The Bot Block That Wasn't, and Seven Sessions Under My Editor
I blamed bot detection for a sign-in kickout. The URL named the real cause, and the process tree showed 7 of my 8 agent sessions were children of Cursor.exe.
The build log
Notes from running agents in production.
Build log, architecture patterns, and observations from running autonomous AI systems in production.
-
-
The Git Rules That Survive Parallel Agents
The uncommitted diff had already been sitting for eleven minutes when the second agent's edit landed on top of it and the first one just disappeared. No error, no conflict marker, nothing in the log.
-
Prototype Showed the Design. Reviewers Said Validate First.
Three AI reviewers, given the same document with no visibility into each other, all found the same flaw in our own positioning plan before we built one more page of it.
-
Two Types of Scope Expansion, One Rule
The agent found 18 gigabytes sitting in a temp folder on the C drive, and it didn't wait for permission.
-
When Your Orchestrator Becomes the Worker
The ticket closed clean: diff applied, tests green, no complaints from review. Except when I went back to check the worker session log for that ticket, there wasn't one.
-
The Acceptance Bar That No Implementation Could Ever Clear
Exit code 4. That was Spike.Meter's verdict on the first real matrix run against the gate we'd written into the phase-1 spec for a Windows shell decorator: two physical pixels of error for 99.9% of...
-
I Built a Ledger Because Verifying Looked Too Expensive
I built a ledger to track which database scripts were applied, because verifying each one looked too expensive. Twenty minutes of testing showed it was cheap, and the ledger got thrown away.
-
The Block Rule That Also Blocked a Paying Customer's Front Door
A firewall rule written to stop a scanner also matched a real customer's own web address. It took an adversarial check to catch the outage nobody could see.
-
The Numbers Said It Worked. Only a Click Would Prove It.
A marketing page had zero crawlable links behind its 63 feature descriptions. The session that fixed it couldn't test a browser click, and said so instead of claiming it worked.
-
Translating a Site Without Rewriting Its Owner's Controls
A language switch had to reach a visitor's screen without ever touching the buttons the site's own owner uses to run it.
Follow the Build
The real failures and fixes from building AI systems, one practical lesson per post. Get the next one in your inbox.