09 — EXTRACT LESSONS LEARNED
Routes IP from this session to cyp/tpc/lessons-learned/ for use in Practice Builders OS — General. The mechanism that turns TPC into a live IP source.
When this step runs
After open-actions is updated (file 08). Auto-advances from there.
Why this step exists
TPC is the live diagnostic engine for the broader value ladder. Things that surface in 1:1s with TPC members — what works, what would be done differently, what's universal vs. tax-specific — feed Practice Builders OS — General's design.
Without this step, lessons stay in Kathryn's head. They get lost. The cohort being designed for PBOS-General loses the empirical grounding TPC provides.
Output
File: cyp/tpc/lessons-learned/[lesson-name].md Naming: Topic-named, not date-named (e.g., lesson-onboarding-knowledge-stays-in-head.md, not lesson-2026-04-29.md). Lifecycle: Living docs. New evidence appended as the lesson recurs across members and sessions.
The test for a lesson
For each notable observation in the recap, ask:
Would this be useful as a baked-in framework, playbook, safeguard, or cohort-design choice for Practice Builders OS — General?
- Yes → extract as a lesson
- No → leave it in the recap, don't extract
Lessons are NOT:
- Member-specific situations (those go in the recap and member CPM)
- Operational follow-ups (those go in open-actions or drafts)
- Universal truths everyone already knows
- One-off oddities
Lessons ARE:
- Patterns that recur across members
- "If I were designing the next cohort fresh, I'd handle this differently" insights
- Framework gaps in PBOS-General that this session reveals
- Safeguards that, if missing, would have prevented this issue
- Member behaviors or cohort dynamics that the kit/program design needs to anticipate
Lesson file template
# Lesson: [Short, clear lesson name]
**First captured:** [YYYY-MM-DD]
**Source pattern:** [one sentence on the underlying pattern]
**Routes to:** Practice Builders OS — General [framework/playbook/safeguard/cohort-design]
---
## The lesson (one sentence)
[The takeaway. Should stand alone.]
## What it changes for PBOS-General
[How the program design / curriculum / cohort onboarding / kit content needs to reflect this.]
## Evidence
| Date | Member | Observation | Source |
|------|--------|-------------|--------|
| [YYYY-MM-DD] | [Name] | [What surfaced this session that demonstrates the pattern] | [recap path] |
## Status
- [ ] Pattern observed in 1+ session
- [ ] Pattern observed in 2+ sessions or members (graduates from "watch" to "lesson")
- [ ] Translated into PBOS-General design decision
## Linked PBOS-General work (when applicable)
- Concept brief / kit / curriculum doc the lesson informed (link)
---
*Lesson maintained by /extract-lessons-learned step in cascades.*
Workflow
- Read the recap
.mdfrom this session. - Read existing files in
cyp/tpc/lessons-learned/so you know what's already captured. - Scan the recap for observations matching the test criteria above.
- For each match:
- If a matching lesson file exists → append the new evidence row to the table, update "First captured" if appropriate, advance the status checkbox if 2+ data points exist.
- If no matching lesson exists → create a new file with the template above. One data point.
- Don't force lessons. A session can produce zero lessons. That's fine.
Quality gate
- Every lesson passes the test ("would this change PBOS-General?")
- Every lesson has at least one specific evidence row
- No member's name appears in a lesson title (lessons are about patterns, not people)
- File names are slugified topic names, not dates
- Member identification in evidence rows is OK (it's the source — not the subject)
Confidentiality / framing rule
Lessons are private notes that translate into PBOS-General design. They reference TPC members as the source of the observation but the lesson itself is structural — not "Bev was bad at X" but "members will struggle with X if the cohort doesn't address Y." Frame structurally.
Auto-advance
After extracting (or confirming nothing to extract), auto-advance to /cascade-report-1on1 (file 10).