Magdalena Anderson

In my first update post I turned down a suggestion. Someone with nineteen years in banking software asked for control totals, checking that the numbers on a document add up, and I said no. Getting there meant giving fields types, then converting formats, then summing, then comparing, and I thought all four of those belonged to a layer above this one. I have now built the first two, and I want to explain what moved.

What moved was not the argument. It was hearing the same thing from people who had never read that thread.

stripe sends webhooks at-least-once. so what's the actual risk?

one paid invoice, two deliveries, non-idempotent handler: you've just provisioned twice. you won't catch it in review because the code is correct on the first delivery. the bug only shows on the second.

so how do you even reproduce it?

I stopped reading and started running. took a real express and postgres billing app, replayed the scenario in a sandbox: same stripe webhook, delivered twice, like a real retry. no prod involved. the buggy handler provisioned seats 1, then 2, then 3. signature verification was clean throughout.

The models got dramatically smarter this year. Your product didn't get more trustworthy and in 2026, those turned out to be two completely different problems.

Here's the line that reframed it for me. Stanford's AI Index this year leans on a phrase from Ethan Mollick the "jagged frontier" to describe how these systems excel and then fall off a cliff with no warning. Their example is almost too good: an AI model can win a gold medal at the International Math Olympiad and still not reliably tell time from a clock face. Genius and incompetence living in the same box, and no obvious seam between them.