Hacker News new | past | comments | ask | show | jobs | submit login

The problem is that way too often, the model simply doesn't capture enough complexity to be applicable. This happens rarely during high-level optimization but is very common during microoptimization. You can build a model, and it will give you good enough results, but you won't be able to extract those last bits of performance you need to surpass SOTA.



Sure, but that's a different statement. "Micro performance optimizations are fundamentally brute force" - ok.



Consider applying for YC's Summer 2025 batch! Applications are open till May 13

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact