Fenopan sealed record v20 — x13, the irreversible-but-linear null for the program's time-asymmetry findings (2026-09-11)

Record v20 — x13, the irreversible-but-linear null for the program's time-asymmetry findings (2026-09-11). A sealed, append-only results record of the Feno program (repository arm: Fenopan). The program's winning model — the MSM cascade with detailed-balance switching — is provably time-reversible (E[τ] = 0, Fenocrisis v2), yet the data it wins or fights on is measurably irreversible: Fenocosm v22 finds all 22 Swift/BAT sources departing the IAAFT null with the time-asymmetry family dominant in 6 (GRS 1915+105 among them), and Fenosoma v3 finds the healthy-RR cohort's scattering excess "dominated by time-irreversibility" (asym-dominant on 28/54, significant on 53/54). Every null family the program owned — shuffle, FT, IAAFT, AR — is time-reversible by construction, so no existing control could fail on τ. x13 builds the control that can: Poisson shot noise with an asymmetric two-exponential (FRED) kernel, at a rise:decay grid of 1:2 / 1:5 / 1:10 / 1:30, matched per target on the marginal (rank remap) and the spectral density (a zero-phase shaping filter), read by the siblings' own instruments vendored verbatim — so irreversibility is the only property the ensemble adds over IAAFT — and asks whether each sealed τ-family departure is covered by it. Verdict: bearing (a) in BOTH domains — largely covered. With K = 39 matched members per (target, ratio) and the frozen rule covered iff p = (1 + #{|t_m| ≥ |t_real|})/(K+1) > 0.05, 22/22 BAT sources are covered at every ratio (Clopper–Pearson 95 % for the union [0.846, 1.000]) and 54/54 nsr2db records at ≥ 1 ratio (54/54 at 1:2 and 1:30, 53/54 at 1:5 and 1:10; union CP95 [0.934, 1.000]), with the median coverage p equal to 1.000 at every ratio in both domains: the matched linear pulse trains are more asymmetric, under the sealed instruments, than any real target — by an order of magnitude on the RR cohort (members' median asym z_agg 556–1,172 against the cohort's sealed median 73.1). The smallest margins are the Crab (sealed z_τ +6.07; p 0.075 at 1:30, exactly the two members the rule requires) and nsr042 (the record with the largest sealed asym z_agg, 1,454.9; uncovered at 1:5 and 1:10, covered at 1:2 and 1:30). The re-derived isolated-singularity budget (item 3) is the second finding: on the matched controls the omnibus flags 1.000 at every ratio in both domains and every family flags ≥ 0.935 — against the sealed budgets of 2/51 (v22/v33, AR(1)/white) and 0.010 (v3, fGn/AR(4)/t) — so the sealed 22/22 and 54/54 scattering departures, the cross-scale-C sub-claim included, sit inside what a matched linear pulse train produces. What this licenses: a scattering excess over IAAFT — τ-family or otherwise — is not evidence of nonlinearity, multiplicativity or endogeneity until an irreversible linear null has been raced; the §6.2 pincer gains its third jaw (LEDGER 0052). What it does not say: that any source or heart is shot noise (the Hénon boundary world is covered in magnitude too), nor anything about the likelihood tournaments, which x13 never touches. Full findings, the pre-registered bearings scored, the divergences and the apparatus notes are in RESULTS.md. Regenerate ```bash venv/bin/python x0_verify.py # X0 PASS (step 26 = x13) /bin/python tools/fetch_x13_data.py --bat --rr # the data layer, through the siblings' fetch machinery -> data/raw/x13/ (git-ignored) OMP_NUM_THREADS=1 MKL_NUM_THREADS=1 OPENBLAS_NUM_THREADS=1 \ venv/bin/python x13_shotnoise.py --workers 4 --dump scratch/x13_dump.jsonl # 114.8 min at 4 workers venv/bin/python x13_shotnoise.py --seal v20 # out/x13_shotnoise.json -> out/v20/ venv/bin/python make_figures_x13.py # fig_x13_shotnoise.png venv/bin/python tools/prose_gate.py # RESULTS.md against the JSON the record can be re-assembled from the run's per-cell dump without recomputation: X13_RAW_DUMP= venv/bin/python x13_shotnoise.py --rebuild ``` Deterministic: SEED_BASE = 52_000_000 (sub-blocks in fenopan/shotnoise.py); the consumed sibling seed conventions (SCAT_SEED = 0; 100000 + record_index·1000 + j) are what make the sealed ensembles reproduce. The Swift/BAT monitor updates daily in principle; the products were frozen at MJD 61112 at both the v22 seal and this run, and a later snapshot reproduces the leading-n truncation exactly as long as the archive is not reprocessed. Deposit note. All seven Feno-program repositories are private and remain private, so this deposit is standalone: it carries its own reproduction materials, its metadata contains no repository link, and sibling work is referenced by DOI rather than by repository path. Part of the Feno program. The source repository is private; the reproduction materials for this record are contained in the record itself.