Set it and forget it. Weekly gc keeps entropy low. Pair with your Sunday routine for maximum effectiveness.
Environments don't stay clean on their own. Toxic patterns develop gradually — too slowly to notice day-to-day, fast enough to do real damage over months. A scheduled scan gives you a regular forcing function to check what's actually in your environment, rather than waiting until something is bad enough to be impossible to ignore.
It's the same principle as a weekly code review, a monthly one-on-one, or a quarterly financial check. You're not looking for crisis — you're preventing it through regular maintenance.
Sunday is the recommended default, and not arbitrarily. The transition between weeks is a natural point of reflection. You're already doing some version of a mental reset — thinking about what worked, what didn't, what next week needs. Slotting a quick scan into that window adds almost no overhead and keeps the data current.
The scan takes as long as your honest attention. Five minutes of real reflection per subject is more valuable than a rushed daily log.
The most common failure mode with any tracking system is inconsistency. You log diligently for two weeks, then stop for a month, then log again after something goes wrong. The result is a dataset that's too sparse and too emotionally loaded to be useful — you end up with a record of incidents rather than a record of patterns.
Scheduled runs solve this by making the cadence automatic. Even if a week is uneventful, logging that normalcy is data. A subject with seven neutral weeks followed by a bad week looks very different from a subject with seven unlogged weeks followed by a bad week.
The best use of a scheduled scan is as the closing step of a weekly reflection, not an isolated task. By the time you open the tool, you should already have a rough sense of how each subject has shown up this week. The tool then takes that impression and turns it into structured data — which is more honest than pure impression and more actionable than pure data.
Some people find it useful to write a sentence or two about each subject before logging the scores. Not for the tool — just for clarity. What would you say about this person's behavior this week if you had to describe it to a friend? That's usually a good proxy for an honest evaluation.
If you run the scan consistently and find that no one is ever flagged, that's not a failure of the tool — that's the goal. A clean scan is a good scan. Environments with low toxicity tend to stay that way when they're maintained. The scan is part of the maintenance.
It also calibrates your expectations. When you know what your environment normally looks like, you notice deviation earlier.