The lack of fast GPU kernels written by AI does not lend credence to your theory... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

saagarjha 3 months ago | parent | context | favorite | on: CODA: Rewriting Transformer Blocks as GEMM-Epilogu...

The lack of fast GPU kernels written by AI does not lend credence to your theory.

boulos 3 months ago [–]

Perhaps you missed work like https://crfm.stanford.edu/2025/05/28/fast-kernels.html ?

saagarjha 3 months ago | [–]

Comparing against torch.compile is not particularly impressive

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