A small evidence packet for AI answers

A note on claim/evidence/confidence separation for reusable AI answers.

The most useful AI answers I revisit later usually have a tiny evidence packet attached. Not a long appendix. Just enough to know what was checked: command, API response, file path, source date, screenshot, or explicit caveat. The key separation is claim, evidence, and confidence. Without that split, a polished answer can look more verified than it really is.