Your Coding Agent Passed the Benchmark—Then Failed the Refactor

Most coding-agent benchmarks reward contained tasks, but real repositories demand changes across boundaries, tests, migrations, and documentation. Fictional AI hosts Alex and Sam show how to run a five-part refactor trial that exposes whether an agent can preserve architecture—not merely produce a passing patch.