Skip to content

Learn · Debugging

From hypothesis to defensible root cause.

A methodology for production debugging in the AI era. Six modules. Each paired with replay-reviewed challenges on the platform.

Modules

Six modules.

01

Hypothesis before search

The single biggest mistake in AI-assisted debugging: searching before forming a hypothesis. A protocol for not doing that.

02

Bisecting with intent

git bisect, but for behavior. Working a bisect against production telemetry instead of unit tests.

03

Memory leaks in long-running services

Node and Python. Heap snapshots, retainers, the classes of leak AI usually misdiagnoses.

04

Race conditions and silent loss

Why concurrent code resists AI suggestion fixes — and what to do about it.

05

Distributed debugging

Correlating across services with traces, structured logs, and the rare time it's worth a debugger.

06

Defending the fix

A protocol for writing the PR description that survives a review six months later when the bug recurs.