Post #2082841

Another project involving a vibe-coded rust-rewrite of a project that’s already pretty good. Postgres, this time. There’s a graph-goes-up-and-right of a kind I’ve never seen before: llm tokens spent vs test suite pass percentage. But… why wouldn’t you do this with something that isn’t already well supported project involving some extremely complex underlying technology that’s both cryptic and performance sensitive? Presumably the projects that would most benefit from rewrite-it-in-rust don’t have a high quality and comprehensive test suite, and you can’t just vibe up a good test suite, because the technology just isn’t good enough?