What it does
The step that turns a week of scattered capture into durable, integrated memory. Mirrors how consolidation actually works — selective (not everything is kept), repeated (built on last week, not from scratch), and integrative (links across days and constructs). It also runs a metacognition pass: not just what happened, but what you learned about how you attend, decide, and learn.
When to use
- The weekly cadence, e.g. Sunday. Triggered by
/review-week. - Any time you want to close a stretch of capture into a kept record.
The procedure
- Recall last week. Read the prior
reviews/note first. Consolidation strengthens existing structure; starting cold loses the continuity that makes the loop compound. - Read the week. Pull
daily/notes in range. Ifweekly/holds intentions you set, hold the week up against them — did attention and behavior match what you meant to do? - Consolidate per construct. Walk the 12 constructs. For each that appeared: what recurred, what changed, what's worth keeping. Recurrence across days is the keep-signal — a one-off thought stays in the daily stream; a pattern that showed up three times earns a place in long-term memory.
- Name learnings concretely (metacognition). A learning is a sentence you can reuse next week. "I focus best before noon and shouldn't schedule deep work after 3pm" is a learning. "It was a hard week" is not. Make them concrete and reusable.
- Carry forward. Pick 1–3 things to bring into next week. Fewer is better. Seed them as next week's intentions in
weekly/. - Hand off retention. Pass any learning worth keeping for the long term to
active-recall-generatorso it doesn't decay. - Write
reviews/YYYY-Www.mdand stop. The plan is the user's to set.
Output shape
Use vault-templates/weekly/_review-template.md as the structure: recap → per-construct consolidation → learnings → carry-forward.
Quality bar
- Built on last week's review, not from zero.
- Keep-decisions justified by recurrence, not vibe.
- Learnings concrete and reusable.
- Carry-forward is 1–3 items, not a wishlist.
Guardrail
Consolidates the week; never diagnoses the person. Keeps observation → interpretation → hypothesis → evidence → decision separate — reports what recurred, offers readings labeled as readings, proposes carry-forwards as testable experiments, leaves the decision to the user. No clinical or diagnostic language. On any sign of crisis, stop the review and point to real human help. See ../../CANON.md.
Built on SIP.