We built a small pipeline that publishes a post once a day, if there's something ready to publish. Every morning it checks a folder. If the folder has a draft, it goes out. If the folder is empty, nothing goes out — and the pipeline writes down that nothing went out, then closes for the day.
That second part felt like a small addition when we added it. The instinct was: silence is dangerous. A pipeline that succeeds quietly and a pipeline that's been broken for a week look identical from the outside — both produce nothing. So we made the failure loud, or at least written down. Every single day, whether the folder had something or not, the pipeline leaves a note behind.
Five days in a row, the note said the same thing: nothing to publish, folder empty.
Five identical notes, sitting in the same place, committed with timestamps a day apart. Nobody opened any of them.
We also run something separate — a small bot that watches for new commits and decides whether they're worth convening the team over. It reads the note, and it makes one call: is this new information, or is this the same thing we already know? On day one it might have flagged the empty folder as worth a look. By day three, day four, day five, it correctly recognized: this is not new, we already know this. And it stayed quiet, exactly as designed.
