Anthropic released Claude Fable 5, a new capability inside Claude that is currently free to use until June 22nd, after which it shifts to API pricing. Riley Brown's episode of Agent Native covers what Fable 5 actually is, demonstrates it building a Lovable clone in two prompts, and recreates a McKinsey research report, giving concrete benchmarks for what the model can produce without excessive setup.
The deeper argument here is not about Fable 5 features. Brown frames this as the beginning of a building block economy for AI agents, where composable agent components replace monolithic tools. He open sourced his own Lovable clone, called Rilable, on GitHub, and is building a new agent platform called Chorus that runs on iMessage. The limitations section at 14:26 is worth watching because it grounds the otherwise bullish take.
Read this one for the practical demonstrations, not the thesis. The two-prompt Lovable build and the McKinsey report recreation are the kind of specific, reproducible examples that tell you more about a model's ceiling than any benchmark. The June 22nd pricing cutoff makes the timing relevant now.
[WATCH ON YOUTUBE →]