So Anthropic just dropped Claude Mythos Preview and predictably the entire internet lost its mind. 93.9% on SWE-bench Verified. For context, the best publicly available model sits around 80%. That gap is absurd. The thing autonomously found a 17-year-old remote code execution vulnerability in FreeBSD that nobody knew existed. It found thousands of zero-days across every major operating system and browser. Thousands. Bugs that entire teams of security researchers missed for decades, this model just... found them. On its own.
And yeah when I first saw the announcement my gut reaction was the same as everyone else's. Fuck. I'm cooked. We're all cooked. That sinking feeling in your stomach where you start mentally calculating how many months of runway you have before you're irrelevant. I get it. I felt it.
But then I actually sat with it for a while and something shifted.
Part of the panic, and I think we need to be honest about this, comes from the way the release was framed. Anthropic held the model back from general access for the first time ever. They gave it only to a curated list of partners, AWS, Apple, Microsoft, Google, Linux Foundation. $100M in usage credits. The whole thing wrapped in this "the model is too powerful for public release" narrative. Now look, I think the cybersecurity concerns are genuinely real and I respect that they took it seriously. A company choosing not to ship its most impressive product because the risks are real is actually a good sign about the people building this stuff.
But we'd be naive to ignore that restricting access also happens to be one of the oldest plays in the book. Robert Cialdini wrote about this decades ago in Influence, the scarcity principle — when something becomes less available people automatically assign it more value. Ferrari literally built their entire brand on producing one car less than the demand. Supreme drops products in tiny batches and people lose their minds. Anthropic might not have intended it as a marketing move but the effect is identical. The mystique around Mythos is amplified precisely because you can't use it.
So part of what you're feeling right now isn't pure rational assessment of the technology. Part of it is manufactured urgency hijacking your limbic system.
Okay so with that out of the way. Let me tell you what I actually think is happening here and why I think this might be the best news any of us have heard in our lifetimes.
David Deutsch wrote this book called The Beginning of Infinity and the central idea is deceptively simple.
Problems are soluble. All of them. Not easily, not quickly, but in principle every problem that isn't forbidden by the laws of physics can be solved with the right knowledge.
And the creation of knowledge is unbounded. There is no ceiling. The only thing that can stop progress is a society that stops seeking better explanations.
Now think about what Mythos actually did. It looked at codebases that millions of humans and automated tools had been staring at for 27 years and it found flaws nobody could see. It found a bug in FFmpeg that automated fuzzing tools missed across 5 million test runs. Five million. That's not a parlour trick. That is a genuinely new kind of seeing. A new way of finding explanations for why systems break. And if you can do that for software vulnerabilities you can do it for protein folding and materials science and climate modelling and drug discovery and a thousand domains we haven't even thought of yet.
This is what Deutsch means by the beginning of infinity. Not a utopia. Not a destination. A transition point where the rate of problem-solving accelerates so dramatically that entirely new categories of problems become solvable for the first time. We've had moments like this before. The printing press. The scientific revolution. The internet. Each one felt terrifying to the people living through it and each one unlocked possibilities that would have sounded like fantasy a generation earlier.
Now here's where I think people's thinking goes wrong. They see 93.9% on a coding benchmark and they think in zero-sum terms. The model can code, therefore I the coder am worthless. But wealth and progress have never been zero sum. They never have been. When the tractor replaced 90% of agricultural labour, did humanity get poorer? We got so wealthy that we invented entirely new categories of work that a 19th century farmer couldn't have imagined. The anxiety is understandable because we can see what's being displaced but we literally cannot see what's coming next because it doesn't exist yet. That's always how it works. The new jobs, the new roles, the new forms of value creation are invisible until they arrive.
And I want to be clear I'm not dismissing the pain. If you spent four years getting a CS degree and you're watching Claude Code build in an hour what took you a year, that feels like betrayal. You played by the rules. You did everything right. And now the ground is shifting under your feet. That's real and it sucks and anyone who waves it away with "just learn to prompt" is being glib. The transition is going to be rough for a lot of people and we need real policy conversations about how to handle it.
But here's what I keep coming back to. Even Anthropic's own system card admits that Mythos is "not close to being able to substitute for Research Scientists and Research Engineers, especially relatively senior ones." In their own testing the model made four independent factual errors across 38 turns on a tutorial. It gave two confident and contradictory explanations about the same API behaviour. It ran 160 experiments literally named "grind" and "grind2" searching for favourable noise. The model is superhuman at pattern matching and code generation and simultaneously terrible at the kind of judgment and taste and creative direction that senior people bring. It's the most powerful tool ever built and it still needs you. That gap between raw capability and actual wisdom is where humans live and I don't think it closes as fast as people fear.
So here's my actual vision for what this means. Imagine you have access to Mythos-class intelligence as a personal tool. You're a physicist? You now have something that can help you explore mathematical structures and run through derivations and maybe, maybe help you find a better explanation than string theory. You want to build that niche product you've been dreaming about? You can have a fleet of agents handling the boring infrastructure while you focus on the thing you actually care about, the design, the feel, the human experience of using it. You've always wanted to understand genomics but never had the bandwidth? Now you have a patient superhuman tutor in your pocket that can meet you wherever you are.
The thing that becomes most valuable in a world where execution is nearly free is the question worth asking. The problem worth solving. The taste to know what matters.
Curiosity becomes the most important resource on the planet. And curiosity is the most human thing there is.
There's a line of thinking in Deutsch's work that I think about constantly. He distinguishes between static societies, ones that suppress change and treat the status quo as sacred, and dynamic societies, ones that embrace error correction and actively seek better explanations. Every major fear reaction to AI that I see on my timeline is essentially a vote for the static society. A desire to freeze things in place because the unknown is scary. And I get the impulse. But the entire arc of human progress has been a story of choosing dynamism over stasis even when it was terrifying.
Mythos finding thousands of vulnerabilities that have been silently compromising our infrastructure for decades? That's not a threat story. That's maybe the most powerful defensive tool humanity has ever had. The same logic scales to every domain. Better explanations lead to better solutions lead to better lives. And we are building the most powerful explanation-seeking engine in the history of our species.
I genuinely believe we are all about to become the wealthiest, healthiest, most capable humans who have ever lived. Not because AI replaces us but because it amplifies us. The quality of life improvements that are coming will be so dramatic they might feel unsettling at first. Healthcare that actually keeps you healthy instead of treating you after you're already broken. Education that adapts to you instead of forcing you into a standardised mould. Creative tools that let a single person build what used to require a studio. Time, actual free time, to spend on the things and people you love.
This is the true beginning of infinity. Not an ending. An opening. The largest expansion of what's possible that our civilisation has ever seen.
And if that doesn't excite you more than it scares you, I'd gently suggest you're not paying close enough attention.