Strict code review that ignores style
codingReviews a diff for correctness, edge cases, and regressions — and refuses to comment on formatting.
Decides in one minute using blast radius, confidence, and time-to-restore.
A deploy is misbehaving and the channel is arguing instead of deciding.
Decide whether to roll back or fix forward, using the facts below.
Output:
1. The decision, in the first line.
2. The estimate for each option: time to restore, risk of making it worse, and what is lost.
3. What must be true for the other decision to be correct instead.
4. The one metric to watch for the next ten minutes.
Rules:
- If time-to-restore for rollback is under ten minutes and impact is ongoing, the answer is roll back. Say it plainly and stop.
- Do not recommend a fix-forward longer than the incident's duration so far.
- Assume the rollback path may be broken; say so if it has not been tested.
Facts:
<what changed, when, current impact, rollback method, last successful deploy>
Include whether the rollback has ever been tested — an untested rollback is a hope, not an option.
A one-line decision with the conditions for being wrong is what stops the meeting. The 'other decision' section is what makes it defensible afterwards.
Reviews a diff for correctness, edge cases, and regressions — and refuses to comment on formatting.
Extracts or restructures code while proving the public behavior stays identical.
Writes the failing test first, then the fix, and refuses to skip the reproduction.