Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
62 changes: 36 additions & 26 deletions docs/overhaul-execution-runbook.md
Original file line number Diff line number Diff line change
Expand Up @@ -146,7 +146,7 @@ invalidate these bindings. Section 5.2 records the deployed runtime separately.
| Repository | Durable checkpoint | Protection state |
| --- | --- | --- |
| `lean-eval` | Completion-plan and runbook checkpoint `59c0c18b2d14015589927b6e810386025c93ba4b` | Required `verify` |
| `lean-eval-submissions` | Retained-State binding `36e405e558be69d50e3093d3e188d24d6fc7cfa1`; protected main and replay packet `d26a3090a338358915cc94651ec7efddde71d241`; deployed production Worker `6e0aeb2b5c71fb857f09feff6172c4ee7bdfae08` | Required `verify` |
| `lean-eval-submissions` | Retained-State binding `36e405e558be69d50e3093d3e188d24d6fc7cfa1`; replay packet `d26a3090a338358915cc94651ec7efddde71d241`; protected main `b686eb49743ff5340a34801c0f8f36d921aa9094`; deployed production Worker `6e0aeb2b5c71fb857f09feff6172c4ee7bdfae08` | Required `verify` |
| `lean-eval-leaderboard` | Protected and deployed main `939d69c88292358adf60b124f29605215a1e422a` | Required `build`; exact Pages deployment and live readback complete |
| `lean-eval-state` | Retained-baseline checkpoint `76b3b3e54f4be69161a00cd81576a58df8eae815` | Required `validate`; append-only descendants allowed |
| `lean-eval-state-staging` | Launch-acceptance checkpoint `0849a95026ea3491ec55f1e0ef3b6ff2dff00fd5` | Required `validate`; append-only descendants allowed |
Expand All @@ -158,9 +158,9 @@ invalidate these bindings. Section 5.2 records the deployed runtime separately.

Production serves deployed submissions implementation
`6e0aeb2b5c71fb857f09feff6172c4ee7bdfae08` with durable intake and exactly
the approved lifecycle routes enabled; protected `main` is the
documentation-only replay-packet descendant
`d26a3090a338358915cc94651ec7efddde71d241`. Deployment, CI, readiness,
the approved lifecycle routes enabled; protected `main`
`b686eb49743ff5340a34801c0f8f36d921aa9094` retains the replay packet and adds
bounded exact-shape private poll-response retries. Deployment, CI, readiness,
health, non-mutating authorization denial, and protected-State validation pass.
Staging intake and public lifecycle routes remain disabled, with its promotion
canary enabled. Model consolidation, publication opt-out, production promotion
Expand All @@ -169,17 +169,22 @@ canary, and general replay remain disabled.
The source-App admission repair, private replay stream transfer, and bound
start diagnostics are included in the launch commit. The exact post-repair
private replay canary is terminal accepted; its credential, artifact, and
temporary-executor cleanup readbacks pass. The post-launch replay packet is
bound at protected submissions `main`
`d26a3090a338358915cc94651ec7efddde71d241` to controller source
`6e0aeb2b5c71fb857f09feff6172c4ee7bdfae08`. At activation State checkpoint
temporary-executor cleanup readbacks pass. The post-launch replay packet was
protected at submissions `d26a3090a338358915cc94651ec7efddde71d241`.
Protected main `b686eb49743ff5340a34801c0f8f36d921aa9094` updates the current private
controller binding to implementation `7c79adcfe020c7ab2258662647dd5518358ae075`
and workflow SHA-256
`0d86af8b627ca474fc882ef6f93996a29099d9349e2700d84c5b4aa4acf34c78`.
At activation State checkpoint
`4fae55f7699e80d5b50314cf678bcf6caa020ad8`, the retained queues contained
161 public and 630 private queued entries. Both bounded sustained chains are
active with independent controller variables and concurrency groups. Repeated
exact-packet public and private tasks are reaching terminal accepted State,
performing their required source/credential/resource cleanup, and starting
exactly one successor per lane. The private non-replenishing proof is terminal
accepted at protected State
161 public and 630 private queued entries. The last all-terminal checkpoint
before the current runs, protected State
`e5751a5b22b5663e1cdad7d87e9fa6f551852b40`, public accounting is 23
accepted, one terminal failed, and 150 queued; private accounting is 14
accepted, three terminal failed, one retryable failure, and 621 queued. Both
controller variables are true, and exactly one current run in each sustained
lane is active on protected submissions `main`. The private non-replenishing
proof is terminal accepted at protected State
`24bcb65c8849b95f569c8ee037503049c7fe568f`; its source-free artifact,
credential scrub, sandbox destruction, and temporary-executor deletion pass.

Expand Down Expand Up @@ -683,20 +688,25 @@ The exact post-repair private replay canary is terminal accepted. Its terminal
State, artifact, credential cleanup, and executor-absence readbacks pass.
Retained terminal accounting includes accepted and safely failed executions.

The post-launch replay packet is protected at submissions
`d26a3090a338358915cc94651ec7efddde71d241` and binds controller source
`6e0aeb2b5c71fb857f09feff6172c4ee7bdfae08`. Activation State checkpoint
The post-launch replay packet was protected at submissions
`d26a3090a338358915cc94651ec7efddde71d241`. Protected main
`b686eb49743ff5340a34801c0f8f36d921aa9094` updates the current private
controller binding to implementation `7c79adcfe020c7ab2258662647dd5518358ae075`,
workflow SHA-256
`0d86af8b627ca474fc882ef6f93996a29099d9349e2700d84c5b4aa4acf34c78`,
and bounded exact-shape poll-response retries. Activation State checkpoint
`4fae55f7699e80d5b50314cf678bcf6caa020ad8` materialized 161 public and 630
private queued entries. The first public sustained task is terminal accepted
and exactly one successor continues that chain. Repeated monitored public and
private terminals preserve that invariant and complete the required cleanup.
Private non-replenishing proof
private queued entries. The last all-terminal checkpoint before the current
runs, protected State
`e5751a5b22b5663e1cdad7d87e9fa6f551852b40` records 23 accepted, one terminal
failed, and 150 queued public tasks; it records 14 accepted, three terminal
failed, one retryable `runner_lost` failure at attempt two, and 621 queued
private tasks. The materialized actionable queues therefore contain 150 public
and 622 private tasks. Both controller variables are true, with exactly one
current run active in each independent lane. Private non-replenishing proof
run `33823645564` is terminal accepted at State
`24bcb65c8849b95f569c8ee037503049c7fe568f`, with its redacted artifact and
resource/credential cleanup reviewed. Both sustained lanes remain active on the
same exact baseline. Both controller variables are installed; a future safe
stop deletes the relevant lane variable first, before any other recovery
action.
resource/credential cleanup reviewed.

- [x] Install the production replay credential without exposing its value.
- [x] Merge and deploy the bounded public/private two-lane controller in the
Expand Down Expand Up @@ -840,6 +850,6 @@ Update this table in place; do not append a history beneath it.
| Production launch readiness | Complete | — |
| 4. Launch | Complete: backend `6e0aeb2b5c71fb857f09feff6172c4ee7bdfae08` live with durable intake; leaderboard `939d69c88292358adf60b124f29605215a1e422a` protected, deployed, and read back | — |
| 5. Four-week overlap | In progress; automatic publication, durable server intake, and server-primary entry live; calendar-bound; future cutoff variable installed | Keep issue intake open through at least `2026-09-30T06:57:10Z`; merge and verify the reviewed cutoff guard after the retained drain and before that timestamp; the conditional closure notice matures `2026-09-16T23:06:35Z`, and the canary automatic-release checkpoint is `2026-11-02T03:50:01.002Z` |
| 6. Historical completion | Packet bound; both bounded sustained drains active from a 161-public/630-private activation checkpoint after terminal private proof; repeated exact-packet terminals and single successors pass; final-delta and issue-retirement drafts are open, including the matching leaderboard-copy retirement; not ready; calendar-bound | Complete both retained drains; then merge the audit companion and rebind the submissions final-delta implementation before the cutoff; process the separately bound delta only after issue-intake cutoff |
| 6. Historical completion | Packet bound; retained private migration and all required image/profile work complete; both bounded sustained drains active on protected submissions `b686eb49743ff5340a34801c0f8f36d921aa9094`; final-delta and issue-retirement drafts are open, including the matching leaderboard-copy retirement; not ready; calendar-bound | Complete both retained drains, then merge the audit companion and rebind the submissions final-delta implementation before the cutoff; process the separately bound delta only after issue-intake cutoff |
| 7. Remaining product completion | In progress | Open problems and editorial work are complete; final live release/replay presentation and issue closure retain their calendar, stability, adoption, final-delta, and readiness gates |
| Final audit | Preparatory cleanup complete; final audit pending | Repeat the audit after all phases and confirm only explained launch and retirement work remains |