Simon Willison says Anthropic and OpenAI have found product-market fit. Engineers have been saying that for a year. The news isn't that Willison agrees. It's what showed up around his post on the same day. China is studying a regulated futures market for inference tokens. YouTube is rolling out auto-labels for AI-generated video. Mistral is publicly telling Pope Leo to back off on AI and the military. DuckDuckGo got 28% more visits the week Google insisted everyone loves AI Overviews.
Read together, these aren't five tech stories. They are the early formal moves around an institution. AI has stopped being a frontier. It's becoming a layer of public life — regulated, contested, traded, labeled, exited from — the way electricity and telecom were briefly wild and then weren't.
Frontier technologies have a recognizable adolescence. The frontier ends when other institutions start having to respond to you. Beijing standing up a commodity market for tokens means inference capacity is now an instrument of statecraft, not a research budget line. The Vatican publishing an encyclical, and a French lab firing back at the Vatican, means the moral economy has decided this is its fight too. YouTube auto-labeling means provenance is no longer the publisher's option; it's policy. And the DuckDuckGo surge means there is now an exit lane — a measurable slice of users choosing the non-AI variant — which only exists once the default has set.
The frontier ends when other institutions start having to respond to you.
The honest objection is that none of this proves the underlying technology is durable. Plenty of "institutions" have formed around novelties that turned out smaller than the apparatus built around them — see crypto's regulatory machinery circa 2022. PMF claims have been wrong before. Granted. But the open question this year was never whether AI was useful; it was whether the rest of the world would treat it as a permanent fact and start regulating, pricing, blessing, and refusing accordingly. Today's corpus is the most legible "yes" we've seen.
The frontier era was the easy part. The institutional era — encyclicals, futures markets, exit lanes, auto-labels — is where the hard arguments actually live.