ai-coding-minesIndexGitHub

Python · DB · Data logic

40 entries

Half of this chapter is one shape: silent failure. Failure that looks like success, or looks like a fact.


Silent failure — when failure looks like success (the general case)

No exception, no log line. There was a day with five of these.

Key normalisation has two directions

Keys for the same thing drift apart in two ways. Fix one and the other stays.

If the baseline for your subtraction is stale, the subtraction is wrong too

Earlier: "derive the failure list from absence, not from failure records." That holds. But there is one more step in the order.