CyrilXBT|Nov 18, 2026 14:46
Alibaba just open sourced the code review tool it's used internally for 2 years
and on its own benchmark it beat claude code on precision
using about 1/9 of the tokens
here's the trick
most AI reviewers are just a long prompt
on big PRs they skip files, comments land on the wrong lines, and quality changes every time you tweak the prompt
open code review hardcodes the boring parts
the system picks the code, bundles related files and matches the rules
the AI only makes the final judgment call
what it does:
→ comments on the exact line
→ built in checks for null pointers, thread safety, XSS and SQL injection
→ reads whole files and searches your codebase for context, not just the diff
→ can scan an entire repo, not just PRs
→ works with claude code, codex, cursor and more
→ runs in github actions, gitlab CI and gerrit
the tradeoff: it finds fewer issues overall
but when it flags something, it's far more likely to be real
free. apache 2.0. bring your own model
Share To
Timeline
HotFlash
APP
X
Telegram
CopyLink