Engineering articles.
9 articles in engineering.
Agentic Code Review vs Deterministic Quality Gates: Use Each for the Work It Can Prove
Agentic reviewers reason about intent and context. Deterministic gates enforce repeatable facts. Learn how to combine them without duplicate comments, unstable merge rules, or wasted review time.
Read more →AI Code Review Metrics That Expose Noisy Tools
Measure whether AI code review reduces risk and human work. Track verified precision, issue coverage, dismissal cost, duplicates, latency, resolution, escaped defects, and cost per useful finding.
Read more →CodeSlop: Why AI Coding Agents Leave Abandoned Edits Behind
A new study names a specific form of AI code waste: speculative edits that survive after the agent has found the real fix. Here is how CodeSlop happens, how to review for it, and how to keep it out of main.
Read more →A Good AI Code Reviewer Knows When to Stay Quiet
AI code review is useful when it removes work from human reviewers. Here is how engineering leaders can separate signal from activity and measure whether a review bot is genuinely helping.
Read more →We scanned gstack. The score was brutal, but the useful part was the verdict.
A real-world scan of gstack showed why AI-slop tooling needs more than a single score. aislop found two confirmed defects, several conservative security patterns, and a large amount of reviewable quality debt. Then we changed the output so teams can see the difference.
Read more →What I fixed after that score, and what I kept
A clean library scored 1 out of 100, and the score was my bug, not the code. Here is the week I spent fixing it: rule by rule, measured on real projects, including the scoring change I built, liked for an hour, and then reverted because it was lying.
Read more →If your function needs 80 lines, your agent gave up. If your file needs 400, it never started.
If your function is more than 80 lines, your agent is doing something wrong. If your file is more than 400, same story. We found a 981-line component in the wild.
Read more →Agents add. They don't subtract. That's the bug.
Agents add. They don't subtract. The chrome-extension repo went 9 to 68 to 88 almost entirely on dead code removal. Orphan imports. Unused exports. Half refactored functions the agent forgot to delete.
Read more →What 25 real projects taught us about aislop 0.5
Before tagging aislop 0.5 we ran it against 25 real projects from our own backlog. Here is what broke, what got fixed, and why we moved destructive fixes in-house.
Read more →