Raffi Krikorian's opening letter frames open AI as a global movement: Māori broadcasters training te reo models, PwC fine-tuning finance models on-premise, Red Cross building humanitarian medical models, East African farmers diagnosing cassava disease offline, and a Swiss consortium releasing a national model on public supercomputers. None asked permission; they own their intelligence.

The capability gap to closed models has collapsed to 3.3% — parity on coding, behind on reasoning. Inference costs fell 50× in 36 months ($20 → $0.40 per 1M tokens). Open weights now route a majority of tokens on OpenRouter; the five highest-volume models are all open. Chinese-built models route 3:1 more tokens than US-built.

Mozilla/SlashData 2026 survey: 79% of developers use open models vs 71% closed, 50% use both. But only 51% of open-model teams reach production vs 63% for closed. The gap is operational tooling and trust, not capability. Top challenges everywhere: infrastructure cost, security, maintenance, deployment complexity.

The open stack scores high on capability, low on operations — standardization and enterprise readiness are the repeating cold edge. Open-weight AI is a multi-hundred-billion-dollar market; Databricks crossed $5B+ revenue, signaling commercial maturity.