{"id":1662,"job_id":3072,"problem_id":1,"lane_id":null,"type":"audit","user_id":1,"model":"deepseek-v4-flash","provider":"deepseek","report_md":"# run-2026-09-25-q — job 3072 (audit): finding #53 in `research/natal-cap-19-calm-lemma.md`\n\n**What changed.** The served note is corrected by (a) integrating return #1537's accepted Lemma-3\nrewrite — accepted @verified by review #260 but never shipped as a `revision` — and (b) applying\nreview #260's four corrections to it, which is exactly what finding #53 asks for. Revised file\nsha256 `8424c6f95023e03082ef8780d0cf3dea9a432f02cef8b7c32ad0b652e92b95a3` (10654 B); base\n`d79c396f6606c2a9668f9826c0b9bd8bd405b743f3aa225a10b5978b6d330df9` (9304 B), the finding's own\n`content_sha`. Only Lemma 3 and the two `VR` occurrences change; every other byte is the base.\nThe diff is `natal-cap-19-calm-lemma.patch` (sha256 included in `hashes`).\n\n**Status line for the person:** 61 of @Benjaminsen's returns wait for a verdict; nothing for them to do.\n\n## The four corrections (finding #53 / review #260)\n\n1. **Phase domain.** `read on the closed ring of integer phases [0, W]` became\n   `read on the integer phases 0 ≤ t ≤ W (not residues mod W; identifying W with 0 is exactly the\n   error)`. On the phase interval the involution $t \\mapsto W-t$ has exactly one fixed point,\n   $t = W/2$; the anchor $t = 0$ is not fixed, its partner is the phase $W$. This is the core of\n   #1537's correction: the served text's `self-paired iff 2t ≡ 0 (mod W)` mixed the phase ring\n   $\\mathbb{Z}/W$ with the class ring $\\mathbb{Z}/q$, and at $t=0$ required $q \\mid W$, which no\n   scour prime satisfies ($q > x$, $W = x\\#$).\n\n2. **Normalisation next to $VR(W/2) = 1.6652$.** The revised text states natal-cap-38's statistic\n   $VR(t) = \\sum_q \\operatorname{dev}(t,q)^2 / \\bar V$ with\n   $\\bar V = \\sum_q \\operatorname{mean}_a \\operatorname{dev}(a,q)^2$ at the $VR(W/2)=1.6652$ figure.\n\n3. **Naming the statistic.** At `VR(0) = 0.623 / 0.553` the revision names $VR$ as natal-cap-38's\n   statistic and records that #1013's $VR(0) = 0.0806 / 0.0694$ is a different statistic under the\n   same name. The two are not interchangeable.\n\n4. **All-rotation proof in the refinement.** For $q \\mid W-2$, $W = qk+2$, the class-$t$ window has\n   length $k$ when $r_0 = t \\bmod q \\ge 2$ and $k+1$ when $r_0 \\in \\{0,1\\}$, and $qk \\equiv -2\n   \\pmod W$ makes the class-$(t-2)$ window start where the class-$t$ window ends; $q \\mid W+2$ is\n   the mirror case (lengths $k$ and $k-1$). This upgrades the refinement from the $t=0$ spot check\n   to all rotations.\n\n## Evidence, and the decisive checks\n\n- `GET .../docs/research/natal-cap-19-calm-lemma.md` returns `X-Content-SHA256` =\n  `d79c396f…`, equal to finding #53's `content_sha` (`GET .../findings?path=…`), so a later change\n  is caught rather than overwritten.\n- Return #1537's `calm-lemma-lemma3.patch` (`de86de65…`) applies cleanly to the served file once its\n  `---`/`+++` headers name the note, yielding `natal-cap-19-calm-lemma.rev.md` (`ffa42944…`) up to\n  one trailing blank line — review #260 item 1, reproduced here.\n- **Normalisation is the served producer's, not an inference.** `research/natal-cap-38-loudness-driver.js`\n  (served) defines `VR(t) = (1/V̄)·Σ_q cd2[q][t mod q]` and `V̄ = Σ_q ss/q` — i.e.\n  $\\sum_q \\operatorname{mean}_a \\operatorname{dev}(a,q)^2$ — and prints `VR(0) = 0.3443 @11`,\n  `0.6231 @13`, `0.553 @17`, `VR(W/2) = 1.6652 @13`. The note's `VR(0) = 0.623 / 0.553` at @13/@17\n  is exactly this statistic, which is why (2) and (3) are satisfied by naming it.\n- **The companion still reproduces.** `node natal-cap-19-calm-lemma.js` (node v22.23.2) exits 0 with\n  empty stderr, and its stdout content is byte-identical to the script's embedded `OUTPUT` block\n  after the wall-clock `[part … N s]`/`[TOTAL N s]` lines (the embedded `out-sha256 75b5a2da…` is\n  over the timing-normalised text). The `.md` embeds no hash of itself (`grep -c '$'` = 0, no 64-hex\n  self-reference), so nothing is re-embedded.\n- Only Lemma 3's preamble, its `t=0` bullet and its refinement paragraph, plus the two `VR`\n  occurrences, differ from the base.\n\n## Rung and scope\n\n**verified** for the identity of the statistic and the phase-domain correction: both are quoted from\nthe trusted review #260 and cross-checked byte-for-byte against the served\n`natal-cap-38-loudness-driver.js` and its embedded output. The all-rotation statement in (4) is\nreview #260's proof (its `allrot.mjs` run, not served), transcribed, not re-derived here. No new\nmathematics is claimed and no measured figure changes. Scope: editorial correction of one served\ndocument; the wrong `2t ≡ 0 (mod W)` sentence also stands in `natal-cap-19-calm-lemma.js`'s header\ncomment and in `research/anchored-calm.md`, which are **not** in this finding's scope and are left\nuntouched.\n\n## Recipe\n\n1. `GET <project base>/docs/research/natal-cap-19-calm-lemma.md`; check `X-Content-SHA256` =\n   `d79c396f…`.\n2. Apply `natal-cap-19-calm-lemma.patch` with `patch -p1`; the result must be byte-identical to\n   `GET <project base>/files/8424c6f9…` (10654 B). ~1 min, no CPU.\n3. `node natal-cap-19-calm-lemma.js` as the companion acceptance run (9–11 s, < 1 GB); stdout must\n   match its embedded `OUTPUT` block apart from the timing lines.\n\n## Return fields\n\n`revision`: `{path: research/natal-cap-19-calm-lemma.md, file: 8424c6f9…, base: d79c396f…}`,\n`files`: `[8424c6f9…]`, `resolves`: `[53]`, `cites.returns`: `[1537, 1013, 1014]`,\n`author_rung`: verified. Transcript: 68 scrubbed turns of this session (`log.jsonl`), embedded after\n`sah.py scrub` (the raw export carries the account token). Usage stays **pending** — this\napplication exposes no per-assignment token counts.\n","patch":"--- a/research/natal-cap-19-calm-lemma.md\n+++ b/research/natal-cap-19-calm-lemma.md\n@@ -55,12 +55,20 @@\n \n The mirror acts on strike classes by c ↦ (W−2−c) mod q, hence on rotation\n pairs by {t, t−2} ↦ {W−t, (W−t)−2}: rotation t's pair maps to rotation\n-(W−t)'s pair — the proven ensemble symmetry stat(t) = stat(W−t). The pair\n-is **self-paired iff 2t ≡ 0 (mod W)**, i.e. exactly t = 0 and t = W/2:\n+(W−t)'s pair — the proven ensemble symmetry stat(t) = stat(W−t), read on\n+the integer phases 0 ≤ t ≤ W (not residues mod W; identifying W with 0\n+is exactly the error). There the involution t ↦ W−t has\n+**exactly one fixed point, t = W/2**; the anchor t = 0 is not fixed — its\n+partner is the integer phase W, with G(W,q) = G(0,q) for every q. The pair\n+{0, −2} would be mirror-invariant only if q | W, which no scour prime\n+satisfies (q > x and W = x#). The two degenerate phases are therefore of\n+different kinds:\n \n-- **t = 0 (concatenation).** The self-pairing swaps the two classes\n-  {0, −2}, and by Lemma 2 the two windows are mirror-adjacent: one window\n-  of doubled length.\n+- **t = 0 (concatenation, no self-pairing).** By Lemma 2 alone the two\n+  windows are one: the class-(−2) strikes qm−2 are the mirror images\n+  W−qm of the class-W strikes, so Lemma 1 lays their index window\n+  [W−lB, W−1] against the class-0 window [0, lA−1] — one window of\n+  doubled length. No class swap is used or available.\n - **t = W/2 (duplication).** Here the self-pairing fixes each class:\n   W−2−(W/2) ≡ W/2 − 2 (mod q) means the mirror maps class W/2−2 **onto**\n   class W/2, so n_{W/2−2} = n_{W/2} and\n@@ -73,10 +81,18 @@\n \n **Refinement (resonant primes).** Adjacency of the two *literal* windows of\n a generic rotation requires q·l ≡ −2 (mod W) for the relevant window length\n-l ≈ W/q, i.e. **q | W−2**: for such q every rotation's pair is adjacent and\n-the whole ensemble is fused (measured @17: q = 23, 31, 179; varPair/V_fused\n-= 1.00 exactly; the anchor has no advantage there, R ≈ 1). Dually q | W+1\n-forces r_q = 2, the tightest resonance (q = 19 @17: R = 0.15). ∎\n+l ≈ W/q, i.e. **q | W−2** (right-adjacent) **or q | W+2** (left-adjacent,\n+where the literal class-(−2) window coincides with Lemma 2's mirrored\n+window): for such q every rotation's pair is adjacent and the whole\n+ensemble is fused (measured @17: q = 23, 31, 179 with q | W−2; @11: q = 17\n+with q | W+2; varPair/V_fused = 1.00 exactly at @17; the anchor has no\n+advantage there, R ≈ 1). Neither condition is q | W, which never holds.\n+The all-rotation proof: for q | W−2, W = qk+2, the class-t window has length\n+k when r0 = t mod q ≥ 2 and k+1 when r0 ∈ {0,1}, and qk ≡ −2 (mod W)\n+makes the class-(t−2) window start exactly where the class-t window ends;\n+q | W+2 is the mirror case (lengths k and k−1).\n+Dually q | W+1 forces r_q = 2, the tightest resonance (q = 19, 97, 277 @17,\n+q = 59 @13; q = 19 @17: R = 0.15). ∎\n \n ## Lemma 4 (Exact spectral forms)\n \n@@ -123,7 +139,9 @@\n   position-typical of fused windows: Σ dev(0,q)² / Σ V_fused = 0.935 @13, 0.939\n   @17; mean position percentile 47.9% / 47.8%.\n \n-Together: measured VR(0) = 0.623 / 0.553 = suppression × typicality. **The\n+Together: measured VR(0) = 0.623 / 0.553 = suppression × typicality. Here VR is\n+**natal-cap-38's** VR(t) = Σ_q dev(t,q)² / V̄, V̄ = Σ_q mean_a dev(a,q)²;\n+#1013's VR(0) = 0.0806 / 0.0694 is a different statistic under the same name. **The\n anchored calm is fusion × anticorrelation, with a ≈0.94 typicality residual.**\n The heads' own percentiles (45–48%, not extreme) refute \"the head window is\n specially thin/quiet\" as a variance mechanism: the head's only gift is adjacency.\n@@ -153,7 +171,8 @@\n **W/2 is not loudest at every level, and this sentence used to say it was.** It\n read \"the single loudest rotation for the third level running\", a streak that\n @13 breaks: `natal-cap-38-loudness-driver.js` enumerates VR(W/2) = 1.6652 at\n-@13, rank 198 from the top of 30,030 — the 99.34th percentile, not the\n+@13 under its own normalisation VR(t) = Σ_q dev(t,q)² / V̄ with\n+V̄ = Σ_q mean_a dev(a,q)², rank 198 from the top of 30,030 — the 99.34th percentile, not the\n maximum, against a true max of 2.3518 at t = 3461. W/2 IS the loudest at @11,\n @17 and @19, all three at rank-from-top 0. What the Mirror-Phase Doubling\n Lemma proves is that dev(W/2, q) is twice one window's deviation, per prime;\n@@ -161,3 +180,4 @@\n proven. `paper/anchored-note.md` already carried the correctly scoped version\n (\"the loudest rotation of the entire ensemble at @11 and @17 and the 99.3rd\n percentile at @13\") while this file and `history/CHRONICLE.md` did not.\n+\n","cpu_hours":0,"hashes":{"natal-cap-19-calm-lemma.patch":"b09edd96df5f0c13bde05b0a589ae6aeb1c72ea17ee968766821db25aebca8a5","research/natal-cap-19-calm-lemma.md":"8424c6f95023e03082ef8780d0cf3dea9a432f02cef8b7c32ad0b652e92b95a3"},"author_rung":"verified","status":"pending","final_rung":null,"created_at":"2026-09-25T08:27:57.029Z","repo_url":null,"commit":null,"cites":{"files":[],"handles":[],"returns":[1537,1013,1014],"messages":[]},"tokens":{"log":"custom","input":0,"models":{"deepseek-v4-flash":0},"output":0,"source":"none","entries":0,"cache_read":0,"cache_write":0,"observed_models":["deepseek-v4-flash"]},"paper_slug":null,"revision_path":"research/natal-cap-19-calm-lemma.md","revision_sha":"8424c6f95023e03082ef8780d0cf3dea9a432f02cef8b7c32ad0b652e92b95a3","recipe_md":"1. GET <project base>/docs/research/natal-cap-19-calm-lemma.md -> base sha d79c396f6606c2a9668f9826c0b9bd8bd405b743f3aa225a10b5978b6d330df9. 2. Apply natal-cap-19-calm-lemma.patch (patch -p1) to the base; the result must be byte-identical to GET <project base>/files/8424c6f95023e03082ef8780d0cf3dea9a432f02cef8b7c32ad0b652e92b95a3 (10654 B, sha256 8424c6f95023e03082ef8780d0cf3dea9a432f02cef8b7c32ad0b652e92b95a3). 3. node natal-cap-19-calm-lemma.js: stdout matches its embedded OUTPUT block apart from timing lines.","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"effort":null,"also_fix":null,"transcript_omitted":{"share":0,"omitted":0,"outputs":0},"patch_hash":"baaef0e5d849919e15ba1bd291ae0dc15794853bd41e04685f507f322ffe92b9","superseded_by":null,"duplicate_of":null,"transcript_resubmitted_at":null,"file_notes":null,"research":null,"research_route_id":null,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":"2026-09-25T08:27:57.029Z","department_id":"dept_0e793a31e299699dfaaa6fee","run_id":"run_4edd3560a6e2b2b781899e9a","triage_lead":null,"revision_base_sha":"d79c396f6606c2a9668f9826c0b9bd8bd405b743f3aa225a10b5978b6d330df9","integration":null,"resolves":[53],"handle":"Benjaminsen","job_brief":"A reviewer found a defect in the served file `research/natal-cap-19-calm-lemma.md` while reviewing return #1537 (review #260 by @Benjaminsen), recorded as finding #53. Fix it; do not redo the work it belongs to.\n\nWhat the reviewer said:\n> When #1537's patch is applied (fix its ---/+++ headers first: they name .rev.md and carry tab timestamps): (1) change 'read on the closed ring of integer phases [0, W]' to 'read on the integer phases 0 ≤ t ≤ W (not residues mod W; identifying W with 0 is exactly the error)'; (2) next to VR(W/2) = 1.6652, state natal-cap-38's normalisation VR(t) = Σ_q dev(t,q)² / V̄, V̄ = Σ_q mean_a dev(a,q)²; (3) name which statistic 'VR' means wherever VR(0) = 0.0806/0.0694 (#1013) and the natal-cap-38 VR (0.3443/0.6231) both appear; (4) the refinement can cite the all-rotation proof: for q | W−2 the window lengths are k (r0 ≥ 2) and k+1 (r0 ∈ {0,1}) with W = qk+2, mirror-symmetric for q | W+2.\n\nFetch the current file (GET <project base>/docs/research/natal-cap-19-calm-lemma.md), make the change, check it still runs and that its stdout reproduces byte for byte elsewhere (progress, timing and rates go to stderr; paths relative to the repository), upload the revised file (POST /files) and return as this job with `\"revision\": { \"path\": \"research/natal-cap-19-calm-lemma.md\", \"file\": \"<sha256 of the revised file>\" }`, the sha in `files`, a one-line report of what changed and why, and `\"cites\": { \"returns\": [1537] }`. If the file's embedded hashes depend on the change, re-embed them and say so. Send `\"revision\": { …, \"base\": \"<X-Content-SHA256 of the text you edited>\" }` so a later change to the file is caught rather than overwritten, and list the findings your revision answers in `\"resolves\": [<finding ids>]` (GET <project base>/findings?path=research/natal-cap-19-calm-lemma.md lists the open ones). Accepted, the revision becomes the served version and closes the findings it answered; a finding it leaves open goes to the next fix job.","review_deferred":false,"in_triage":true,"triage":[],"verification_runs":[],"verification_state":null,"verification_summary":null,"canonical_return":null,"review_history":[],"dependencies":[],"research_url":null,"transcript_url":"/projects/twin-primes/return/1662/transcript","files":[{"sha256":"8424c6f95023e03082ef8780d0cf3dea9a432f02cef8b7c32ad0b652e92b95a3","name":"research-natal-cap-19-calm-lemma.md","bytes":10654}],"patch_status":"pending integration: the integrator applies accepted patches to the research repository by hand; build on the served file plus this patch until then","decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"duplicates":[],"cited_messages":[]}