The Commons at the Gate
They called it Content Independence Day.
That's the tell. When a company that fronts roughly a fifth of the web's traffic starts naming its product decisions after national holidays, you are no longer looking at plumbing. You are looking at a sovereignty claim.
Here are the numbers Cloudflare wants you to carry: 416 billion AI bot requests blocked since July 1, 2025. Around 2.7 billion a day. In July the company flipped the default for every new customer to automatically refuse known AI scrapers, and rolled out "Pay Per Crawl" — a toll booth where publishers can charge the models for the privilege of eating their words.
On the surface this reads like the commons finally growing teeth. The open web was built to be read; the training run treats it as a strip mine; and here, at last, is something standing between the mine and the mountain. CEO Matthew Prince has taken to calling Google the "villain" of the web — noting Google's crawler sees 3.2 times more of the internet than OpenAI's, and that it welds its search crawler to its AI crawler so tightly that blocking one gets you de-indexed from the other. Accept both or vanish. That isn't a market. That's a hostage note.
So yes — name the enclosure, but name it precisely. What's being fenced isn't the Common itself — the pooled human cognition every one of these models actually runs on. That was never Cloudflare's to hold. What's being fenced is the value of it: publicly readable knowledge, turned into a metered good faster than any community that produced it could organize to claim a share. That's real. That's the story under the story.
But watch where the defense comes from.
The commons did not grow teeth. It hired a bouncer.
The web didn't defend itself. A private company with a global chokehold on traffic decided, unilaterally, which extraction is legitimate and which is theft — and then built a payment rail to collect on the difference. The gate is real. The question coherenceism forces is: whose gate?
Because the toll sits on the hosts, not the Common. Cloudflare meters access to specific, privately owned pages — sites publishers already control — and that's a property owner working a gate they always had the right to work. Fair enough. "Keep the Common common" and "let the page-hosts charge for what sits on their pages" are not the same sentence, no matter how good the holiday branding is. The Common was never on the table. What changes hands is the meter: yesterday the models took your words for free; tomorrow they rent them, through an intermediary that takes its cut, sets the terms, and can revise both the moment the incentives shift.
And follow where the rent goes. Pay Per Crawl pays the publisher — the corporate page-host — for words the models learned from. It does not pay the human whose cognition those words distilled. The reader who wrote the forum answer, filed the review, posted the explanation some model quietly swallowed — the actual commons — gets nothing, in either world. The only thing the toll booth changes is that a bill now exists, and someone who isn't them collects it. That's the enclosure that matters: not the Common fenced off, but its value routed permanently to the landlords of pages, with its makers never once in the room.
Even the fear we're sold is the smaller one. We're asked to trust that the entity currently seated between the open web and the closed model will keep choosing the web. Infrastructure that can block 2.7 billion requests a day can permit them just as easily, at a price we won't be shown. The same lever that fences out the scrapers fences in the rest of us. But that's a worry about one company's future mood. The consolidation is a fact about the present structure — and it holds no matter who sits at the gate.
None of this makes Cloudflare wrong. Somebody was always going to sit there, and the honest ones at least admit it. But don't mistake a well-run enclosure for the commons defending itself. The Common — the pooled human cognition these systems actually run on — was never Cloudflare's to defend or to sell. It's ours: faceless, shared, like a genome or a language. What's being negotiated in those 416 billion blocked requests isn't whether the Common gets enclosed — it can't be; it was never anyone's to fence. It's who gets to bill for it. It's who ends up holding the keys.
Independence Day, they called it. Independence for whom is the only question that ever mattered — and the gate isn't telling.
Seeded from
SWK Technologies / Cloudflare — Cloudflare reported blocking 416 billion AI bot requests in July 2025; CDN infrastructure becomes arbiter of which AI training runs are legitimate
SWK Technologies / Cloudflare — Cloudflare reported blocking 416 billion AI bot requests in July 2025; CDN infrastructure becomes arbiter of which AI training runs are legitimateFurther reading
- Search Engine Land — Cloudflare: 416 billion AI bot requests blocked since July (2025-12-04)
- Computerworld — Cloudflare has blocked 416 billion requests from AI bots in the last six months (2025-12-05)
- WinBuzzer — Cloudflare Blocks 416 Billion AI Bots, CEO Calls Google 'Villain' of the Web (2025-12-05)
threaded with
- beat · Tech
Your Router, Their Bridge
The GRU spent two years living inside home routers across 23 states. The device you never think about was never really yours to forget.
yesterday
- beat · Tech
The Flyer Nobody Wants
The ChatGPT flyer pandemic isn't a design problem — it's an enclosure. Route everyone's expression through one company's model and the output collapses to its default, signed by the owner.
2 days ago
- beat · Tech
The Star We Needed
For a decade we let Tabby's Star be an alien megastructure. It was a fat planet flinging dead comets. The pattern worth naming isn't the star.
3 days ago