diagnose

Structured bug diagnosis solves performance regressions by guiding you through reproduce, hypothesise, instrument, fix, and regression-test phases. It requires a disciplined approach and a fast, deterministic agent-runnable pass/fail signal to isolate and resolve the regression.

You build a deterministic pass/fail signal by constructing a fast, agent-runnable test that the structured debugging loop uses to verify hypotheses. This build feedback loop ensures the bug reproduction and subsequent fixes are validated automatically and reliably.

Yes, you can use this structured debugging approach for performance regressions. The diagnostic loop handles both hard bugs and performance regressions by applying the same disciplined reproduce, hypothesise, instrument, fix, and regression-test methodology to isolate the root cause.

You verify bug diagnosis quality during the debugging process using adversarial verification. The system applies a 4-lens verifier for each qualifying pattern to ensure high-quality analysis and prevent false positives before finalizing the fix.

A limitation of this structured approach for software maintenance is that it requires a fast, deterministic pass/fail signal to function effectively. Without a reliable way to automatically reproduce the bug, the disciplined diagnostic loop cannot verify hypotheses or validate fixes.