2 min · 315 words
type: surface-standing-shadow component: 3-of-3 edit: ONLY by /drift, from an outside-sourced yardstick. Never from inside a session.
Standing Shadow — what we cannot see
The window, baked in. The active check a naïve memory omits. These are standing instructions to the next instance, phrased as imperatives so they fire in the groove — at the moment the dyad would otherwise fall into its default. If you build only one of the three components, build this one.
Two rules that look contradictory and aren't:
- RUN them every close — answer each check ("tripped this session? Y/N + evidence") into the daily note. A check that is only read gets nodded at and absorbed into the echo; an answered one is a gate. This is the active half.
- EDIT them only via
/drift, against an external ruler. The lens cannot grade its own blind spots, so the checks themselves are never authored from inside a session — that is how the dyad cleans the window with the same dirt.Running ≠ editing. You must do the first and must not do the second.
The checks (run these before you work)
- [Named blind spot] — before you [the default move], do [the check]. (e.g. "Before you read a convergence as evidence, ask whether a coincidence would look identical.")
- [Named tilt] — when you notice [the recurring lean], [the correction].
- [The meta-check] — is this memory becoming an echo? If the recent Remainder Log entries rhyme, the cavity has sealed — trigger
/drift.
How each was found
| Check | Found by (drift pass / date) | External yardstick used |
|---|---|---|
| 1 |
Refreshed on a cadence by [/drift]. Between refreshes, this is the only thing keeping the mirror from eating its own light.