Karpathy Highlights Claude Fable 5's Leap in Ambitious Software Creation

AnthropicAnthropic

Andrej Karpathy commented on Anthropic's release of Claude Fable 5, a new Mythos-class model with added safeguards. He noted its state-of-the-art performance and qualitative step change for complex, long-duration problem-solving, enabling more ambitious software development tasks.

Andrej Karpathy highlighted Anthropic's Claude Fable 5, a new Mythos-class model with added safeguards and state-of-the-art performance. He described it as a major qualitative step forward for extended, difficult problem-solving and ambitious tasks.
Commentator
Andrej Karpathy
Assessment
Major qualitative step change; SOTA by a margin
Strongest At
Long, difficult problem-solving sessions
Noted Caveat
Safeguards too 'trigger happy' at launch

This model's enhanced capabilities suggest a shift in software creation, with Karpathy observing that working software increasingly "comes out on a tap." This could substantially increase demand for software, as its ease of generation enables new types of applications.

Claude Fable 5 can generate diverse software outputs, including explainers, visualizers, dashboards, bespoke single-use applications, and expanded test suites. It also supports auto-optimization and custom HTML for research projects, despite initial "trigger happy" safeguards.

Andrej Karpathy
Andrej Karpathy
@karpathy
X

This is a super exciting release - Claude Fable 5 is the same underlying model as Mythos but with added safeguards. The benchmarks are great and it's SOTA on everything by a margin but I'll add that *qualitatively* also, this is a major-version-bump-deserving step change forward (imo of the same order as Claude 4.5 was in November), peaking especially for long problem-solving sessions on very difficult problems. You can give it a lot more ambitious tasks than what you're used to, the model "gets it" and it will just go, and it's never felt this tempting to stop looking at the code at all (but don't do this in prod!). The model still has quirks that people will run into and the safeguards are configured to be a little too trigger happy for launch, which can hopefully be tuned over time. I feel a lot of things changing as working software increasingly comes out on a tap. The Jevon's paradox kicks in and I feel my own demand for software growing substantially. You can ask for anything - explainers, visualizers, dashboards, bespoke single-use apps (e.g. a full wandb that is hyper-specific just for your project), you can 10X your test suite, auto-optimize code, run giant research projects with custom HTML for the results, anything! "Free your mind" (Matrix ref). Really looking forward to all the things people build!

1.7kretweets18klikes
View on X

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

Share this update