what did your AI coding assistant completely miss while building core logic?
by•
I spent an entire morning building an automated time-zone scheduling tool using an AI coding agent. everything looked polished, smooth, and ready to deploy during local testing.
but the moment a colleague from a different region tested it across daylight saving boundary lines, the entire calculation shifted by an hour. the code was syntactically correct, but the logic completely lacked real-world context.
how do you catch these subtle domain blind spots before launching publicly?
24 views
Replies
Test a weekly 9am New York meeting across the spring clock change. It should still be 9am for the organizer even though the UTC time shifts. That one case catches calendars that store a recurring meeting as a fixed UTC time.
It fixed a failing edge-case test by changing the assertion from strict equality to checking if the array wasn't empty. The suite stayed green for days while the actual bug sat untouched. Test file diffs have to be reviewed before looking at the implementation.
I keep a record of every real mistake, 161 so far, and the rule each one taught me. The trick I learned is that pointing out a case doesn't stop it. The same mistake comes back in another form. What stops it is writing down the rule underneath, like "store times in UTC, convert only on display", and making it a check that runs on every project.