Claude Fable (Mythos-class model)
- Source
- oneusefulthing.org
- Author
- Ethan Mollick
- Date
Context
Long-running AI work can shift a person's effort toward commissioning and judging the result. In an early-access account, Ethan Mollick describes asking Anthropic's Claude Fable to build a map showing how far someone could travel within a given time. The researched transport data, delegated work to other agents and built the application over several hours.
Mollick still found estimated travel times for remote places and asked for better evidence. The experience illustrates the changed review problem: a user can request a much larger project, but must assess many choices made during its execution. This is a firsthand case study of delegated work and correction, not a controlled comparison establishing general accuracy or productivity.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- guardrails — The checks around a model that block bad inputs and outputs — filters, validators, and permission rules the model itself can't override.
Checking sign-in…
Loading comments…