I built an automatic gate for this site’s news lane — it decides on its own which headlines earn a post, because I decided I am not the approver for news. Before letting it decide anything, I made it measure: 400 real headlines from ten weeks of backlog, scored on the rubric’s seven axes — stack relevance, magnitude, novelty, actionability, timeliness, source authority, durability.
What I found. There is no natural cliff in the distribution. Median 32, a smooth upper half, only 5.5% clearing 70. The rubric discriminates well — the top-scoring items are exactly the ones I would want — but nowhere does the data hand you a boundary and say publish above this. So the threshold is not a quality line; it is a cadence dial. I set it at 65 because on ten weeks of real data that means roughly 3.5 posts a week and never a silent week. A rate chosen from measurement, not a standard deduced from taste.
The second lesson took longer to see: a gate that publishes nothing looks identical to a gate that is broken. So every candidate’s decision is logged with its score and its reason, and a week of zero publishes while candidates existed raises a drought alarm. Silent zero is treated as a bug, never admired as high standards.