12:24Nate Herk | AI Automation · Talking Head
Is Claude Mythos Coming?
A 12-minute reality check on the Mythos API leak and why one buried Anthropic sentence beats all the hype.
June 6thThe argument in one line.
The leading AI labs are calling for a global slowdown not because they will stop, but because they admit competitive incentives prevent them from stopping alone — and a pause only holds when breaking it costs more than winning.
Read if. Skip if.
READ IF YOU ARE…
- You use AI tools daily and want to understand the geopolitical stakes framing the tools you depend on.
- You are curious why the companies most invested in AI speed are also publicly asking for a brake.
- You have family members not yet using AI and want to understand how urgent the skills gap really is.
- You want a non-hype, builder-perspective read on the OpenAI IPO-path and Anthropic pause proposals from June 2026.
SKIP IF…
- You want deep policy analysis — this is a solo builder thinking out loud, not a regulatory breakdown.
- You are already deeply read on AI governance; the treaty section covers basics only.
The full version, fast.
In the same week both Anthropic and OpenAI published requests for an international mechanism to slow AI development, each company also moved toward going public. The host reads this as a structural confession: the competitive incentives are too strong to resist unilaterally, so both labs want a third-party referee to make everyone stop together. He argues verification is solvable (training runs have a visible physical footprint; chips come from one supply chain), but the real wall is incentive — a treaty only holds if defecting costs more than winning. The practical conclusion: you have no vote on the treaty, so invest in judgment and taste, the human skills their own plans identify as the durable edge.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 01:52
01 · Both Labs Want to Slow Down
Hook establishes the central paradox: the two fastest-moving AI companies both published requests for a global slowdown in the same week they each moved toward IPO.
01:52 – 03:49
02 · The Massive Public Gap
The gap between AI builders (alarmed, deeply capable) and the general public (fearful, barely touched the tools) is enormous. Commencement speakers are getting booed for mentioning AI.
03:49 – 05:33
03 · Why They Cannot Stop
OpenAI own words: the incentives around commercial and national competition are hard to escape. Neither lab is stopping unilaterally; they want a third-party referee.
05:33 – 08:15
04 · Could a Global Pause Work?
Treaty mechanics examined: all serious AI countries must sign, inspectors must verify like nuclear weapons inspectors, chip supply chain is the uranium analogy. Real wall is incentive, not logistics.
08:15 – 11:10
05 · Bet on Yourself, Not a Company
Labs are converging and going public — you are a user seat to either. The skill stack that stays durable: use AI, use it safely, develop judgment and taste. Run existing workflows through the new tools.
11:10 – 12:01
06 · Claude Mythos and Fable Taken Down
US government export control directive suspended access on June 12. Related to the video theme but not the slowdown the labs asked for. Dedicated follow-up video promised.
12:01 – 12:27
07 · Free Community and Wrap-Up
400,000-person free AI community CTA. Like request.
Lines worth screenshotting.
- Neither Anthropic nor OpenAI is saying they will stop — they are saying they cannot stop alone and need someone else to build and hold the brake.
- OpenAI published their call for coordinated AI slowdown on the same day they announced their IPO filing.
- Training a frontier model leaves a physically visible footprint: tens of thousands of chips, massive power draw, construction — harder to hide than most people assume.
- The entire advanced chip supply chain runs through essentially one source on the planet, making chips the uranium-style leverage point for any AI treaty.
- A slowdown only holds when breaking it costs more than winning — until someone solves that incentive equation, every pause request is theater.
- The consumer/enterprise split between major AI labs is blurring as both chase the same public-market dollars; your loyalty to a brand is not reciprocated.
- The human role their own plans identify as irreplaceable is judgment and taste — deciding what is worth doing, not doing the doing.
- The simplest AI adoption move is not building new workflows but running your existing work through these tools instead of doing it the old way.
- A global AI pause agreement faces the same structural problem as any contract: it is only as reliable as the least trustworthy signatory.
- Both labs publishing safety appeals while simultaneously raising money from growth investors is a costly signal — they are paying a reputational price to ask for help they know they probably will not get.
The incentive problem no treaty has solved.
WHAT TO LEARN
Both leading AI labs have publicly admitted they cannot stop competing on their own — and until someone makes defecting more costly than winning, every pause request is a prayer, not a plan.
- Neither Anthropic nor OpenAI is claiming they will stop — they are saying the system they are inside prevents unilateral stopping, so they need an external enforcer nobody has built yet.
- Training frontier models leaves a physically visible footprint (power, chips, construction), so verification is technically more tractable than most assume — the real obstacle is national incentive, not detection.
- AI chip supply is concentrated through a single global supply chain, which makes compute the leverage point for any treaty, analogous to uranium in nuclear nonproliferation frameworks.
- A global slowdown agreement faces the same structural problem as any contract: it holds only as long as the party with the most to gain from defecting chooses not to.
- As both labs move toward public markets, the consumer/enterprise distinction between them is eroding — both are now chasing the same growth investors and the same user seats.
- The human skills their own roadmaps identify as irreplaceable are judgment and taste — knowing what is worth doing — not the execution of tasks AI can increasingly handle.
- The most accessible AI adoption move is not inventing a new workflow but running the work you already do through these tools instead of doing it manually.
Terms worth knowing.
- AGI
- Artificial General Intelligence — a hypothetical AI system capable of performing any intellectual task a human can. Both OpenAI and Anthropic use this as their stated end-goal.
- Frontier model
- The most capable AI models at the current edge of what is technically achievable, typically requiring massive compute resources to train.
- Compute bottleneck
- The framing that training large models is constrained primarily by access to specialized chips, not by ideas or data.
- Export control
- Government restrictions on the sale or transfer of technology — the legal mechanism the US government used to suspend access to Claude Mythos and Fable on June 12, 2026.
Things they pointed at.
Lines you could clip.
00:18
“The two companies racing the hardest are also the two who have turned around and said, hey world, we need to slow down.”
04:16
“The break they're asking for isn't a break that they're willing to pull. It's a break that they want somebody else to create and hold.”
07:40
“A slowdown only holds if breaking it costs you more than winning ever could.”
10:00
“If you're renting that intelligence from one of three companies that can change the price, change the rules, or cut you off whenever they feel like it, do you actually get handed that power?”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphoranalogy
00:00So a few days ago, Anthropic asked the whole world for a way to slow down AI. This week, OpenAI published their big plan for the future, and they asked for basically the exact same thing. So both of these companies also just took a real step towards going public.
00:13So the two companies racing the hardest are also the two who have turned around and said, hey, world. We need to slow down. So what the heck is going on?
00:20That's what I wanna get into today. Now let me just start off real quick by being straight with you. I use both of these tools every single day.
00:26Right? Claude, JGBT, whatever.
00:28I don't own any stock. I'm not on anybody's payroll. I don't have insider information.
00:32I don't even have a horse in the race to see, like, who wins. I'm just a guy who builds stuff and pays attention. So I'm not here to defend any of these companies.
00:39I'm just here to figure out what does this actually mean for people like you and me. So, anyways, on June 8, OpenAI put out this thing called built to benefit everyone, our plan.
00:49And it's Sam Altman and their chief scientist laying out their whole vision, essentially. And the mission line right at the top is to ensure that AGI benefits all of humanity. And that same day, totally separately, they also announced that they'd filed to go public.
01:03But, anyways, in this first article, they lay out three main goals. Number one is an AI automated researcher. So, basically, AI that can do a lot of the AI research itself, and they predict maybe by March 2028, they will have that.
01:14Number two is accelerating the economy and sharing the gains and basically giving every single person on Earth their own personal AGI. And they do say a lot of stuff that I agree with here. They say they don't want to automate everything.
01:25They say that they don't think power should be concentrated with just a few companies. And then number three is they straight up call for some kind of international group that could let the world, and this is the quote, take coordinated action, including slowing frontier development when needed.
01:40And it's interesting because a few days before this article from OpenAI, Anthropic put out their own report asking for the same exact thing, a verifiable way to slow or pause AI if it just gets too scary. So it's just really, really interesting to me because you've got the spectrum. Right?
01:56And on one end, you've got the two leading AI labs and people that are in the weeds every day, the people that understand this tech better than anyone. And they're basically saying, this stuff is moving so fast and it's getting so powerful that we need the entire world to come together and help us slow it down. And then on the other end of the spectrum, you've got a ton of just regular people who think AI is just some funny tool that makes an image or, you know, a fancier version of Google search.
02:19And they basically barely touched it, and they have no idea what it can actually do. And a lot of them think that it's a bad thing. There's a very negative, you know, association with AI.
02:28So that gap between the people building this stuff and the people who are going to have to live with this and go through this whole, you know, kind of, like, paradigm shift, not to throw out a buzzword, is enormous. It's an enormous gap, and I don't understand how it got so wide.
02:41Because the brand of AI right now is honestly terrible. And I was gonna make a separate YouTube video about this, and I probably will. But, I mean, you've got commencement speakers.
02:47Right? And they're getting booed at the second they bring up AI. And I think it's because a lot of people haven't really used it.
02:52So when they hear AI, they picture some society where robots are, you know, their masters and taking their jobs and, you know, all of this negative stuff rather than thinking, wait. I could actually just do way more, way faster, and my life actually gets a lot better because I'm using AI. And the way that I think about it is .
03:10On its own, it has no good or bad, you know, intentions. It basically just does whatever you point it at, and it helps you do that faster. So you wanna cure cancer.
03:18It'll help you do the research. It'll help you do all this stuff. If you want to hack into systems and you have malicious intent, it might help you do that too.
03:25Now most of these models are getting essentially filtered. They're getting guardrails baked in to protect against that malicious intent, but you get my point. The intent is always the humans.
03:34And just like these companies say, the judgment, which is, you know, knowing when and where and whether to even use this stuff, all of that stays us, the humans. The AI is more of the tool, and that's exactly why giving society more time to catch up is not a crazy thing to ask for. So why would the two companies with the most to gain from going as fast as possible both turn around and say, hey.
03:56Let's slow this stuff down. And OpenAI actually tells you why. The reason that they give for wanting this international group is, in in their exact words, especially because the incentives around commercial and national competition are hard to escape.
04:09They're basically telling us right there that they cannot escape the incentive to race because the competition and the incentive is just too strong. And Anthropics says basically the exact same thing, just in different words. Their whole pause idea hinges on everybody being able to verify everybody else is actually pausing too.
04:24So what's interesting there is neither of them are really saying we're going to stop. What they're both saying is we can't stop on our own. We can't give up our lead.
04:34So somebody needs to build some sort of group or like a referee that makes all of us stop together and actually verifies that everyone is stopping, and that's a completely different thing. You know, the break they're asking for isn't a break that they're willing to pull. It's a break that they want somebody else to create and hold.
04:50And not to throw out conspiracy theories, but it doesn't seem like a super easy thing. So I wonder, like, are they just throwing that out there for good press because they know it's not gonna happen? But, anyways, I don't necessarily think that it's them being evil, by the way.
05:02And, you know, to be fair, it's not like they've done nothing. They've each you know, they published their own safety rules, you know, for for years already. They say they won't cross without certain protections in place.
05:12And so they're essentially inviting more regulation into their own industry, right, as they're both going out to raise money from public investors who just want growth, which is it could be considered a costly thing to ask for. So, anyways, the people building what they themselves call the most powerful tech ever just told you on the record that the system they're in won't let them slow down on their own, so they're paying a price to ask for help.
05:33The So next thing I start wondering is, could that sort of treaty or referee could something like that even exist?
05:40And I'm not, like, a foreign policy guy. Right? But I just started thinking about actually what it would look like and what it would take to build what they're asking for.
05:48You'd basically need every single country that's serious about AI, probably every country in general, to agree. Not just The US. You'd need China, Russia.
05:56You need everybody all signing some sort of treaty saying that they would play by the same rules. And, you know, let's just say somehow that actually happens. Everyone agrees.
06:04Now you need someone who can actually enforce that. You need, like, inspectors who can go look inside these companies, inside these countries, verify nobody's secretly training a bigger model in some basement somewhere, you know, kind of like nuclear weapons inspectors, but for AI.
06:18Now you'd think the AI part would be the impossible thing to track. You know? It's just code.
06:22It's a bunch of chips in a building running math. It's actually the opposite. I think that training one of these frontier models takes a huge visible footprint.
06:31You know, tens of thousands of specialized chips, a single building pulling as much power as a small city, and that whole footprint is probably kind of hard to hide. You know, there's construction.
06:41There's power it's gonna pull. There's massive chip orders. It all does leave a trail, and those chips basically come from one supply chain on the entire planet.
06:49So the chips are kind of like the uranium here. Anthropic has said something about the bottleneck basically just being compute. So, anyways, the point I'm trying to make there is I don't know how you would verify it and how that would work, but that's one issue.
07:01Right? And that's probably solvable even if it's not super, super, you know, trivial.
07:07But I think the real wall is the incentive, just like opening I mentioned, because nobody who's in the lead is incentivized to slow down and let everyone else catch up. The whole game is to win. Really?
07:16I mean, so even if every country signs the treaty, the second that one of them thinks that they can quietly sprint ahead and grab the biggest prize in history, the whole treaty falls apart. It's the same way I think about a contract. Right?
07:28No matter how bulletproof a contract is, if it covers every single scenario, it's only as good as the person on the other side of it. So, anyways, that's the trap that OpenAI is pointing out when they say that those incentives are really hard to escape. A slowdown only holds if breaking it costs you more than winning ever could.
07:43So the real question isn't whether we could build that treaty and we could build that verification system. It's how do you actually flip that incentive? How do you get to a world where the winning move is to respect a slowdown?
07:54And you actually, like, think about society and, like, how can everyone be doing this AI thing together rather than competing. Because until somebody cracks that, I think that asking for a global pause button is either a genuine cry for help or just the safest thing you can say when you already know you can't actually stop.
08:10Maybe a little bit of both, and I don't obviously know which, but that's just what I'm thinking about right now. And whichever one of these labs you root for between OpenAI and Anthropic, you're not really the mission.
08:19I mean, I think people kind of feel like OpenAI is for the consumer, and, you know, Chattypi is the AI that your parents know or, like, the general person knows. And Anthropic's more of like the enterprise one. You know, Claude is built for businesses and engineers.
08:31And that split was real, but it's starting to blur, I think, now that they're both going public and they're chasing the same dollars. And either way, you're the same thing to them. You're a user to sell to.
08:41You're, you know, a seat on in some company's bill. So, anyways, the point I'm trying to make when I'm saying all this is don't really I wouldn't bet on a company. I'd just bet on yourself.
08:50Like, the skills. The skills that you need to learn on how to use AI, how to become AI native, how to do it safely, how to think about it, how to communicate about it. Those sorts of things is what you should be betting on.
09:00You know? And I don't have kids yet. A lot of you guys probably do.
09:03And if you've got a kid right now, you're probably thinking about all of this, and you're wondering what a career is gonna look like when they're 22 and entering the workforce. And I wonder the same thing, just kind of one step removed. Because I know the world in twenty years isn't going to be some slightly upgraded version of the one that we've got right now.
09:18It's going to be fundamentally different. How business works, how the economy works, where the money sits, where the power sits, all of that is just gonna look a lot different. And right now, it can feel like the whole future is getting decided by, I don't know, let's just say eight people that have the whole world in the palm of their hands because of this this huge, you know, tech wave.
09:37And plus, you know, you've got a couple of chip companies and stuff like that, but they're moving so fast that the rest of the world can't even react in real time. So how does the upside of all of this not just pool up around them? They basically said that they don't want it to be a huge divide like that.
09:53They want a personal AGI for every single person on Earth. And on paper, that's the great equalizer. Everybody gets the same superpower.
10:00But if you're renting that intelligence from one of, you know, three companies that can change the price, change the rules, or cut you off whenever they feel like it, do you actually get handed that power? Anyways, I'm not gonna pretend that I have the whole answer. You know, anyone who tells you they know exactly how this stuff is gonna pan out is probably trying to sell you something.
10:17But I don't know. You don't really have a vote on the treaty right now. You don't have a vote on whether they slow down.
10:21The only thing that you fully control is how good you get at this stuff and what you know about it. And their own plan basically tells you where to aim. They say the human role that stays valuable is judgment, taste, deciding what's actually worth doing.
10:33Because the moment everybody's got a personal AGI, the tool stops being the edge. So I think, like, if you are really obsessed with this AI stuff, which I bet if you're watching this video, you are, and you have a lot of friends, a lot of family, a lot of people that you care about that aren't, really start pushing them to learn it.
10:48You know? I just think it's really, really important. And you can just pick one thing you do over and over.
10:52Your email, planning out your week, putting together a report, and actually just run it through one of these tools instead of doing it the old way. That's the whole move as you just start trying what you already do every day with these new tools rather than trying to invent new workflows for yourself and invent new, you know, passions.
11:07Just do what you already do with this technology. Also, I'm sure you guys have heard the news, but Claude Mythos and Claude Fable did get taken down at the moment of filming this.
11:18They say it's gonna come back. We will see. But the US government basically told Anthropic that they needed to suspend access.
11:26So what's interesting about this is that the US government is getting involved, not in the exact same way that Anthropic and OpenAI have asked for as I talked about in this video, but it is something to mention.
11:37And I will be making more videos dedicated towards what happened here and what this means for the future, but I just wanted to at least let you know if you hadn't heard about this, this did happen. This was on June 12. You can obviously come here and read this tweet, and then put out a whole statement here, which you can read and dig into why that happened.
11:56Like I said, a video will be coming on it, but I wanted to fill it out there. So there you go. Anyways, if any of that got your brain going and you want to get good at using these tools yourself, I've got a free community.
12:06It's linked down in the description. We've got about 400,000 people in there who are building with AI, beginners, business owners, you know, engineers, whatever it is, a bunch of free resources in there, and a bunch of people who are talking about this kind of stuff, and it's completely free. But, anyways, that is gonna do it for this one.
12:19So if you guys enjoyed the video or you learned something new, please give it a like. Helps me out a ton. And as always, I appreciate you guys making it to the end of the video, and I'll see you on the next one.
12:26Thanks, guys.
The Hook
The bait, then the rug-pull.
The two companies racing hardest in AI published slow-down requests in the same week they each moved toward going public. Nate Herk, a builder who uses both tools every day and owns no stock in either, reads the fine print — and finds a structural confession buried in plain sight.
Named ideas worth stealing.
03:10concept
AI as Amplifier
AI has no intrinsic good or bad intent — it accelerates whatever human intent is pointed at it. Judgment stays human.
Steal forany video or post reframing AI fear into AI as a neutral tool
07:40model
The Incentive Flip Problem
A slowdown treaty only holds when the cost of breaking it exceeds the value of winning. Until that equation is solved, any pause is voluntary and fragile.
Steal forany argument about coordination problems, contracts, or competitive dynamics
06:40analogy
Chips as Uranium
Training frontier models requires a visible physical footprint. Chips come from one global supply chain — the leverage point for verification, like uranium for nuclear treaties.
Steal forany explanation of why AI governance is more tractable than it sounds
08:50list
The Skill Stack
- Use AI
- Use it safely
- Think about it clearly
- Communicate about it
The four skills that stay valuable regardless of which lab wins or what policy emerges.
Steal forany CTA around AI education
How they asked for the click.
VERBAL ASK
11:59link
“I've got a free community. It's linked down in the description. We've got about 400,000 people in there who are building with AI.”
Soft, conversational. Positioned as a community of peers, not a pitch. Delivered after the intellectual payload is complete.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
AFFILIATECommission earned if you click.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
