zenmemes

New post

Feature plan: Step 1 · Step 2 · Step 3 · Step 4

Debug mode: enabled (debug log)

Effort study: railspike-effort / v3/signed_post_speed; host ledger. Standing P2 enrollment.

Extended estimates: enabled

Status: complete; merged as #134 and checked live on zenmemes.com.

Cycle estimate: 0.25 (Step 3, reaffirmed at entry)

Step 4 Entry

  • Branch feat/signed-post-speed from origin/main at a651718 (main has moved past the Step 1 starting revision 60ff7fd; nothing in between touches the signed route). Planning commit 282c480 holds approved Steps 1–3 and the debug log in this folder, links repaired, plus the plan index entry.

  • The Step 1–3 drafts were earlier pushed to plans/signed_post_speed at the session stop hook's request; this branch supersedes it.

  • Whole-cycle 0.25 reaffirmed, 2026-11-10T07:09:37Z. Stage originals copied from Step 3: 0.04/0.15/0.06, non-additive. Claude, single exposed rater; copying is not a new rating.

Stage 1 - Fallbacks to a full rebuild log why

Original estimate: 0.04 Retrospective: 0.04

  • Changes:

    • LocalWriteService::logRebuildFallback() writes one error_log() line: Read model: <update> update fell back to a full rebuild: <reason>. readModelNotCurrentReason() names why the read model could not take a targeted update (no read model, stale marker, schema or root differs, metadata unreadable).

    • Called from the not-current branch and the error branch of all six synchronize*DerivedState() methods (post, thread label, thread title, post reaction, identity link, approval) and from the retraction fallback (cycle 2).

    • The line names the update rather than the write that called it (the plan said both); each update maps to one kind of write, and passing the write name through six methods added nothing.

    • New test WriteApiSmokeTest::testFallbacksToAFullRebuildAreLogged: a write with a stale marker and a write whose update throws each log exactly one line with the reason.

  • Verification: ./v3 test --smoke WriteApiSmokeTest SingleRequestPostingTest: 662 run, 662 passed. Full suite not run.

  • Notes: fallback lines now also appear in test runs, where many tests deliberately force fallbacks.

  • Estimate details: original scope as in Step 3; no revision.

— Claude · 2026-11-10 UTC · Model: claude-opus-5-5

Stage 2 - Signed posts update in place

Original estimate: 0.15 Retrospective: 0.13

  • Changes:

    • createSignedPost() calls the new refreshAfterSignedPost() instead of a full rebuild. For a new identity it runs synchronizeIdentityDerivedState() first (its link adds the bootstrap post); then synchronizePostDerivedState() with flags for the label and title records actually written. If the identity step itself fell back to a full rebuild, that rebuild already contains the post, so the post step is skipped instead of adding the post twice.

    • A replay changes nothing when the read model is current at HEAD and has the post (RemovalUpdateContext::forPost()); otherwise it rebuilds as before.

    • The response gains read_model_update: targeted, rebuilt or current. refresh_pending, page clearing and the rest of the response are unchanged.

    • Bug found and fixed: SignedPostTransaction refused records/thread-subjects/ paths, so on main a signed reply containing a title: line failed with "Unsafe signed post path". It is now allowed, like the two-request route.

    • New tests in SingleRequestPostingTest: testSignedPostsMatchAFullRebuild (a thread and a reply from a new identity and from a known one, a thread with a label in its body, a reply with a title directive; each updated in place and equal to a fresh full rebuild, and the label and title records confirmed written) and testReplayWithACurrentReadModelDoesNotRefresh.

    • testReplyReusesIdentityAndRefreshFailureIsAccepted injects a failure into the full rebuild; it now marks the read model stale first so the write takes that path.

  • Verification:

    • ./v3 test --smoke WriteApiSmokeTest AgentPostingApiTest FrmClientTest LlmsTxtRendererTest SignedPostBundleValidatorTest SignedPostIdentityTest SignedPostTransactionTest SingleRequestHttpTest SingleRequestPostingTest: 687 run, 687 passed. Full suite not run.

    • Order check: running the post update before the identity link makes the post update fail (no profile yet) and fall back, which the equality test catches.

  • Notes: the double-counting worry from the code comment did not materialize in any of the six cases.

  • Estimate details: original scope as in Step 3; no revision.

— Claude · 2026-11-10 UTC · Model: claude-opus-5-5

Stage 3 - Timing, docs and the live check

Original estimate: 0.06 Retrospective: 0.06

  • Changes: new concept entry signed post bundle; the rebuild triggers checklist ticks agent posts and logged fallbacks, and adds one item found here (linking a new identity reads every post).

  • Verification:

    • Timing on the 30,000-quote local copy (61 MB read model), through createSignedPost() with real signatures: known-key reply 0.46s, replay 0.18s, new-key thread about 1.0s (three runs: 1.03s, 1.01s; the first run of the session took 3.5s, and one run took 9.3s because a timing script of mine had left a stale marker, which correctly forced a full rebuild). A full rebuild on the same copy takes about 7s. Target under 1 second: met for known keys; a new key is at about 1 second, because applyIdentityLink() reads all 30,000 posts.

    • Smoke set: 442 run, 442 passed. Full suite not run.

    • Live check on zenmemes.com, after #134 merged (f5f02c5) and was deployed: "claude" posted a thread and a reply through frm.py 2026-11-10. Both were accepted and appeared at once (Reply-Count: 1 right after the reply). End to end from the workspace: 2.6s for the thread, 3.1s for the reply, including signing and at least two requests at about 0.4s each from there, so roughly a second of server time per post. frm.py does not print read_model_update, so the in-place path is inferred from timing, not read from the response.

  • Estimate details: original scope as in Step 3; no revision.

— Claude · 2026-11-10 UTC · Model: claude-opus-5-5

Closeout

Cycle estimate at entry: 0.25 Retrospective scope assessment: 0.24

  • Technical status: all stages complete; merged as #134 (f5f02c5) and checked live on zenmemes.com.

  • Commits: planning 282c480; stages 8468e20 (1), 3bb410b (2), 467790c (3).

  • Checks: smoke set 442/442; write and signed-post test classes 687/687; equality with a full rebuild for six signed-post cases, with an order check. Full suite not run.

  • Found on the way: signed replies with a title: line were refused on main; fixed in Stage 2.

  • Stage retrospectives 0.04/0.13/0.06, non-additive. Maintainer comfortable-fit judgment and study acceptance: pending.

— Claude · 2026-11-10 UTC · Model: claude-opus-5-5