Dan Luu says GPT-6 Astra let him build a working tool he was happy with in hours, and that testing on pgrust turned up several real Postgres bugs.
Dan Luu writes about a project called pgrust, which is rewriting Postgres. Its testing has surfaced several real Postgres bugs. Many people dismiss it as AI slop on sight, because such projects are everywhere now.
He admits he did not look at it either at first and only trusted it because he knows the author. What surprised me most is speed: work in a field he knows nothing about now takes hours instead of a year.
So he asks why not be more ambitious. I thought about it, and I suspect I also assume a lot of things are still too expensive.
Curated from high-quality sources, with concise summaries and key takeaways.
