@swyx So true the real winner will be the platform that watches your app like a teammate.
>auto-flags issues
>suggests fixes
>sets up monitoring without extra steps.
I reckon that’s when vibe coding actually becomes reliable for real projects, not just demos.
@apples_jimmy I mean I think Fable 5 is an incredible orchestrator for other models rather than a “does all” kind of model.
But yes I can’t wait to see all the Minecraft knock offs made with 5.6 🙄
So why did Anthropic add guardrails to Claude Fable 5?
Well @thoughtfullab tested AI models to improve a weak AI on a frog puzzle.
Fable 5 worked alone for 17 hours and boosted the model to 68% success (avg. 34%)
Other top models? Only 1–4%.
This is real autonomous self-improvement an AI making a better AI by itself.
Exactly why they released Fable 5 with limits and kept the full Mythos 5 version restricted.
If it weren’t for restrictions, with enough time we could have made a free version 🥲
@testingcatalog This + Antigravity feels like the real shift from using “generate clips” prompts to actually produce videos.
Hopefully longer video generation will be supported with it