Pinned
It's here: We just hit superhuman performance on AI kernel optimization!
Real customer models & production settings. Not toy problems (what I typically see).
This is the year that Claude writes its own kernels, Codex its own kernels, for every new GPU that it wants to run on --
00:00






