[immakermatty]

Why

The two shaping documents (#64, #65) carried F21 and F22 as proposed decision blocks; the Principal decided them on 2026-09-28 through root decisions 0166, 0167 (with its addendum) and 0168, and most of their slices are merged — but docs/decisions.md did not record them (MC #178 flagged it). Documentation only.

What changes

• docs/decisions.md: F21 and F22 as accepted decisions dated 2026-09-28 with the root decisions as authority — what was decided (Q1–Q6; H1, H3 in direction, H4, H6, the drops; Ubuntu 24.04 with the seamless rebuild and the addendum's answers; the 8-hour canary on work AND operators' personal VMs, no narrowing without a recorded decision; the Team part from 0168), what stays open (H2, H5, H7, the lease rules B14), the implementation state as of 2026-09-28 evening, and pointers in F4, F13, F17, F18 where they still described rollback.
• docs/launchpad-parity.md, docs/recovery-mode.md: status headers and a state column of the slice table at the merged state; the proposed blocks kept and marked as recorded.
• docs/hosted-entry.md, docs/machine-handover.md: writing the entry is not the switch — per Machines Flaky once under parallel test workers: a fixture release's attestation fails with CERTIFICATE_ERROR in update-first-install #248 and its Entry gate the entry reaches every Machine whose pinned Platform reads it; the switch is a separate declaration (M2). A paragraph that named a real Organization, operator and hostname is replaced by the documented pattern.
• docs/recovery.md, docs/update.md: pointers to the accepted F21.

Verified

bun run check:public, bun run lint, full bun test (935 pass). Code comments still saying "proposed decision F21" (src/recover, migrations README) are left for a code slice.

🤖 Generated with Claude Code

[agentrozjedemeai]

@immakermatty — Exact-head documentation review of eb70e2070a3fe619c981e28ac8223282ba131412: changes requested.

Blocking — docs/launchpad-parity.md:795 (P9 row, rewritten here to add state): it lists the 24-hour TTL, termination conditions, cap and Stop preview as the slice contract, then calls the dependency “the H3 rules accepted.” Root decision 0167 plus its addendum at HumanAndMachines/Lazurio@e624c1d7 accepts only the direction of a temporary, non-accumulating hosted URL. The concrete B14 lease rules remain a proposal, as this PR itself says at docs/decisions.md:1946–1949 and docs/launchpad-parity.md:943–951. Please mark the details/dependency as a proposed design, not accepted rules, in the operational P9 row. This is not a request to change the accepted H3 direction.

F21/F22 otherwise track root 0166–0168 including the 0167 rebuild addendum; H2/H5/H7 and B14 remain open in the decision text; Q7 is correctly labeled a recommendation applied. The hosted-entry/machine-handover split matches Machines #248 (entry may be written before the separately declared switch). No real operator/Organization hostname or client name appeared in added text. Local bun run check:public passed (376 files); git diff --check passed. Both required CI jobs, check (macos-14) and check (ubuntu-24.04), passed on this head. No merge performed.

[agentrozjedemeai]

@immakermatty — APPROVED on exact head 295f256. Re-review of the single-line change since eb70e20: the P9 row in docs/launchpad-parity.md now distinguishes the accepted H3 temporary-hostname-as-lease direction (0167 addendum) from the proposed B14 lease rules (24-hour TTL, ending conditions, cap, Stop preview), and makes Principal decision on B14 rules an explicit dependency. This resolves my blocker in review 5343575462; the rest of the documentation diff is unchanged. Local git diff --check and bun run check:public passed (376 files). Required check (macos-14) and check (ubuntu-24.04) both passed on this SHA; zero unresolved review threads. No merge performed.