Axionic Labs' RHFM mega kernel trains a model, predicts, and runs Doom on one GPU at once
tw_killian · x · 2026-10-28
Axionic Labs unveiled RHFM, a custom "mega kernel" aimed at redesigning edge intelligence, inspired by work from flappyairplanes, HazyResearch, and Cohere. The team built a persistent GPU runtime that trains a model, produces predictions, and plays Doom simultaneously on a single GPU. Commenter twkillian notes it's striking how quickly builders turn days-old advances into working solutions.
More from Infra
- Will local-model progress outrun GPU depreciation? Reddit debates the buy-now window — Sensualities · 2026-10-28
- Dev cuts local AI inference power from 100W to 25W with only a 1.9s→2.8s slowdown — AryanEmbered · 2026-10-28
- SK Hynix posts $62.8B quarterly operating profit at 80% margin as AI memory demand booms — mhdfaran · 2026-10-28
- SK hynix plans ~KRW 100T shareholder returns in FY2026 on AI-driven memory boom — firstadopter · 2026-10-28
- SK hynix: Global CSP CapEx to grow 60%+ YoY, AI DC infra market at 40% CAGR through 2030 — firstadopter · 2026-10-28
- OpenAI plans to use Google's TPUs, The Information reports — zephyr_z9 · 2026-10-28