That’s already largely covered by “it’s reliable,” to the extent that is objective, at least.

Show Post

hooooooly shit

Show Post

The OTHER post I want to write is about “automatic programming” in Hopper’s day, letting us produce code faster than we could understand it. It was a great slopification that followed, and how “GOTO Considered Harmful” and strict control flow eventually tamed it.

Show Post

The metrics you optimize on to reduce code bloat are LOC and # of tests, holding as an invariant full test coverage. I agree with you in one key dimension though: you still need to care about *human legibility* in a codebase. That's something you can optimize away and no LLM can preserve it.

Show Post

The splatting stuff it is whole own amazing world. I promise I'll write a big post soon about the other things but... I got sniped by halloween crafts and catoids. I'll just hand you the papers: arxiv.org/abs/2410.20285 arxiv.org/abs/2310.04406 arxiv.org/abs/2607.25333

Show Post