daily-held: #1639 missing #1310 assistant-test fold

Why this needs you

Filed by the /daily 2026-07-25 problem sweep as a route-3 judgment call under the "Scientific-meaning changes" carve-out: folding a result that dissolves a clean-result's headline effect is an interpretation change on a body sitting at awaiting_promotion.

#1639's clean-result does not contain the #1310 assistant-test result, which landed on main on 2026-07-23 and reportedly collapses the headline assistant-vs-character gap.

What was found

Session 63122023 @ 2026-07-25T18:07:54Z, while answering an unrelated question, discovered that scripts/issue1310_xpersona_assistant_test.py had run on 2026-07-23, was committed (9e65fe09ad), and was never folded into #1639. Its reported result takes the headline 3–5× assistant-vs-character gap down to 0.04 R². The finding was surfaced in chat; nothing was folded and no task was filed.

Verified at filing (2026-07-25)

• task.py view 1639 --json → status: awaiting_promotion; title: "The four fiction-character context→dialogue maps are one dominantly shared linear operator: a pooled…".
• The body's own **Context:** row states its scope verbatim: "consolidating the #1310 cross-persona similarity rounds 1-3: cross-character battery, principled re-analysis, and the assistant test" — so the assistant test is in scope by the body's own account.
• Scanned the body for any assistant-test result: 0 lines mention "assistant" together with an R² value or the word "gap". Its **Repro:** row cites issue1310_xpersona_similarity.py @82d85db5ee (round 1) and issue1310_xpersona_similarity_v2.py @9edaab4fa4 (round 2) — round 3 / the assistant test is absent from the code list.
• The artifact exists on main: git log --oneline -1 9e65fe09ad → "issue #1310 round-3 assistant-test: artifacts + summary + figure (auto-harvest; summary_rc=0 cert_rc=0)"; scripts/issue1310_xpersona_assistant_test.py present.

One premise I could NOT verify — treat as unverified hypothesis: the specific "3–5× gap → 0.04 R²" figure is quoted from the session's chat text, not read from the round-3 summary JSON. Read eval_results/issue_1310/ round-3 outputs before acting on the magnitude. The structural fact — a landed, in-scope round absent from the body — is verified above.

The decisions that are yours

1. Fold or don't. If the result holds, #1639's headline ("one dominantly shared linear operator") may need requalifying before promotion — that is an interpretation call on your text.
2. Promotion timing. #1639 has been sitting at awaiting_promotion with an incomplete evidence base for 2 days; decide whether it parks until the fold or promotes with a stated scope caveat.

Related work already routed tonight

A route-2 workflow-fix task from this sweep proposes the mechanical half — an audit pass flagging any task body whose named pending artifact/commit already exists on main, so a landed-but-unfolded round cannot sit silently. That task does not touch #1639's body.