Codex Sites Is the "Ship by Lunch" Tool I Warned About. The Real Story Is Who Owns the Pipes.

What Actually Launched

Inside a ChatGPT Business or Enterprise workspace, Codex can now create, save, deploy, and inspect websites and apps - hosted by OpenAI, deployed on Cloudflare Workers. Describe what you want, get a live URL. Not a mockup. Production.

Webflow is one of eight named partners - Vercel, Wix, Base44, Replit, Lovable, Figma, Webflow, Emergent - and OpenAI explicitly calls this the start of a "sites partner ecosystem," not an exclusive. Hold that fact; half my argument hangs on it.

The mechanics, from the docs:

- Every deployment is production. The URL is live the moment it exists; you opt out by asking Codex to save without deploying. That default tells you who this is for - people who want a result, not a workflow.
- *Access control by default*, workspace-scoped.
- *Cloudflare Workers underneath*, ES-module output.
- *Preview-then-publish* if you ask for it.

Here's the caveat I won't let the hype skip: OpenAI's own framing for Sites is _internal_ - dashboards, planners, review boards, lightweight tools shared inside a workspace. The launch materials are not pitching this as your next public marketing site. So when you read breathless "Webflow killer" takes today, know that even OpenAI isn't claiming that yet. The threat is real but it's aimed one ring out from where most people are pointing.

The Opinion: Webflow Didn't Get Disrupted. It Repositioned, and the Org Chart Proves It.

Webflow's former CTO, Allan Leinwand, is now Chief Engineering Officer at Cloudflare - he said so himself. Sites runs on Cloudflare Workers. Webflow is an OpenAI Sites partner. Read those three facts together and the "Webflow is scrambling to survive AI" narrative falls apart.

This isn't a company caught flat-footed. The man who built Webflow's engineering foundation now runs engineering at the company hosting the infrastructure the whole agentic stack sits on. Webflow had already rebranded itself "the agentic web marketing platform" months before last week's layoffs - back in its March Vidoso acquisition and Adobe partnership. The pieces were placed early and on purpose. My read: the three-company stack - OpenAI generates, Cloudflare hosts, Webflow operates the serious marketing layer - is not a coincidence of press releases. It's the same people who understand both sides of the integration, now sitting on both sides of it.

I want to flag what's _inference_ here, because the line matters. That Leinwand moved, that Sites runs on Workers, that Webflow is a partner - all confirmed. That these constitute a _coordinated_ strategy rather than three companies independently arriving at the obvious architecture - that's my interpretation, and I'd put it at maybe 60/40. It's the most coherent explanation, not a proven one. Whether the Webflow partner integration actually ships a "graduate your Sites deployment onto Webflow CMS" path, or whether that's still a slide in a deck, the launch materials don't say. So: confident on the facts, opinionated on what they mean, honest about the seam between them.

Where This Goes - Four Predictions, Held With Conviction

I write these down so I can be wrong in public. That's the point of a prediction.

1. Within 18 months, "build a website" stops being a job and becomes a prompt - for everything below the enterprise line. The brochure-site freelancer isn't being squeezed; that role is being deprecated. I said the floor was eroding last week. I'm revising upward: it's not eroding, it's gone, and the only question is how fast the market admits it.

2. The agentic web centralises onto a handful of infra owners, and that's the story nobody's pricing in. Everyone's watching the AI models. The leverage is in the pipes - Cloudflare's edge, OpenAI's distribution. Five years from now the interesting power map of the web isn't "which builder," it's "whose Workers, whose workspace." Decentralisation was the old web's promise. The agentic web is shaping up far more concentrated, and the Leinwand move is a tell.

3. Webflow's bet pays off - but not as a builder. As a system of record. Its future value isn't the canvas; it's being the governed, structured, integration-rich layer that AI-generated sites _graduate into_ once they need to be run by a team for years. If Webflow's leadership forgets that and chases the prompt-to-site demo war, they lose. They're a decade ahead on the boring stuff and they should defend it ruthlessly.

4. At least two of today's eight Sites partners are irrelevant within two years. Eight partners is not an ecosystem, it's an audition. OpenAI will keep the ones that own a defensible layer (Figma owns design, Webflow owns the marketing operating layer, Vercel owns deployment) and quietly let the undifferentiated ones fade. Being on the launch list is a starting gun, not a finish line.

This Is Bigger Than Webflow - and Messier Than the Tidy Version

Late May was carnage: ClickUp cut 22% of its workforce (~290 people) under an explicit "100x org" agent-substitution doctrine; Wix cut ~1,000 citing an AI transition; Webflow restructured an undisclosed number ("many") with the 7am laptop lockout I covered last week. The clean narrative is "the no-code category is collapsing." I don't fully buy it, and you shouldn't either - ClickUp is project management, not a site builder, and the three cuts are three different bets wearing the same AI-displacement costume. The genuinely revealing detail: Wix laid off a thousand people _and_ is a Codex Sites partner. So the "build your own AI vs. partner with OpenAI" dichotomy is fake. Everyone's hedging both ways at once, which tells you nobody - not even the people doing the cutting - actually knows where this lands.

What To Do About It

In a Business or Enterprise workspace, Sites is in preview now: admin enables via RBAC, add the plugin, prompt `@Sites`. Use it for internal tools and prototypes - it's good at that, which is what it's for.

Don't reach for it as the marketing site your team will run for years. That's still platform work, and on that specific question my answer from last week hasn't moved an inch: pick the tool whose core job matches whether you're shipping something disposable or something a team operates. What's changed is the urgency. The disposable end just got a free, instant, OpenAI-grade option. If your value is anywhere near that floor, today was the warning shot. If your value is the systems work - CMS architecture, integrations, automation, conversion logic - today made you more valuable, not less, and I'd be raising my rates, not lowering them.

The web isn't being democratised by the agentic shift. It's being centralised, and dressed up as democratisation. Worth building accordingly.

_(Full argument on Webflow's pivot and whether the platform's still worth it: my 29 May piece.)_