Triton creator Phil Tillet on Gluon: handing GPU decisions back to AI models
TheTuringPost · x · 2026-09-10
The Turing Post launches its "AI Builds AI" series with an interview with Phil Tillet, creator of Triton, on why OpenAI is handing GPU-level decisions back to models via Gluon.
- Core thesis: AI is already helping improve the systems it runs on, and new programming tools like Gluon give models more low-level control; cheaper implementation makes previously impractical experiments worthwhile.
- Key judgment: whether this becomes recursive self-improvement depends on whether these gains make the next round of research more effective. Choosing worthwhile problems and reliably checking results remain the hardest parts.
A substantial deep-dive worth following for anyone tracking kernel programming and the boundaries of AI self-improvement.
More from Infra
- colibri: Pure-C Zero-Dependency Engine Streams MoE Experts From Disk to Run Frontier Models Locally — JustVugg · 2026-09-10
- HBM seen topping 40% revenue share next year as Samsung challenges SK Hynix's lead — zephyr_z9 · 2026-09-10
- A quiet CPU shortage is hitting cloud users as hyperscalers shrug, devs report — mgill25 · 2026-09-10
- "Going for infinite energy": the Kardashev II mindset meme in AI circles — MarvinTBaumann · 2026-09-10
- Redis LangCache claims 90% LLM cost cuts; KV, prefix, prompt and semantic caching explained — blaizedsouza · 2026-09-10
- Raspbian creator laments $300 Raspberry Pi as AI demand prices out young programmers — evilsocket · 2026-09-10