5º. Heat scoring: teaching your data to forget (gracefully)

Here's something that bothered us early on: a search for "project meeting" returned every meeting note we'd ever written, sorted by relevance. The note from this morning and the one from eight months ago scored identically — because semantically, they're equally "about project meetings." But they're not equally useful. The one from this morning matters right now. The one from eight months ago is a fossil. We needed a way to express that difference without breaking search when someone genuinely wants to find old records. The solution was to give every record a temperature. The concept Every record in every domain table — notes, events, contacts, emails, files, diary entries — has a heat score. It's a number between 0 and 1 that answers one question: how relevant is this record right now? Heat rises when you interact with a record (open it, edit it, create it). Heat decays exponentially over time when you don't. A note you wrote this morning is hot. A note you haven't touched in six mont