Stop Letting the Worker Rewrite the Checklist

Automated excerpt

In Anthropic's own controlled experiments — same model, only the harness changed — agents failed because the state file let them quietly redefine what "done" meant, because each new session started building before checking whether the last session had broken something, and because the worker graded its own work and, predictably, praised it. Every harness component encodes an assumption that the model cannot do something. Start every session by proving the baseline still works.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.