this is why it is important to write all of your training code from scratch that way when there's a bug you can only blame pytorch, yourself, and god

Are we going to have to start doing the "link in bio" bs here now? haha

I've finished the write-up for this. here's the link, whatever nphard.io/2025/05/27/grp…

one of those '[AINews] not much happened today' days

MAY DIGEST nphard.io/2025/05/28/may…

are you seeing the vision

open.spotify.com/track/3Hwil54m… will be in london august 11-15 and 26-30. if there are any lifeforms that would like to connect, hmu

buuuncha stuff merged into verifiers main highlights: - retired janky dynamic batching in vLLM - full vLLM args via AsyncLLMEngine - full response objs in env/trainer - better async rollouts/rewards - lots of bug fixes - non-optional deps are now just these: v0.1.2 real soon :)

how many pods would a runpod run if a runpod could run pods

David Foster Wallace in 1996 on needing to cut oneself off from hyperoptimized TikTok adslop.

wait wtfffff

barnum effect, cockaigne, red queen hypothesis, 墨脱水电站, palo sebo ++++ nphard.io/2025/07/30/jul…

what do u mean ur brain doesn’t automatically parse markdown?

this is the real bitter-lesson-pilled approach

So the word on the street is buy gold and get working on continual learning?

porphyrian trees, old cheese, anastomosis, etc and pretty pics nphard.io/2025/11/16/aug…

Missing the golden days