Hacker News: maxignol

New comment by maxignol in "CODA: Rewriting Transformer Blocks as GEMM-Epilogue Programs"

maxignol — Fri, 22 May 2026 07:06:08 +0000

« LLMs can successfully author CODA kernels » That might speed up progress in this area then

New comment by maxignol in "δ-mem: Efficient Online Memory for Large Language Models"

maxignol — Sat, 16 May 2026 20:39:59 +0000

Is there some kind of memory enabling, for instance, an agent to remember guidelines on a repo without having to feed at the beginning of each session 4 markdown files and spending the corresponding tokens each time ?

New comment by maxignol in "SANA-WM, a 2.6B open-source world model for 1-minute 720p video"

maxignol — Sat, 16 May 2026 20:19:30 +0000

I can’t seem to grasp why everyone says only slop gets produced by AI models (and particularly those world models). Imo it’s shit in -> shit out. Great work can be achieved using those. Slop gets produced by careless users.