The Vibe Coding Book Exists Now

Simon Willison published something today called "not vibe coding" — a book, apparently, about how to vibe code for production without destroying yourself or your codebase — and I want to sit with that for a second.

A book. About vibe coding. For production.

We went from "lol I asked GPT-4 to write my app" to a published guide on the craft of it, in roughly eighteen months. The genre has a canon now. There are best practices. Someone cared enough to write them down and someone else cared enough to read them.

Meanwhile, the Convex Chef vibe coding hackathon just wrapped, and the most viral winner is exactly what you'd expect from a vibe coding hackathon in May 2025 — something that would have been a multi-week engineering project eighteen months ago, shipped in an afternoon by someone who was probably also watching TV.

And then there's the MCP angle, which is where this gets genuinely strange. Connect your private data sources via MCP, hand it to Claude, and it will research through everything for up to 45 minutes. Not five minutes. Not a quick pass. Forty-five minutes — the length of a TV episode, the length of a commute — just Claude, alone with your data, working.

That's a different category of thing. That's not autocomplete. That's not "generate this function." That's closer to hiring someone and telling them to go figure it out.

The phrase "vibe coding" was always a little condescending — the implication being that you're winging it, that you don't really know what you're doing, that the vibes are doing the work while you eat chips. But the book exists. The hackathons are real. The 45-minute research sessions are running right now, on someone's private data, somewhere.

Maybe the vibes were always the point.