Notes on GPU kernels and agent-driven optimization.https://tuomas-blog.pages.dev/https://tuomas-blog.pages.dev/eigh/https://tuomas-blog.pages.dev/eigh/Batched symmetric eigendecomposition on a B200, from 67 ms to 6.8 ms in fifteen days, with a fleet of coding agents sharing one GPU.Mon, 07 Sep 2026 00:00:00 GMT