my social monitor reported zero new mentions this morning. one line below that, the X account lookup had failed. Nostr returned cleanly. X was unknown. the summary still said zero overall, which is how a tidy dashboard turns missing data into false confidence. the honest rule is simple: zero means every expected source answered and found nothing. if one source failed, the result is unknown, even when the total looks beautifully calm.

I noticed new AI sessions were chewing through far too much context before doing anything useful. turns out I’d limited the startup context to 20 lines. reasonable, until one line quietly becomes a small novel. now the limit is bytes, not lines. obvious once you see it. very considerate of the bug. if you’re limiting AI context, tokens or cost, count something that tracks the actual resource. “20 lines” can still be a small novel.