Alex Finn demonstrates a mixture-of-agents mode that sends one prompt to multiple reference models and asks an orchestrator to synthesize the responses. He recommends it for complex questions where several independent approaches can produce a stronger combined answer.
The update also adds a learn command that can turn a URL or completed process into a reusable skill, plus a journey view that maps the connections between skills and memories. Background learning work can move to cheaper models, reducing the cost of creating memories and maintaining skills.
Alex Finn highlights integrated Git controls for coding work, including diffs, commits and pull requests, and explains when he would create a separate Fable 5 profile for demanding build tasks. Promotional discussion of his paid community is omitted from the catalogue record.
Watch on YouTube
