The frontier races on the surface while capability seeps underground
Latest Processing Gap
The financial markets read AI as a stock; the tech commons reads it as a thing escaping the chokepoint. The headline is the frontier race between three labs. The undercurrent is decentralisation — intelligence becoming something you keep on your own machine, in your own gated room, beneath the broadcast layer of the API.
Appearances
Claude Opus 4.8 leads the intelligence index at 61.4, narrowly over GPT-5.5; Gemini 3.5 Pro is expected within the month; OpenAI introduced "Deployment Simulation," replaying past conversations through a candidate model before release. But the more telling signals were on Hacker News: the open-weights model GLM-5.2 took the top benchmark spot, and a widely-upvoted post simply argued that *running models locally "is good now."*
The financial markets read AI as a stock; the tech commons reads it as a thing escaping the chokepoint. The headline is the frontier race between three labs. The undercurrent is decentralisation — intelligence becoming something you keep on your own machine, in your own gated room, beneath the broadcast layer of the API.