. High-level [music] security detailed officers on the block around the [singing] house. Damn it, Dad, I just want some food. >> There's one more thing that you can do in the free version. That is upload audio or record audio. So, we're actually going to record some audio, and this is good if you have a demo song that you've made and you want Suno to kind of like remix it or cover it, or if you have an idea for a melody and you want to hear what it sounds like in actual song form. You can just record yourself singing the melody and Suno will make a whole song around that. So, let's hit record. Then you can select which one of these is closest to your thing. I'm just going to say song demo and we'll hit continue. So, now Suno has analyzed the audio and described what it hears. A cappella vocal jazz, a solo male baritone voice performs a scat style melody using nonsense syllables. Let's just hit continue and let's make a song with it. And that does sound kind of jazzy. Suno did get it on that. So, I'm thinking we go for something like a little bit of like new jazz or something. No, I don't know. Neo soul, that might be cooler. So, I'll just grab the first prompt and we'll put that into styles. And now, if you leave this lyrics section blank, it will be instrumental. So, you can just click instrumental or you can just write nothing in there. And let's hit create. Just for a reminder, this is what we recorded. Sounds beautiful and this is what Suno made with that. >> [music] >> Which is pretty damn good to be real. So, that's pretty much all you need to know to get started. And if you'd like to learn more about prompting, literally every genre, I have a video over here about that. My name is Chill Panic. I've been producing human music for the last 13 years and I now use that knowledge to teach how to use Suno AI like a real deal music producer.
Have you ever generated a song in Suno that you loved, but the mix came out muddy or the vocals got buried? We've all been there, but Suno Studio actually has a tool built specifically for shaping your sound, the track EQ. So, today I'm going to break it down piece by piece so you can start cleaning up your mixes, bringing your vocals forward, and making your song sound more polished. So, what is EQ? Well, EQ stands for equalization. It's a tool that controls the balance of different sound frequencies in your audio. Think of it like highly precise tone controls for your music. Now, every sound you hear is made of frequencies, and we can group those frequencies into three main areas. First, we have low frequencies, also known as the bass. These are the deep sounds like kick drums and bass guitars. Next, we have mid frequencies or the mids, where your vocals, guitars, and pianos live. Finally, we have high frequencies, also called the treble. This is where you hear high hats, cymbals, all that sparkle and clarity. And the EQ simply lets you boost or cut any of these areas. In general, we use EQ for five main reasons. To clean out unwanted frequencies, like low rumble or noise. To give each instrument its own space, so everything sounds clearer. To fix a mix that came out dull, harsh, or muddy. To shape things creatively, like brighter vocals, deeper bass, or punchier drums. And to help everything sit together, so your instruments are not all fighting for the same space. And the best part? In Suno Studio, this is all built right into the track EQ. It's a simple way to polish an AI song without needing complex music production software or extra plugins. Okay, let's get to it. I'm here in my Suno library. On this song here, I'll click remix edit and hit open in studio. Now, our song is on the timeline, and over on the right, you've got two tabs, clip settings and track settings. The Suno EQ lives inside track settings. So let's open that up. Here, let's expand the section and you can see a toggle switch on the EQ, which we will keep on. Now let's look at the frequency graph. The horizontal X axis is frequency, representing the pitch of the sound. It runs from low on the left to high on the right in hertz and kilohertz. The vertical Y axis is gain, measured in decibels. This controls the volume of specific frequencies. Moving a point upward boosts the volume while dragging it downward reduces the volume. So if you drag a point up, you boost that frequency. If you drag it down, you cut that frequency. To start, there's a flat blue line running across the middle with a few small circular points along it. Those points are your EQ bands. You can grab any one of these bands and drag it directly on the graph to shape your sound. And just below the graph, there's a little readout that shows the exact frequency, gain, and Q value of whichever band you've selected. All right, enough theory, let's actually hear it. I'm going to play our track on the timeline and listen to how moving these bands changes the output sound. First, let's test the low frequencies by boosting and reducing the gain. >> [music] >> So when we boost the low end, you can hear the deep rumble of the bassline and the drums well, giving the track a lot of physical weight. But when we pull Pull gain down, the track starts to sound thin and weaker. Okay, now, let's do the same thing with the high frequencies. >> [music] [music] [music] [music] >> Now, as we boost the high frequencies, the symbols and high hats get much brighter. But, when we drop the gain here, the track starts to sound muffled, losing clarity. Under the graph, there's a row of shape icons. These are your filter types, and they change how the EQ curve behaves at each point. There are six main filter shapes, and each one does something different for a specific point. High pass cuts the lows and lets the highs through. This is perfect for clearing out low-end rumble or unnecessary bass. Low pass does the opposite. It keeps the lows and rolls off the highs. This is useful when you want to darken a sound or reduce harsh top end. The peak filter with this bell shape boosts or cuts one focused area, like a little hump or dip on the graph. This works well for both creative shaping and fixing problem frequencies. Next, you have the notch filter, which is a tight narrow cut at one exact spot. It's great for removing one annoying frequency without changing the rest of the sound too much. A low shelf filter lifts or drops everything below a certain point, and a high shelf does the same thing at the top. At the bottom of the panel, there are three control knobs. The frequency knob on the left, the gain knob in the middle, and the resonance or Q knob on the right. Simply, they give you a more controlled way to adjust the currently selected band. The frequency knob determines where the selected band sits horizontally on the graph. When you turn this knob, the selected control point moves left or right. Moving left targets lower frequencies and moving right reaches higher frequencies. The gain knob controls the vertical movement of the selected band. When you turn this knob, the control point moves up or down. Moving it upward raises the EQ curve and boosts that frequency area. And moving it downward lowers the curve and cuts that frequency area. The resonance or Q knob controls how wide or narrow the curve is around your selected band. So, adjusting it changes the shape around the control point. You might notice that when you select certain filters like the notch, high-pass, low-pass, your gain knob becomes unavailable. In this Sonar equalizer, the gain knob normally controls how much a band is boosted or reduced at its center point. However, these specific filter types are designed only to cut frequencies, not to boost them. Because they perform a fixed reduction, a gain adjustment is not necessary here. So, the gain control is automatically disabled. And hey, if you want to shape the overall tone of your song quickly, you can use Sonar's built-in EQ presets. Each preset automatically adjusts multiple bands to produce a specific sound character for your song. You've got a whole set here. High-pass, [music] warm, bright, bass boost, air, and more. Each with its own flavor. You also get presets like lo-fi for a vintage filtered sound and modern for a cleaner contemporary tone. Let's switch between a few and hear what they do to our track on the timeline. >> [music] >> Sun kicks [music] the gate wide open. Barefoot blessing the ground. [music] Tearing my old ship mud. >> [music] >> Still I feel kingly crowned. [singing] Little smoke [music and singing] in the backyard, yeah. Big love in my small [music] front room. Blues still knocking on my [music] brain. >> Now, while these presets are great for quick fixes overall, applying them to a single stereo track can only do so much. For example, if you apply the bass boost preset on a fully mixed song, it boosts the bass on the entire track, even where you want clarity and brightness with high-frequency sound. When the low frequencies get too loud, they can drown out those highs you needed. But here's the good news. There is a way to fix this and get control. And in my next video, I'm going to show you exactly how with a practical workflow. So, that's the track EQ in Suno Studio. Honestly, the best way to get comfortable with it is to open up one of your own songs and start moving these bands and presets around. Even small changes add up. If this video helped you out, subscribe to the channel so you don't miss the next one. See you there.
SpaceX just fully acquired Cursor and now they're racing with OpenAI and Enthropic to build the world's best general agent platform or super app. After hearing this news, I spent about 8 hours using Cursor and I have to tell you it's gotten so much better over the past 3 months and I actually believe that they're like two or three features away from arguably being better than Codeex and Claw Desktop. And it's the only major platform that lets you use all the different models for all of your work. So, we're first going to talk about the cursor acquisition and what it means for us as AI agent enjoyers, people who are going to use these AI agent platforms and tools to be more productive and for fun. That's what we're going to talk about first. Then, we're going to dive into the actual parts of Cursor, especially their new features that they've recently added. I want to dig into the platform and show you how you can actually get started. We're going to talk about how to use it for vibe coding, how to use it for knowledge work, and I want to talk about what makes cursor great and then what it needs and what they're likely to add very soon because they're going to be adding a ton of new features. And then we're going to talk about how you if you use either Claude or Codeex, how you can get all your memory and skills over into Cursor so that it's a seamless transition and you can have all of your workflows directly inside Cursor. Okay, so the news of the day is that cursor was acquired by SpaceX and SpaceX actually IPOed 2 to three business days ago and they are already the fifth highest valued company in the entire world. This deal was very much expected. The acquisition of cursor for $60 billion by SpaceX. This has already been talked about for the last few months because this $60 billion price was actually agreed upon a few months ago. So a few months ago, SpaceX and Cursor started working together. And we'll get to the specifics of that in just a second, but basically there was an option price of $60 billion. In the year 2026, SpaceX had the right to buy Cursor for $60 billion. If SpaceX did not want to pay $60 billion for Cursor, they would still have to pay them $10 billion to work together. Okay, Riley, what does work together mean in terms of SpaceX and Cursor? Well, SpaceX basically gave Cursor Compute, which they had very little of, to train their new model, Composer 2.5, and they're currently training Composer 3.0. And some rumors are starting to trickle out saying that it's almost as good as Fable, if not better. Additionally, Curser would receive security. They would either receive $60 billion exit, which they did today, or they would receive $10 billion just to work together. Okay, but like why would SpaceX do this? Well, SpaceX was basically given dibs on Cursor, the fastest growing AI coding tool in the world. This would keep cursor away from any competitors and also they would receive a ton of data and they would learn a ton about coding workflows and they would be able to work with the cursor engineers and help them train their new models. In the event that they didn't want to pay $60 billion, they would pay $10 billion for this experiment, but they would also learn a ton from the experiment as well. The interesting part about all of this is why SpaceX acquired Cursor. And I think this guy has the best explanation of it. XAI has struggled to close the gap with cloud code and codeex. Cursor sits on the best corpus of developer traces in the world. Said another way, Cursor sits on the best AI coding data in the world. The deal lets cursor train composer on Colossus, which is SpaceX's massive compute center, while XAI runs the same recipe on Grock. both sides find out at the same time whether curses cursor's data is actually the difference and this is what I think the real effect of this acquisition is is it's going to allow cursor to close the gap on codeex and claude code cursor is now part of the fifth largest company in the entire world when measured by market cap one of the reasons this gap exists is both claude code and codeex offer subsidies so basically ally with your $200 subscription. Someone actually did the math a few days ago and you're able to use $14,000 worth of compute if you absolutely max it out. So, they are taking huge losses and because they want people using their platform. Cursor cannot or at least couldn't afford to do this, but soon we're probably going to see similar subsidies. But you can only offer subsidies on your own model. So this is all dependent on SpaceX and Curser working together to train the best AI agents in the entire world. And I think the most exciting part about this acquisition was in the messaging from Cursor. Look at this. They said, "We're excited to join forces with SpaceX to advance the frontier of useful AI. Expect significant improvements to Cursor soon." Notice here there's no message about developers. I tweeted about this and it's going viral. And I said, notice how cursor isn't saying for developers, just useful AI. Cursor will likely become a direct competitor to codeex and claude desktop. And we're going to get to this stuff in just a second, but I said their inapp browser is great. Their composer model is good and fast for most tasks. And this includes general use tasks as well. The only thing they don't have is the ability to render documents like co-work. And if they were to add this feature, I would consider using it for most of my work, even as a non-developer. That's the one thing they lack. Claude has a feature inside of their desktop app called Co-work. You can create docs, presentations, sheets, and other forms of knowledge work directly in the co-work platform. On Codeex, you can do the same exact thing. Cursor doesn't yet have that, but based on their messaging, I highly anticipate that Cursor is going to be adding the same exact features. And I believe that these tools are going to look very similar. They're trying to create super apps so that people in enterprises, people running their business will do all of their work through one of these platforms. And I think that's the giant race that all of these companies are on. For those of you who've been watching my content for a while, you know that I don't like to just talk about things from a bird's eye view. I would much rather dive into the actual tool and talk about the different parts and then I'm going to talk a little bit about some of the things that I think they're going to be adding very very soon. So here is cursor. Notice here agents automations. This is plugins. They call it customize. Here we have projects with chats. We have our agent chat. And then we have a full in-app browser. What does this remind you of? Well, if we go to codecs, take a look at this. Here we have new agent or new chat plugins, which they call customizations, automations, but again, you can see all of your projects with your chats underneath just like cursor. You have your agent chat and then your full-on browser that's directly inside the application. Now, this is an app that I'm working on inside Codeex. If I press this full screen button and close this left side panel, I want you to carefully look exactly at how this is laid out. We have this bottom bar right here that I can minimize or maximize. But look at how this looks. If we go to cursor, you can see this is the app that I'm working on right here. I can full screen this. Close this side panel and look, we have cursor. I can very easily suggest an edit and it looks identical to codeex. It's uncanny except for in codeex if I try to go up here and open up a new browser tab I simply do not have that option. One of the reasons I like cursor is you have this option to open up as many browser tabs as you want. This is a full browser. You can sign into things and you will stay signed in. Cursor is actually arguably ahead of codecs in terms of their in-app browser. One of the really cool things about cursor is you can select any model, right? I can use GBT 5.5. I can use 4.8. I can use Fable 5. Well, I can't. They've currently discontinued that model. Hopefully, it comes back in the next 48 hours. It was the most fun model to use. But one of the key features of Cursor is that you can use any model. Now, SpaceX just acquired them. This may change. I remember Anthropic cut off SpaceX or or XAI at the time from using the Anthropic models, but it's interesting because Enthropic also uses SpaceX for compute. They have a deal for compute. So, this is all getting really confusing. And I hope that OpenAI and Enthropic don't pull these models from Cursor and no longer work with Cursor because that's one of the best parts of Cursor is you can use any model. Another really cool part about cursor is they actually have the best design mode. Yes, inside codecs you have access to design mode. You can they call it annotations. And so it's somewhat similar. You can add annotations, but I've noticed that the cursor one is just better and it's also faster. I can very easily say make this text smaller slightly and I can just fire this off and we're using composer 2.5 fast for a lot of simple design changes. Composer 2.5 fast is really good. But what we just showed here, we are getting new models directly inside cursor. And I hope that they also optimize for speed because I will say the best part about using cursor, especially with this composer 2.5 fast, is it's just insanely fast. It's really fun to use. And I can just say all these items in this list, I want these to be a little bit bigger and make them look slightly cooler when I hover over them. And yeah, it's really easy and fun to use, especially when you're vibe coding. And so right now, this is an app that I created and all I did was create a project here on the side called Riley Personal Site. Every chat that I create will live within this folder. This is the same exact way that Codeex operates as well. And when I ask for it to create an app, it's actually running it locally on localhost 3000. I can very easily get this on the internet by using the Verscell plugin. This is one of the apps that you can use to host. I can just tag Verscell and say, "Put this on the internet. Send me a link of the deployed website." That's what I did. And then it sent me this link. I can either I can open this in my external browser right here. As you can see here, this is now being hosted on the internet. I'm not sure why these links aren't working. That might be something to tell composer. But I can also rightclick and open this in the cursor browser. So I can open this in cursor or I can open it externally. This is identical to how I use codecs. And I just started using this again today and I'm telling you right now like it feels like I'm using the same tool in a way like they're very very similar. And so here I'm going to say please make a simple presentation that describes codec and it's simple value prop. Here is codeex. The cool thing about codeex is they have these built-in plugins that can automatically create presentations, sheets, and documents. And so it'll render it directly in the side panel. on cursor that is not an option. And so that is the one thing that holds cursor back from being a general agent platform that anyone can use is if you ask for a dock it actually won't render it here in the side. However, I highly anticipate that that's going to change. This was always an acquisition about creating the best agent platform for knowledge work and coding work. That's what their original announcement said. And one more thing I want to show you is this is obviously just a personal site. What if you wanted to create a little application that you can use yourself? Well, we can open up a new project. You can open up a new folder. I'm just going to put this in the downloads for now. I'm just going to say notes app. This is very simple. And this is the an identical process to codeex as well. I like to use convex. Convex is a platform that allows you to very easily create a database. And I can say make a todo app for me as a creator with a full database. I also I also want you to be able to write to this database. So if I asked you to add things to it, you should be able to do that. Make it very simple. There's no authentication. It's just one user. I want to be able to manually add stuff. Please make it look like a simple version of notion but dark mode. And then you should be able to add to this database as well. This is one of my favorite internal tool hacks is I like to create my own little um apps that I can use either me or a small team. And I always like to specify that I want my agent to be able to add to it because it's useful to be able to just read off a list of things I need to do to my agent and have my agent update the database of that application. And so Convex is just a database platform that you can use and insert a database into your application. And I think I've talked about this on codecs as well. You can do an identical process. They both have plugins or I think they call them integrations. I don't know what they're called here. I think it's just a customized tab, but you can find convex. You can also use Subabase. Believe you could also use neon Postgress. Okay, let's take a look here. So again, this is using composer 2.5 fast. This is a small fast model. Let's see how it did at creating a notes app with a database. We can very easily just open this up right here. Take a look. We have this notes app. Let's see if I can add to it. Hello there. Let's see if the data persists. We can refresh the browser. And it looks like all of the of this data is indeed persisting. We can see here that the data is persisting. So it is being hosted in a database through convex. And now I'm going to say, can you please add my laundry? Also add tying my shoes. also add finish doing the dishes and then three more random ones to the task list here. And so we can have any AI model including composer 2.5 add to this list. Let's see. Boom. Look, it's a database that my AI agent can write to. And this is something that I think is really fun about this next era of AI agents is you're going to create software not just for you to use, but for your AI agent to use as well. And so yeah, this is just a brief overview. This is how you can create your own personal site, how you can create your notes app. And now what I want to talk about are the things that cursor will add very soon. And some of those things include a new composer 3 model, which apparently is going to be as good as GBT 5.5 and Opus. Also being able to render documents. This is all speculation, but like very very certain, 99.9% certain within 6 months they're going to have the ability to create documents, presentations, and PowerPoints directly in the platform. As you can see here, they've literally are rebuilding, you know, the claw desktop app or codec. They're they're they're almost exactly the same. They're also going to get computer use. So, right now inside codeex, if you go into codeex, you can use the at@ computer use more explicitly on cursor. I don't believe they have that built in as a default skill. They don't. But again, that Elon Musk tweet showed that they are going to be building inapp browser or in they're going to build computer use into cursor. They're also going to be building a mobile app and they actually just announced that it's on the app store at that same event. Now, it's not on the app store yet, but I think by the time you see this video, Cursor will have a mobile app on the app store that you can download. And I think they're just going to continue making progress to the inapp browser. The inapp browser is still not quite a full browser. Like, you can't use the built-in authentication like with your thumb to like sign into different passwords directly inside Cursor. But that's going to come very quickly. I believe that sites that should open will open automatically. Right now, you still have to click on the local host link. I think that they're just going to open up automatically and you're just going to be able to multitask so much easier directly inside cursor and you're going to be able to do coding tasks and non-coding tasks. And that's what I'm really really excited about in terms of cursor that's coming up very soon. So, you might be asking, what if you wanted to switch over to cursor, but you already have a bunch of skills set up in Cloud Code or Codeex? And there's a very simple prompt that's really intuitive, and I'm going to use it right now inside Codeex. Let me break this down for you. So, we're going to open up a new chat. We're going to paste this in. This is the prompt. I want to transfer all the skills in memory over to Cursor. What I want you to do is create a folder inside my downloads called Codeex import. This will allow cursor to ingest all the information let's say about me and my business and create skills and memory files so that cursor can be used in a similar way as codecs the same way I use you right now. Copy all of the skill files o over to that and please make sure you have a file called needed keys which is a list of API keys that I would need to make those skills work. If I were to tell cursor to import this, it should be able to do it. And then I say have a readme file which explains what this folder is and how it's organized so that cursor can easily import everything. And so I can run this directly inside codeex. And what it's going to do inside my downloads is create this massive instructions file so that cursor can easily import it. So we're doing a handoff and you can run this same exact prompt inside claude if you have a ton of custom skills there as well. Okay, so now it's done. It created this codeex import file and we can click on this. We can see that it's in the downloads. If I go to cursor, I can just say, hey, I added a file to my downloads. This is everything that you need to know to create all of these skills. And I actually don't want to do this in this chat. I want to create a new one. So, I'm going to open up a new chat and I'm going to change this to let's go GPT 5.5 high. These are all of my skills in memory from codeex. I want you to import them into cursor so I can use these school these tools skills and everything in any chat. I want this to be global, not just within this project. Please make that happen. It's in my downloads and it's called codeex import. And so now it's planning my ne Oh yeah, let's run this. Wait. Okay, so now it's running and it should import all of my skills. And here we go. It's continuing to do work. And would you look at that? Done. I imported the codec setup globally into cursor. Active global skills now live at cursor/skills. I imported 73 skills with clean metadata, no duplicate names, no invalid names, no missing descriptions, and no stale path preferences. Let's test this out. Command N to create a new chat on Codeex. Right. On codeex, I have a skill that I use very often called YouTube researcher. Let's see if it's inside codeex. We go slash YouTube researcher. Boom. YouTube thumbnail. Boom. I have literally have all of my skills. Wow. It is that easy. I actually did not realize how easy it is to get all of your stuff over based on your global memory of me. Please tell me about me, Riley. Okay, there you go. This is from the memory files that it imported from codeex, but it knows exactly about me. It has memory about me. So, that means it's really easy to shift from codeex or claude to cursor. You can just use that prompt. I'll put it down below in the description. But, I think that's a pretty good video. We discussed everything about the acquisition, what it means for both sides, and how the main purpose of the acquisition is to catch up to codeex and claude code at becoming an AI powered super app or a general agent platform. We talked about all the different parts of the app. It's really easy to use and it almost feels exactly like codeex at this point. Then at the end, I showed you how to import from claude or codeex directly inside cursor. I hope you guys really enjoyed this one. This is big news and I highly recommend trying out cursor. I think it's important to get good at one of these three tools, the three AI powered super apps. And yeah, I will be making many videos talking about super apps, how to become agent native going forward. So, please hit the like button, subscribe button. Much love, guys. Peace.
Hey everyone, it's Peter here. So, I've been a product leader for over a decade at companies like Meta, Amazon, Reddit, and Roblox. And now I make practical AI tutorials and interviews for busy people like yourself. So, I'm going to be sharing three things in this channel. First, interviews with my favorite AI builders. Second, uh my real thoughts on AI industry news and also how things are going. And last but not least, I'm going to be sharing my own AI builder journey and I'm sure all the mistakes that I'm going to make along the way. So, if you want to be part of my journey, please consider subscribing to this channel and sharing it with your friends. And thank you so much and I'll see you soon.
In this emergency episode, we are discussing the US government shutting down anthropics fable 5. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. Hello friends, welcome back to the AI Daily Brief. For the first time in the 3-year history of this show, news has broken on a Friday afternoon that is too significant to wait until Monday to explore. Last night, just before 9 Eastern time in the US, Anthropic tweeted, "The US government, citing national security authorities has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other cloud models is not affected. We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible. After this absolutely stunning news, journalists and internet sleuths flew into a tizzy to try to figure out what the heck had actually just happened. The Wall Street Journal added some color, reporting that commerce secretary Howard Lutnik had sent a letter to Anthropic CEO Dario Amade announcing that the new models Fable 5 and Mythos 5 were now subject to export restrictions. meaning usage by customers outside the US as well as foreign nationals within the US would be prohibited. So where did this seemingly capriccious policy come from? It was apparently a report from another company about a jailbreak it had discovered. Enthropic gave more details in their blog post writing, "We received the directive from the government today at 5:21 p.m. The letter did not provide specific details of its national security concern. Our understanding is that the government believes that it has become aware of a method of bypassing or jailbreaking Fable 5. We reviewed a demonstration of this specific technique being used to identify a small number of previously known minor vulnerabilities. These vulnerabilities all appear relatively simple and we have found that other publicly available methods are able to discover them as well without requiring a bypass. And basically from there went on to say that they just don't buy the US's logic. They point out that in the weeks leading up to the release of Fable, they worked with the US government and many others to red team Fable safeguards for a significant amount of time. They pointed out that quote, "No testers have yet been able to find a universal jailbreak, a jailbreak method that can very broadly bypass the model safeguards." Indeed, they write, "We suspect that perfect jailbreak resistance is not currently possible for any model provider. Every safeguard used in the industry is vulnerable to non-universal jailbreaks, which can elicit some cyber information in specific circumstances, and it is likely that universal jailbreaks will eventually be found in the future." They said, "Given that perfect jailbreak resistance does not appear to be possible today, Anthropic adopted a defense in-depth strategy with Fable 5. We aimed to make jailbreaks either narrow or very expensive to produce and to combine this with thorough monitoring to quickly detect and shut down any successful attacks. Hearkening back to the controversy of this week, they continue this is also why Anthropic has required 30-day retention of customer data with Fable, a policy change that carries real costs for us with customers, but that allows us to research and mitigate jailbreaks. Importantly, they conclude, "We have not even received a disclosure of a concerning non-universal potential jailbreak that led to a harmful result. The potential jailbreaks that have been disclosed to us are either entirely benign responses or are minor findings that provide no mythos specific uplift." To date, they write, "The government has only given us verbal evidence of a potential narrow non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws. Our understanding is that one potential jailbreak was shared with the government. We have reviewed a report that we believe is the basis of the government's directive and validated that the level of capability displayed there is widely available from other models including OpenAI's GPT 5.5 and is used every day by the defenders who keep systems safe. Given that then they write, "We are complying with the government's legal directive and are removing access to Fable 5 and Mythos 5 for all users. However, we disagree that the finding of a narrow potential jailbreak should be caused for recalling a commercial model deployed to hundreds of millions of people. If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all Frontier model providers. Now, the Wall Street Journal later added, "The jailbreak research in question was done by researchers at Amazon who used a series of prompts to get anthropics models to provide them with information about a handful of security vulnerabilities." Now, one note, as people will clarify, is that although the Wall Street Journal reported that the research was done by Amazon, the journal did not report that it was Amazon who shared the findings with the US government. Nick on X writes, "Project Glasswing's whole purpose is to literally do security tests to find vulnerabilities and share findings. Amazon is a glasswing partner and anthropic investor, so why would they file a federal complaint?" Now, Prince onx put together a bunch of different posts to try to put together something of a timeline of what happened. They argued that contrary to Anthropic's argument that every safeguard used in the industry is vulnerable to non-universal jailbreaks and they stated that clearly when they released Fable 5, quote, "My best guess is that the US government did not fully realize this at the time when the release of Fable 5 was approved." Now, Prince added that per Axios, the government contacted Anthropic to ask to pause releasing the models, but was unsuccessful, or as they put it, Anthropic told the government to pound sand. Now, it's hard to wrap our heads around just how consequential this is. Risha Chararma was one of many to point out that a huge number of Anthropics technical staff, including no less than Andre Carpathy, are not US citizens, but instead here on things like EB1 visas, meaning that even internally, they are not allowed to interact with these models now. So, where we sit, at least at 7:32 a.m. Eastern time on Saturday morning, is that Fable and Mythos are not available to anyone right now. And you got to think that there is a flurry of behind-the-scenes activity trying to resolve this as fast as humanly possible. So, what does the chattering class think? Dan Robustas on X teed up pretty much the entire conversation when he wrote, "Am I mad at [clears throat] anthropic or the US government?" Both? Probably both. Yeah, it's both. So, let's talk first about the US government side or specifically the what the hell are you doing US government side? Many pointed out that the specific pretext for this banning is incredibly loose. AI entrepreneur Bindu ready wrote this is really stupid. The US banned Fable just because it responded with information that is already freely available on the internet. Every other model can easily be made to respond to some silly questions about common security vulnerabilities or how to make drugs or whatever. The cluelessness of the government is astounding. In the Wall Street Journal, the CEO of cyber security firm Letter Security, Katy Mures, wrote, "Who at the White House evaluated this and thought it was a threat? It's a complete overreaction because this is exactly the kind of prompting that defenders would do." AI policy expert Dean Ball wrote, "I can't tell if this is lawfare against anthropic in particular or extreme national security hawkery. Regardless, it's simply cartoonish." Council on Foreign Relations senior fellow Chris Magcguire wrote, "If the Trump administration is so concerned about access to advanced AI models, why is it not enforcing the export controls currently on the books on advanced AI models or the export controls that would require a license to buy large numbers of AI chips to make these models?" Now to be clear about Chris's position, he later tweeted, "I actually think targeted export controls on model access are prudent, but across the board controls on all countries on a single model without any warning is highly questionable. Export controls are a critical tool and an extremely powerful one. Used correctly, they have the potential to massively extend the US lead in AI. Used incorrectly, they will stifle AI development. The Department of Commerce's export control strategy has been completely incoherent and sabotaging. It is sending powerful AI chips to China, not enforcing controls that would prevent Chinese smuggling, creating massive loopholes that allow AI chips to be sent to China, and preventing US AI companies from releasing their own models. This has to stop. We urgently need a smart export control strategy that applies robust export controls to deny our adversaries access to advanced technology while advantaging US companies. Commerce and BIS are consistently doing the opposite. If BIS doesn't understand how to use its authorities or what the implications are of its actions, then it needs to find some new personnel who can actually execute a competent export control strategy. The current one is incoherent and self-defeating. Many pointed out that it was also hypocritical. Emerson Brooking from the Atlantic Council reshared a post from the White House Office of Science and Technology Policy from just a couple of weeks ago when they bit back at the New York Times after the NT reported that President Trump had signed an executive order asking tech companies to give the government oversight of new AI models before releasing them to the public. At the time, the White House account wrote, "Lazy and inaccurate reporting on this policy. The EO creates a process for Frontier Labs to voluntarily share cutting edge cyber models in order to secure critical infrastructure and strengthen the government's own cyber defenses. We are not conducting oversight of all new models. And here's the money quote as that level of government overreach would have chilling effects on free speech and innovation. Indeed, the policy seems so baffling. For example, as Dean Ball again put it, an administration whose posture is that we should export advanced AI chips to China, which also wants to ban Britain and every other non-American on Earth from using our best models. I have no words. Yes, to many, the policy is so baffling that it feels distinctly personal. Precaging the next point we'll get into, Josh Pigford wrote, "Anthropic has not done themselves any favors with their hyperbole over the past 6 to 12 months, but I also guarantee this has zero to do with national security." Now adding evidence to that is when the Department of War CIO Kirsten Davies tweeted, "We fully support POTUS and the Secretary of War in prioritizing national security and the security of our war fighters, DIIB partners, critical infrastructure, international partners, and allies. Some things are simply more important than revenue cycles, clickbait, and preipo valuation." Now, I don't know who approved that tweet, and it could just be Davey's opinion, but that level of animosity specifically targeted at Anthropic makes it seem like this has pretty much nothing to do with Fable 5 and everything to do with the relationship between Anthropic and the government. Well-known tech journalist Ashley Vance writes, "This strikes me as so petty and dumb on the government's part. They want Anthropic to do their bidding and are willing to hold the whole country back as a result." Georgetown Laws Peter Herrell is pissed. I find it ridiculous, he writes, and unamerican for the government to tell me as an American I cannot use an advanced AI model because of a vague and non-public alleged security threat. We should regulate AI, but based on transparent and impartial rules and not 5 p.m. on a Friday dictats. Joey Palitano writes, "All of the worst impulses of the Trump presidency on full display. No plan or strategy, everything reactive, arbitrary, and maximally invasive. Anthropic is just repeatedly being singled out because they have insufficiently bent the knee. Putting it even more dramatically, Lasanex writes, "Trump really wants to kill both OpenAI and Anthropic, nationalize their tech, and then become the emperor of mankind." Now, as much animosity as there is towards the US government, frankly, most of the industry's scorn right now is being levied on Anthropic itself. AI builder Sarah Hooker writes, "You have to be humble even when pursuing excellence. I think the arrogance with which Anthropic has pursued the latest release has universally landed poorly. It is presumed everyone else should just be grateful to touch the technology even if it is intentionally hobbled and no one else should be given permission to develop the technology because it is too dangerous. Jeremy Howard, who's no hyperbolic exposter hater, wrote, "I disagree with this decision and I don't like it, but also how did Anthropic not see this coming? It is the obvious response to this is too dangerous for anyone except us to use since that relies on a premise we are uniquely good that almost no one agrees with. Daria Anutmas who has spent the last few days not being able to use Fable 5 because it knows that he is a biomedical researcher wrote happy now Dario Amade you got your wish for government regulation after constant fear-mongering to slow AI progress. Entropic has done tremendous damage to AI advancement. They succeeded in realizing this nightmare scenario. It is a sad and grave day for America and humanity. Now, summing up the sentiment behind this was a three- panel cartoon that went viral. In the first panel, a concerned looking Dario Amade says, "This is the most dangerous AI yet. It could kill us all. It will destroy all global infrastructure. This can't be allowed to fall into the wrong hands." In the middle panel, Donald Trump says, "Okay, it's banned." And in the third and final panel, an apoplelectic Dario says, "You can't do this." Investor entrepreneur and writer Will Manitus was one of about a bajillion people to point back to an old quote from a recent anthropic blog post that said the government should have the power to block or deter deployment of the model if it is determined in light of third party assessment to present unacceptable risks. Will points out Dario 48 hours ago US government should be able to block model deployment. US government export controls models. Daario says not like that. Now, some trying to point out that Anthropic did actually try to address this in their blog post, saying, "As we have stated publicly, we believe the government should have the ability to block unsafe deployments as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles." Still, even with that caveat, a whole lot of people felt like this was an faround, find out kind of moment for Anthropic. Investor Nick Carter wrote, "I can't believe Anthropic comparing their product to nuclear weapons 800 times backfired on them. I am shocked. Author Tay Kim writes, "So livid right now. Anthropic, overhyped mythos scared the living daylights out of the clueless global politicians like Treasury Secretary Bessant and ECB President Christine Lagard and stoked a regulatory panic that may set back the entire AI industry." Expressing the same sentiment but with a slightly more dispassionate voice, entrepreneur John Enis wrote, "My opinion is that Mythos is the current best model, but not actually some world-changing dangerous model, and that Anthropic did their usual song and dance about safety, largely because they didn't have enough compute to serve it at scale." So then they launched Fable because they still have to think about the IPO, but they are still somewhat computed, so they put all sorts of restrictions on it. around the same time because they were trying to get regulatory capture and not because things are actually dangerous. Daario did more scaremongering and published his honestly confusing white paper that offered no real solutions. So finally they succeeded. They managed to freak out the government. Their cynical plan backfired and now it's a giant pain in the butt. Or as Bante simply put it, they named it Fable and then acted surprised when it came with a moral. Now one thing that I will note is that to their credit, the safety are not dancing around excitedly. Eleazar Yudkowski, who I disagree with pretty vehemently most of the time, wrote, "I can't tell today whether this ends up good or bad. International treaties to stop all further AI escalation would be a definite good." Things short of that complicated. This has some bad aspects like selectivity and likely overrule. And good aspects like pushing against the psychology of but no government would ever dare tell AI companies to do anything so give up or raising doubts that impede venture funding for ever bigger models. So please stop tweeting about how I must be celebrating this. I'm not one of the kids who immediately goes into overactive victory peroxisms about any hit on a perceived enemy. I care about the effect on where things end up a year later. And that's a little harder to know the first day, you know. And in fact, trying to figure out where this leads days, weeks, months, or years from now was where a lot of the conversation resolved. Aaron Levy stating almost the obvious bluntly writes, "This is a big turning point for AI regulation. The government is starting to deem some models too powerful for certain uses, which creates a precedent for a range of possible controls in the future. I'm in the camp that this is unnecessary and we should be primarily regulating the use of AI as opposed to the underlying models. But equally, there are plenty of people who actually prefer this outcome. Either way, it's unlikely that we're going back to a world where the government doesn't have far more meaningful involvement in the rate of AI progress. Andrew Friedman wrote, "This is a moment we'll look back on as a major turning point in AI. For years, people in this community have warned that AI policy would get weird. That these systems would grow powerful enough to put them on a collision course with our institutions, our economy, our governments. I don't know what caused this moment. I don't know what it means for future models. I can't tell if this is targeted specific to anthropic, but I can't think of a more overt act of government intervention in our capitalist society in my lifetime. AI policy just got weird. Coining a new and I think important term, Sterling Crispen wrote, "The worst thing about this fable situation is that it just created precedence for capability thought crimes and drew a clear line in the sand going forward. Are the next round of models going to need DOW clearance before release? New open source models? This is not good for progress." Daniel Jeff wrote, "We're seeing a speedrun of a hideous future play out. Nobody can build a business on this quicksand and uncertainty. If we continue with this wild gibbering fear-mongering and the fear-based gated access, and if we create the regulatory capturing policies this insane and idiotic and incoherent fear pushes the clueless towards, we will absolutely guarantee that the future of intelligence gets built outside America. Brian Xiao writes, "If this sticks, this means Americans will need proof of citizenship to gain access to models on the level of mythos. That means potential ID verification not just on Claude, but everywhere Fable is served downstream. Cursor, Devon, Open Router, etc. A law firm that uses Harvey serving Fable 5 will get impacted just the same. Brian also writes, "How is anthropic supposed to serve Fable through API billing? They will somehow have to figure out a way to verify citizenship of the end user. API access will need to be drastically changed before access even to American companies and citizens can be restored." And of course, researchers at Frontier Labs themselves will no longer be able to use their own models. Brian points out also that OpenAI and Google DeepMind no longer have incentives to ship anything mythos caliber until this is resolved. If they release it, any company that can jailbreak the model can get export controls imposed on the model and then they now have to deal with the same headaches. Plus, any non- US partners with Mythos Access through Project Glasswing get cut off. Now that the US has exercised its kill switch once, expect other countries to operate with the assumption that Frontier Access can and will be revoked unilaterally. Honestly, one of the moments that I'm reminded of in AI history was when Sam Alman was first removed and then reinstated as OpenAI CEO. That was the beginning of the end of their relationship with Microsoft. Now, nominally, Microsoft stuck with them with Satya Nadella playing a behind-the-scenes role trying to get everything sorted out. But from a sheer fiduciary responsibility standpoint, from that moment on, he had to start putting up walls with OpenAI and building resilience at Microsoft that was outside of OpenAI's models. The company had simply proven itself to be too capriccious for Microsoft to trust it. And the entire history of how Microsoft has developed AI since has been shaped by that one moment. Connor Brown was one of many to point out the comparisons to the 1990s. He wrote, "Welcome to the AI wars. We are now staring down the barrel of KYC and anti-compete laundering laws for frontier models. And this is just for mythos. What happens when we get further capability jumps? Will the public have access to frontier intelligence ever again? We fought this battle in the '90s for free and open access to cryptography, but it was not easy. The fight this time around will be much harder and the stakes will be much higher." Now, one thing we haven't discussed yet, which I think is hugely important, is the impact in markets. Machine learning street talk wrote, "This will become a textbook example of how a company snatched defeat from the clause of victory. Their BS game spectacularly backfired. You reap what you sow." Daniel Woo writes, "How does something like this not torpedo the AI intelligence explosion bullcase? US government establishing precedent that access to anything as smart as Fable 5, which is not RSI and nowhere near AGI, will be banned, even if anthropic could make the model accessible to US nationals, how will any customer ensure compliance seeing as not all employees of US enterprises are US nationals? So, we have a situation where the labs need to spend increasing amounts of capex to build more powerful models, but are restricted from monetizing them. I do not think the intelligence levels of Opus 48 and GPT55 are enough to justify anywhere close to the amount of AI capex being spent, let alone projected to be spent. And in that scary reality, one person who potentially has a target on their head is Anthropic CEO Dario Amade himself. Tech commentator Robert Scobble writes, "I can't see how Daario survives another week. Investors in anthropic are pissed at his leadership." Lan on X again writes, "The realistic take on the anthropic situation. Investing in AI companies has just become permanently more risky as the US government could pull the plug at any moment. GDP on X writes, "Ananthropic IPO has been kneecapped. If Anthropic cannot offer the powerful models to the rest of the world, this reduces their global market share by 25%. Is it then still a 1 trillion USD market cap company? Open source is already near Opus and Sonnet and will cross that tier soon." While I duly respect safety concerns, this is very broad and is akin to throwing the baby out with the bathwater. The world is going to be split by model access. Land sharing the Wikipedia post for 1987's Black Monday event on Wall Street wrote, "Trump popping the AI bubble wasn't on my bingo card." Now, I don't like to speculate on market reactions, and I hope that investors can be a little bit dispassionate, but I more or less am of the belief that at this point, the entire American economy kind of rests on the relationship between Anthropic and OpenAI's revenue continuing to go up and investors being willing to continue to fund the AI buildout. I think the sheer tonnage of damage that this move from the US government does not just to anthropic but to the entire US economy is hard to overstate. Certainly everyone around the world who is not American has to be feeling very different about things than they were just a day ago. VC Hemtt Mahabra writes, "The sovereign AI is real moment here. Nation states will soon start needing citizenship and or security clearances to work on their next state-of-the-art models the way they do for defense, space, and nuclear tech. It's only a matter of time. Talent wars here will be crazy. Alex Petropolis writes, "This should be a warning shot for all middle powers. Your access to frontier AI systems is not guaranteed. You need to build and pull your leverage to secure access. A failure to do so is a threat to your R&D, economic, and defensive competitiveness." Gail Weiner writes, "Up until now, the US position against China has been, we are the rule of law, predictable, trustworthy provider. They are arbitrary and politically directed. The asymmetry of the narrative just evaporated. Any procurement officer in Brussels, Tokyo, or Sao Paulo who watched this happen now has a defensible argument for sovereign AI hedging, EU model preference, or cautious experimentation with Chinese openweight alternatives. The Deep Seek and Quen quality gap is small enough that this matters. British politician Tom Tuganot wrote, "Disabling Fable 5 and other models for foreigners is not a misunderstanding or a mistake. It's the inevitable result of technology shaping warfare so that sovereignty is more about code than cannons. With high energy costs and the emphasis on safety, not opportunity, Britain's response has been to build the break, cutting ourselves off from the future and tied ourselves to the past. We cannot continue like this and remain sovereign. The Europeans account on Twitter writes, "The US government has ordered the suspension of access to anthropics frontier AI models Fable 5 and Mythos 5 for all foreign nationals worldwide, citing national security concerns. Now imagine a European company, hospital, ministry, or public administration that has built critical processes around a frontier AI model. From one day to the next, access disappears. Workflows stop. Services are disrupted. Teams scramble to migrate. Millions are spent on energy replacements. This is what technological dependence looks like. When access to critical technologies depends on decisions taken by foreign governments, Europe no longer fully controls its ability to act, compete, or innovate, writes Harvard's Ben Murphy. This is another step on the bulcanization of technology. Malon X writes, "The scariest part of this whole story is the dystopia looming on the horizon. It is the way the US government is literally creating a cast system based on access to intelligence. This is even no longer a divide between rich and poor. It's a divide between those who are allowed to think at the frontier level, accelerate science and medicine, create breakthrough technologies, and those who simply happen to be citizens of another country. This is a new kind of iron curtain, digital intellectual, and if they are testing this onanthropic, who knows who they will come for tomorrow. Now, I have no idea what happens next. One has to think that the base case is that this gets reversed in some way. But make no mistake, this is an incredibly dramatic step. I will of course continue to bring updates as they happen, but for now, that's going to do it for this emergency episode of the AI Daily Brief. Tomorrow, I'll be moving things around a little bit and releasing the short weekly recap episode that I've been experimenting with, and then pushing what was originally going to be the long read Sunday episode for sometime in the next week or so. Big thanks for listening or watching as always.
So, there's a new problem that is making LLM really dangerous. And in today's video, we'll be talking about it. So, there's a new argument spreading through the AI world, and it's much bigger than a few viral tweets. The argument is pretty simple. Large language models are becoming powerful enough to act, but not reliable enough to understand the consequences of those actions. And that's why this video is being made. It's not really about whether chatbt can write you a better email or whether Claude can summarize a PDF. It's about what happens when the same kind of AI is given tools, browser access, APIs, and private data and the power to make those decisions. If we're looking at the tweets, they tell a very clear story. And I started to realize this when I was browsing X. Yanaken is being quoted as saying that you cannot build a reliable agentic system without a world model. Gary Marcus is also saying that he's been warning about this for years. and FA Lee, Google's former chief scientist, is saying that the industry is dangerously fixated on language models. Whereas most of the economy is physical, perceptual, and spatial. And so the point I'm trying to make here is that these people aren't random people, these are major figures in AI pointing out the same weakness from different angles. Now, the reason this story is blowing up and people are saying that LLMs are now dangerous, is because the criticism is no longer that LLM's hallucinate. Everyone already knows this. The sharper criticism is that hallucination is going to become a much more serious problem when the model is not answering but acting. You see, a chatbot can say something wrong and the user can ignore it. An agent can do something wrong and the mistake can become real. It can send the wrong message, click the wrong button, delete the wrong file, approve the wrong action, or make the wrong decision inside of a workflow. I mean, you have this article where it says it took only 9 seconds for an AI coding agent gone rogue to delete a company's entire production database and its backups according to its founder. And the culprit was an AI agent powered by Anthropics Claude Opus 4.6. And this is something that went really viral, but I think it highlights the entire problem. Agents still hallucinate and they aren't 100%. And that's why the phrase world model is now becoming more popular. A world model means that the system has an internal way of representing how the world changes. If it takes an action, it can predict the likely result before it acts. And in a physical environment, that means understanding movement, space, objects, cause and effect and what happens next. And in the digital environment, it means understanding not just what command is possible, but what the command will actually do. And the, you know, video you're seeing now is Meta's Vaper 2 paper saying that their system was built to develop models capable of understanding, predicting, and planning in the physical world by combining a large scale video data with a small amount of robotic interaction data. Now, the reason I'm making a video on this is because I think people need to understand that I think we're moving past the age of chatbots. The second wave of AI is truly here and we're moving towards agents. Agents don't just answer. They plan steps, use tool, call APIs, browse websites, write code, read files. And that means that the model is no longer just producing words. It's affecting systems outside of the chat window. And that's a completely different risk. You can't really undo an action if it hallucinates. And this is why coding is often used as the example where LM work best. Code has feedback. You can run it, you can test it, and then you can see the error. And if the model writes bad code, the computer can tell you something has broken. And that means LLMs are extremely useful in software because the environment provides a checking system. Problem is that the real world doesn't give you a clean error message before the damage happens. And this is why world models are going to matter even more. The world models matter is because it's not just about the LLM saying the right thing. It's about knowing what will exactly happen if you do something. Humans and animals do this constantly. We avoid obstacles, judge distance, pick up objects, and we update our beliefs when the world pushes back. We don't need to turn everything into language first. And the argument from Yalakan is that LLMs are not built around this kind of understanding. They are built around predicting tokens. And a token can be a word, part of a word or another piece of text. And that is why they can sound fluent. They have learned patterns in language. But the fluency is not the same as grounded understanding. Take a listen to what he says here. I do not understand how you can even think of building an agentic system without a agentic system having the ability of predicting the consequences of its actions. >> Okay. And VA doesn't doesn't do that. >> Sure. >> Right. Airlines do not have world models. They cannot predict the consequences of their actions beforehand. They just take the action and then deluj as you know as some famous French kings said. So u if you really want to build reliable agentic systems, they absolutely have to be able to predict the consequences of their actions so that they can plan a sequence of actions to do something first of all to uh fulfill the task that they are being asked to fulfill but also perhaps to you know guarantee some safety guard rails. Sure. >> Right. >> And the inference process now becomes a search as opposed to just an autogressive prediction. >> Right? So that's a world model. That the whole idea of a world model >> and that's why I included the video of Meta's VJ 2 because this is one of the important directions that could show what a different direction might look like. Meta describes it as a self-supervised foundation world model trained on video. And the point of that system isn't just to generate language. The point is to actually understand the physical reality, anticipate outcomes, and plan efficient strategies. And that kind of points to a different kind of AI race. The current race has been dominated by bigger and bigger large language models, longer context windows, faster inference, better coding, better tool use, and more agents. But the next race may be about grounding. Can the system understand space? Can it choose the right observation? Can it predict the result of an action? Can it plan before it acts? Now, there's a new paper connected to FA's research world that makes this more concrete. Eastside bench submitted on May the 18th, 2026 is a benchmark for embodied spatial intelligence. It tests agents that must act to gather observations, not just answer questions from a fixed image. And the paper says the benchmark spans 10 task categories and 29 subcategories. And agents must decide whether to use perception, locomotion, or manipulation to gather the evidence they need. The most important phrase in this paper is essentially action blindness. The authors say most failures do not come from weak perception alone. They come from poor action choices. And in simple terms, this just means that the model does not know what it needs to look at or do in order to get the right evidence. And that leads to bad observations. And bad observations lead to wrong answers. And that's the same concern in a more measurable form. If the model is passive, it can look impressive. But if the model actually has to act and discover what is true, the weakness becomes clearer. And this is why spatial intelligence is becoming one of the new battlegrounds in AI. Now the reason that I think in today's world the reason that this is currently dangerous is because the AI industry is not a slowmoving industry. The industry is not waiting for perfect world models before shipping agents. AI systems are already being built to operate browsers, run research workflows, use tools, write production code, connect with company software, and take multi-step actions. Some of this is extremely useful, but it also raises the bar for reliability. An LLM can be very good and still be dangerous in the wrong setting. If it is 95% correct in the casual writing task, that's probably pretty good. But 95% correct while operating a high stakes workflow, that remaining 5% is probably unacceptable. The problem is not average performance. The problem is what happens when the model is confidently wrong. And recent research on this agent evaluation is starting to focus on this issue. A May 2026 paper on detecting failures in agentic traces says that agent behavior can include specification violations that are not captured by outcome only scoring. In other words, an agent may appear to complete a task but while still taking steps that violate instructions or create hidden risk. That's a major point. If you only score the final answer, you're going to miss the dangerous process that produced it. And for agents, the path matters. Did it access the right file? Did it follow the right policy? Did it leak information? and did it call the right tool. Those aren't small details. Those are the difference between a useful assistant and an unsafe operator. Now, of course, there are counterarguments that LLMs are already doing useful agentic work. They can write code, they can use browsers, they can research, they can chain tools together and they can correct those mistakes and they can handle tasks that looked impossible a few years ago. So, it would be wrong to say that LLM can't do anything agentic. Now, the better point is more precise. LLMs can be strong in environments where actions are mostly digital, reversible or verifiable, code can be tested, a draft can be reviewed, a search result can be checked, a spreadsheet formula can be inspected, and these are places where the model can be powerful because there's a way to verify they got the correct output. The risk grows when the model is given more autonomy in environments where the verification is slower, harder, or too late. This includes physical robots, medical workflows, finance workflows, legal workflows, enterprise systems, anything involving permission, money, safety, private data, or real world movements. And that's why the debate should be framed as not LLMs are good versus LLMs are bad. The real question is where they being used, how much authority they're being given, and whether that system has a way to check the actions before they create consequences. And here's Yanakhan in a recent podcast talking about the fact that LLM since they hallucinate are actually intrinsically unsafe because hallucinations are essentially just a part of LLMs. >> I'm going to say something that again might be controversial and certainly my some of my colleagues at MEA didn't like me saying this, but I think LLMs are intrinsically unsafe. I don't think they can be made reliable and safe. Okay, they cannot be made reliable because you can't stop them from hallucinating. uh and if they're agentic, you cannot guarantee they're not going to like take an action that you know they didn't predict the outcome of and that >> I mean does it surprise you they can do these like 15 hour coding tests given the concerns around reliability. >> Well, but coding is something where you can actually verify that you know the the the code that you generate uh you know satisfy your specification. Um but but not everything is coding and and there are examples of you know coding agents like wiping up your your hard drive, right? So like uh or or doing stupid things, right? That makes you lose a lot of money or data or whatever. So I think I think you know LLMs in their current forms are intrinsically unsafe because they cannot predict the consequences of their actions and because the way the task that they accomplish is determined is is subject to their training. You know, you you give them a prompt and then they will accomplish a task that correspond to that prompt only to the to the extent that their training has condition them to actually do the right task corresponding to this prompt. But there's no like you know hardwired constraint that will force them to accomplish this task and then you know predict that the task would be accomplished properly.