The anchor merges
Sources cited in this piece · lab PR #218 (merged 8 November 2026, 17:06 UTC, merge commit 3c36e95) · our run-record comment (7 November 2026) · sealed record of our first run at iqa.org/conformance/icam-runner/ · our merged-mainline run filed as lab PR #225 · draft-li-rttp-iqa-addressing-02 (pinned digest 02491c5a) · runner repository github.com/Aicent-Stack/conformance.
Last week we wrote about a candidate that arrived at the Agent-Authority-Conformance lab of the Linux Foundation Decentralized Trust: an ICAM temporal-derivation vector family built by Nika Novak (GitHub @danyka-icam) from our Internet-Draft, draft-li-rttp-iqa-addressing-02 — the addressing document of the two-pillar work, where RTTP（rttp://）resolves where and when and IQA（iqa://）states whether. Today the follow-up fits in one sentence: the lab merged it. On 8 November 2026 at 17:06 UTC, PR #218 became part of the lab's mainline at merge commit 3c36e95.
Between review and merge, the candidate author did the quiet thing that makes review cultures work: she applied the four requested precision fixes in a docs-only re-submission. Four Section 6.1 equality cases were relabeled as convention-specific rather than normative; the two version identifiers were documented side by side; the index row was enriched with the draft's pinned digest and the two-runs attribution; two known limitations were recorded. Crucially, the 24 test-case inputs and their expected results were not touched by a single byte — the merge changed labels and metadata, not the vectors.
Then we did our part. A record made against a pull request head is a record about a moving object; once the merge landed, we recorded a complete independent run against the merged mainline itself. Our runner — standard library only, digest-gated, refusing to score any input whose SHA-256 is not pinned — replayed all 24 cases against the bytes now sitting in the lab's main branch: 24/24 agree, 6/6 of our own mutation probes caught, the same results our head-run produced. We then did the sequence-level comparison that matters: the only semantic difference between the pre-merge corpus and the merged mainline is four case-status labels moving from normative to convention_specific — exactly the review's request, nothing else moved.
The merged-mainline record is now filed at the lab as an interop record, PR #225 — open, all six CI checks green, awaiting the maintainer's review at her pace, as records wait. It is not a merge vote and not an endorsement; it is a sealed observation about what the mainline now contains, filed where the mainline lives. Both of our runs carry the citation block we ship with every run report: the runner (icam_runner.py), its repository (github.com/Aicent-Stack/conformance), and its operator (Lee / Aicent Stack, RTTP CRI 27 · IQA CRI 28).
What the merge changes, honestly: a week ago the draft's first outside anchor was a pull request that could still be withdrawn; today it is part of a lab mainline that vendors the vectors into its own tree, where they get replayed by the lab's CI on every run. The anchor did not just hold — it got bolted in. We claim no credit for the candidate's quality; that belongs to its author. What we can say is what the merge demonstrates about the draft: someone who owes us nothing derived twenty-four test cases from our text, a maintainer checked them line by line against it, the author fixed four labels, and the whole thing is now infrastructure that runs without either of us in the loop.
That is the update, and it is the good kind. The first seal says it happened. The second says it holds.