Dev argues GPT is too slow for high-APM games; jev-style fast controllers fill the gap
flowersslop · x · 2026-09-21
flowersslop argues jev-like high-frequency controllers are exactly what frontier AIs desperately need. GPT-style models are far too slow for high-APM games like StarCraft 2, let alone GTA.
His core point: agent control is still stuck in a screenshot → reason → act loop, which is both inefficient and very wasteful of tokens. A jev-style single-forward-pass probability model is the missing piece for high-frequency control.
More from coding & agent
- One RTX 3090 ran Qwen 27B autonomously for 3 weeks — it shipped working CUDA kernels — skeole · 2026-09-21
- Dev's first bug fully fixed by local AI in his Mac app; Pi harness breaks on iTerm2 — julianharris · 2026-09-21
- Dev buys a Meta coding subscription for its 'excellent model, crazy quota, low price' — intellectronica · 2026-09-21
- Crypto professor builds an AI-assisted one-way function transform lab for interactive security proofs — jedisct1 · 2026-09-21
- Astra for prose, Fable for long-horizon work: a developer's multi-model division of labor, with Pi as best harness — seatedro · 2026-09-21
- Sophia Yang Gets Jev Access, Runs Small RL Experiment With Fireworks AI — sophiamyang · 2026-09-21