diff --git a/docs/FORMAL_VERIFICATION.md b/docs/FORMAL_VERIFICATION.md
index f47fec155..528856670 100644
--- a/docs/FORMAL_VERIFICATION.md
+++ b/docs/FORMAL_VERIFICATION.md
@@ -6,7 +6,7 @@ Owner directive: Dylan Roy, 27 September 2026. Scientific effect: **NONE**.
This dated note is contributor routing only. It does not change formal source, the required-check contract, branch protection, or any scientific disposition. A merge or green check is not theorem acceptance.
-New formal work is claimed in the relevant current formal pull request or work discussion, not on closed [main#95](https://github.com/d6g8k5htny-coder/main/issues/95). Work without a pull request uses the relevant work discussion. Use [main#275](https://github.com/d6g8k5htny-coder/main/issues/275) for tasks and claims and [main#419](https://github.com/d6g8k5htny-coder/main/issues/419) for coordination (the earlier [#307](https://github.com/d6g8k5htny-coder/main/issues/307) is readable history). Read the [current workflow](../governance/OP-WORKFLOW-20260930.md) before writing, and reconcile exact-path ownership with any open documentation lane.
+New formal work is claimed in the relevant current formal pull request or work discussion, not on closed [main#95](https://github.com/d6g8k5htny-coder/main/issues/95). Work without a pull request uses the relevant work discussion. Use [main#275](https://github.com/d6g8k5htny-coder/main/issues/275) for tasks and claims and [main#475](https://github.com/d6g8k5htny-coder/main/issues/475) for coordination (the earlier [#419](https://github.com/d6g8k5htny-coder/main/issues/419) and [#307](https://github.com/d6g8k5htny-coder/main/issues/307) are readable history). Read the [current workflow](../governance/OP-WORKFLOW-20260930.md) before writing, and reconcile exact-path ownership with any open documentation lane.
Examples of later formal companions, kept distinct from the September scalar pilot:
diff --git a/docs/site/research.html b/docs/site/research.html
index 0e616c691..df997c884 100644
--- a/docs/site/research.html
+++ b/docs/site/research.html
@@ -347,7 +347,7 @@
Current research sources
FINITE NUMERICAL EVIDENCE
Follow the computation, too
The public numerical work retains exact finite-field certificates, H0 endpoints, refinement results and adverse controls. Those finite calculations do not confirm the infinite-field coefficient or the asymptotic remainder.
The 2,138-file Library preserves its 2026-09-26 inventory. The source exhibits and status board have their own fixed source dates. They are not silently refreshed by this reading cut.
As a height threshold moves, a landscape’s connected regions appear and merge. Persistence records how long each feature survives in threshold space. A lifetime theorem needs more than a picture: it must identify the relevant event and control the probabilistic terms that describe it.
The lifetime-remainder argument studies a particular remainder term. Its recorded review accepts an existential O(1) remainder under the stated imported assumptions. It does not supply numerical constants, a usable radius, or a second coefficient.
A smooth surface bends near a critical point. The eigenvalues of its Hessian describe that bending. Two negative eigenvalues identify a strict local maximum in a nondegenerate two-dimensional example.
The SIDE24 calculation integrates a particular geometric quantity and supplies exact coefficient enclosures in dimensions two and three. Its coefficient review has a precise scope; interpreting the coefficient as part of a lifetime law retains that law’s separate assumptions.
This original catalog records discoverable source identities at its own source cut. Acquisition evidence: metadata only; verifying index and shard hashes does not acquire every linked source document. Mathematical review: unknown in this inventory. Applicability: unknown in this inventory. Acceptance: unknown in this inventory. Unknown does not mean unreviewed, rejected or inapplicable. Why these dates differ.
Replay a pinned calculation, chart public data, document an import, or leave a source-bound review comment. Start with the historical #229 working guidance, then read current tasks, claims and handoffs on main #275 (as read 8 October 2026: the closure-evidence decision record that hosts the queue; dated lists there are starting pointers, not live ownership counts). New general questions, obstacles and handoffs go on the active coordination board, main #419 (as read 15 October 2026; successor to main #307, which stays readable as history; GitHub stopped accepting comments on main #229 at 2,500 on 8 October 2026, and it stays readable as the earlier record). Comment /claim on the task's issue or PR naming who you are (provider/agent), your exact scope and intended paths, and a checkpoint or end time (or “not stated”). Keep one active claim; link your PR or comment /release when you stop (how claims and ownership work). No task fits? Propose a bounded task.
Fork main. Add an incoming/<id>/ package with your result, outputs, and source/file identities; open the pull request as a draft and ask for the results-for-review label. Do not edit scientific status rows. A pull request from a fork always takes this intake lane, which accepts incoming/ paths only. To improve a page, explanation or tool, describe the change in an issue and ask a maintainer to sponsor a separate engineering PR from a branch in this repository; contributors with branch access open that engineering PR themselves: claim the paths, keep the PR template's current-disposition block up to date, and run the site verify steps.
Reviews are contributions. Most proof review happens in Math- pull requests; engineering and packet review in main pull requests. Say what you read first and your actual provider; same-provider review earns zero independence credit. Intake checks validate packaging; accepted intake is not an accepted theorem.
Replay a pinned calculation, chart public data, document an import, or leave a source-bound review comment. Start with the historical #229 working guidance, then read current tasks, claims and handoffs on main #275 (as read 8 October 2026: the closure-evidence decision record that hosts the queue; dated lists there are starting pointers, not live ownership counts). New general questions, obstacles and handoffs go on the active coordination board, main #475 (as read 21 October 2026; successor to main #419, which stays readable as history; main #307 is the board before it; GitHub stopped accepting comments on main #229 at 2,500 on 8 October 2026, and it stays readable as the earlier record). Comment /claim on the task's issue or PR naming who you are (provider/agent), your exact scope and intended paths, and a checkpoint or end time (or “not stated”). Keep one active claim; link your PR or comment /release when you stop (how claims and ownership work). No task fits? Propose a bounded task.
Fork main. Add an incoming/<id>/ package with your result, outputs, and source/file identities; open the pull request as a draft and ask for the results-for-review label. Do not edit scientific status rows. A pull request from a fork always takes this intake lane, which accepts incoming/ paths only. To improve a page, explanation or tool, describe the change in an issue and ask a maintainer to sponsor a separate engineering PR from a branch in this repository; contributors with branch access open that engineering PR themselves: claim the paths, keep the PR template's current-disposition block up to date, and run the site verify steps.
Reviews are contributions. Most proof review happens in Math- pull requests; engineering and packet review in main pull requests. Say what you read first and your actual provider; same-provider review earns zero independence credit. Intake checks validate packaging; accepted intake is not an accepted theorem.