You built it with AI. Now it’s stuck.
It ran yesterday. Now it won’t deploy, or it does something you can’t explain, and every fix the AI suggests makes it worse.
What usually goes wrong
Most of it falls into a few buckets. A deploy that fails with an error that explains nothing. Data that disappears or shows up twice. A login that works for you and nobody else. A pile of code the AI wrote that nobody understands well enough to change.
None of that means you did it wrong. It means you reached the part where knowing how the web works starts to matter.
Where to start
I made Anatomy of a Web Request for exactly this spot. It walks through what happens between typing a URL and seeing a page: DNS, the secure connection, the CDN, your server, your data. A lot of stuck projects are stuck somewhere on that path, and once you can see it, the error messages start to make sense.
I’ve been a web developer for 15 years, and I use AI tools on my own projects. I know both halves of the mess.
What’s next
I write about building with AI, including what breaks and how I get it working again. The newsletter is where it goes first.
Start with the request
Free, in your browser, no sign-up.