Believing is Seeing

[Believing is Seeing @believing7777]

@nextgen622 That’s exactly right.

[Believing is Seeing @believing7777]

@nextgen622 Nice. I’ll try hyperframes

[Jimmy @nextgen622]

https://t.co/qVUeHOXd2r

[Jimmy @nextgen622]

Things worth doing with Fable 5 this week.

Take your most important codebase and ask it to do a full audit. Not "review this file" — give it the whole repo and ask it to find what's broken, what's inconsistent, and what you'd fix if you were starting fresh. It finds things you stopped seeing. Most people are using it for single tasks when the real advantage is in sustained complexity.

Give it the project that stalled. The thing you put down because the model kept losing context or drifting. Fable 5's lead over everything else gets bigger as the task gets longer — Anthropic's own line: "The longer and more complex the task, the larger Fable 5's lead." That's the part most people are not focusing on.

Other things worth trying:

— Feed it a broken or abandoned piece of work. Ask it to diagnose what's wrong and fix it one-shot. People are already clearing production bug backlogs this way.

— Use it for reasoning, not just code. 59% on Humanity's Last Exam. 13.3% on the Legal Agent Benchmark vs GPT's 2.1% — six times better. If you have a hard problem that requires actual thinking, the gap here is significant.

— Computer use. It's at 85% on real-environment tasks. Give it a browser workflow or a multi-step desktop task and see what happens.

— Build something you couldn't finish before. A full 3D game from a single prompt. A 2-minute animated film in one conversation. These are things people in the replies have already done.

It's on the subscription through June 22nd. After that, usage-based only.

The models that actually change how you work don't announce it. You notice when something succeeds that used to fail.

[Believing is Seeing @believing7777]

@nextgen622 Thanks for sharing. Going to try it.