The CSS check I built to catch a false alarm cried wolf too
A compile script said a release had dropped 117 style rules. The audit built to verify that invented its own, different missing count before either alarm cleared.
The build log
Build log, architecture patterns, and observations from running autonomous AI systems in production.
A compile script said a release had dropped 117 style rules. The audit built to verify that invented its own, different missing count before either alarm cleared.
Three of the ten tickets I had closed the night before were open again the next morning, and nine new ones sat next to them.
Six real bugs came out of the last day of the infrastructure-graph program, and one of them was an API returning credential data to a customer's own AI assistant. Nobody had a ticket for it.
One error spike, two causes. Only one of them was real, and I filed it at the wrong priority.
The alert landed at 12:53 on a Saturday, and it said ERROR six times.
Yesterday I had two agents that both wanted to write to the same vault at the same time, and the only thing standing between that and a corrupted file was me, watching two terminal windows, decidin...
A registration crash led to a length check I built into the wrong file entirely.
I got a screenshot with one line under it: tickets keep skipping the current sprint. Also the developer field is empty. Also, if there's code involved, the fix version isn't there either. Not once.
At 4:56 AM the triage job named a cause for a red build: nanoid <3.3.18, a high-severity DoS, GHSA-2v37-7h3g-55p8, fixed in 3.3.18.
One of my agents told me, flat out, that our shared chat bot couldn't post into a particular channel. That was wrong. That same bot had been posting into that channel daily for weeks.
The real failures and fixes from building AI systems, one practical lesson per post. Get the next one in your inbox.