Six controlled GPT-5.6-sol runs
The same engineering-loop skill lost on a small fix and won on a medium build. Explore the result, inspect the evidence, then run your own replication.
Task-size boundary
Medium implementation
Four browser-game files, ten engine tests, syntax checks, and a post-run evaluator.
What was held constant
Same model, reasoning effort, starting commit, task contract, sandbox, and acceptance criteria. Only repository-skill routing changed.
Acceptance, required checks, and evidence completeness were primary. A cheaper failed run would not win.
No repository skill, the original v0.2.0 loop, and the lean v0.4.0 loop started from equivalent fresh copies.
Token totals are Codex CLI input plus output tokens. Cached input is already included and was not counted twice.
Read this before sharing
Make the evidence better
Fork the fixture, hold the environment constant, report every result, and publish negative findings too.