c147 review stop 3: up and down

Written: 2026-11-13, after Phase 4. Reader: Mark, with peterson.md for the conversation with Mark Peterson. Status: at review stop 3 (Q7: the last of three). Each call has a default; a call you don't answer takes it.

What to look at

• Down, on real rows (Priya): Findings, From HQ first · Dana's follow-up (Acknowledge, "This was a mistake") · a suggested training · Settings, her company's default
• Q4 in the prototype's model: S2, open and resolved · S1 on F-E
• The shape: finding-shape.md (§3 the new event, §4 Q4 in the model, §6a what the plane says besides events)

Priya's machine: Teach me is the default her company chose for Engineering; From HQ lists the suggested training and Dana's note; Observed; Prevented as one quiet group.

Dana's follow-up shown as HQ wrote it, with Acknowledge; "Also under" names the same server's other finding; "This was a mistake" is the third thing she can do.

Acknowledge and "This was a mistake" join the history as her own lines; the finding stays open until the plane resolves it.

The prototype's S2 with Q4 in the model: F-D, F-E and F-F read Resolved: prevented from the moment they were detected.

What Phase 4 did

• Q4 is the prototype's model (review stop 2, call 2). foldFinding() in governance-fixture.ts writes a prevented event right after detect for every Enforced finding, so every surface reads it the same way; prevented joins the event kinds and the resolutions; the lab's qdash copy maps it to a block; S2 (mock/findings-design) opens on every outcome with its new subtitle; the native copy follows. In the prototype's world three findings change: F-D, F-E and F-F.
• The down direction on real rows (call 9). app/data/plane-c147.ts holds the plane's moves, authored as the prototype's operator and policy actions and matched to a world's derived findings by policy and resource: c133's training T-MCP-1 suggested on Priya's Linear MCP finding, and Dana's follow-up on her Postgres MCP finding, which asks for an acknowledgement. The derivation appends them as from: plane events.
• What the plane says besides events (calls 5 and 6). A plain "why" for each of the nineteen policies (WHY_PLAIN, with a whyPlain HQ may set, and whyPlainOf()), carried as why_plain in /api/policies; the noise default per team (Engineering on Teach me, Finance and the organisation on Quietly) as /api/plane-settings.
• The person's view (calls 4, 6, 7): From HQ first on the list; on a finding, HQ's words shown as written with Acknowledge, the suggested training opened in the browser (Q5), "Also under", and "This was a mistake" (a reported-mistake event); the first-run card and Settings name the company's default. Every new string voice-checked.
• Checks: c123's metrics 49 of 49; the tours walk 455 of 455. Simulation's check is 184 passed, 6 failed (it was 190 of 190) and the tour drive check 186 passed, 2 failed (it was 187 and 1), as the c144 session measured it; my own run crashed on a page timeout after 102 passed and the same 6 failed. All eight are Q4's (call 1).
• Recaptured on the widened allow list and Q4: the copy (8) and the person's view (10, three of them new).
• For Peterson: peterson.md, nine questions only he can answer, with what to show him.

Calls

1. Simulation under Q4. The simulation's world goes through the same fold, so a beat's blocked or redacted finding is born resolved and leaves the Findings page's Open list, where the beats' test lines and the tour's callouts look: S1's and S4's replays at Enforce (the approval policy), S2's and S5's sanitized values, and the tour's P-MCP group (F-E). Three ways out, from the c144 session: the simulation shows Closed as well (or the tour opens the row where it lands); the test lines expect prevented; the callouts change. Default: the simulation's Findings shows prevented rows beside the open ones, marked Prevented, and the test lines say so; handed to c149, whose session owns the simulation; c144's session has changed nothing and waits for this ruling.
2. The plane's moves as authored data. plane-c147.ts is the prototype's operator, written by hand like the rest of the governance fixture, matched by policy and resource. Default: keep.
3. The per-team noise defaults: Engineering on Teach me, Finance and the organisation on Quietly. Default: keep, and ask Peterson who sets them (peterson.md 6).
4. "This was a mistake" as the quiet third action, after the two buttons. Default: keep.
5. Phase 5's brief. The qnative cycle (types, nozzle, the Findings pages, c4's inbox retired, the recut) is briefed after Peterson, so his answers are in it. Default: yes; Phase 5 starts from review-3.md's rulings and peterson.md's answers.
6. Still open from earlier stops: resolve on training completion (review 1, call 5), chatgpt.com (review 1, call 7), who raises findings on a standalone machine (Tyler, finding-shape.md §6). Default: Peterson for the first two, Tyler for the third.

Answered (Mark, 2026-11-13, walked)