Instead of treating code review as a safety net, I try to treat it as a final filter. Over time, I’ve built a simple checklist I run through before I ask a colleague for review.
1. The Context Is Already There
I try not to let reviewers ask “what is this PR about?”. A good PR answers:
What problem am I solving?
Why does it matter?
Why this approach?
What this PR does at a high level
No detective work required, no time wasted in back-and-forth clarifications.
2. AI Review Has Been Done First
In this day and age, I let AI write most of my code, and I act as the gatekeeper to what makes it into the PR. Also, in this day and age, AI does a preliminary code review before I pass it to my colleagues to review.
AI isn’t perfect at reviewing code, but it is pretty good at catching:
obvious bugs & unclean code
inconsistent patterns, coding convention
strange edge cases I didn’t think about
My experience in using AI code reviewers has been pretty good thus far — 95% of what the AI code…
