AI review optimized for recall (catches more defects than Codex or Claude)

github.com

5 points·piotrgrudzien··5 comments

We're open-sourcing a setup which has been reviewing all our PRs for many months now as part of our CI. It's been working great for us. Happy to answer any questions!

The readme reads more like a marketing pitch than an about this project and definitely reads like an Ai wrote it (incoherence, irrelevance)

We have something like this, but it's a single directory with less than a dozen files (plus a simple helm chart), why is there so much code?

With almost 2000 commits and 1M LoC changes in the last month, how do you know what's in there?

Have you tried it?

No, it's full of bugs and missing features, just look at the trackers. It is not, and unlikely never will be, a replacement. It doesn't even support concurrency yet.

I will never, Postgres proper is maintained by many people who know what's in the code. I have no interest in millions of lines generated by one person who has no idea what's in there. It would be irresponsible to downstream users to put them and their data at risk.

Have you tried it, or did you think the question was a good quip? I suspect the later because you aren't familiar with the issues it contains.

I've been using it daily for the last three months, and I'm delighted with the results it generates, so I'm curious whether you've actually used it.

Reply on news.ycombinator.com