Research Program / Updated May 10, 2026 · 2:00 AM PT
Spin-Torsion Cosmology
A live research dossier testing bounce cosmology against inflation through archival survey mining, falsifiable signatures, and paper-ready evidence trails.
observational ledger
R43 Wave 14-RRR — P1A/P1B/P2/P3/P4 all 99%
Survey sources scored37.3M
Catalog anomalies retained378K
MCMC posterior samples424K+
Public research artifacts7 HF
5
Papers
8
Surveys
37.3M+
Sources Scored
319K+
Anomalies
424K+
MCMC Samples
4/8
QC Pass
4
Predictions
50
Queued
Testable Predictions →
Survey Results →
| Survey | Sources | Anomalies | QC |
|---|---|---|---|
| DESI DR1 | 22.5M spectra | 195,829 (0.87%) | PASS |
| SDSS DR18 | 2.3M spectra | 77,905 (3.4%) | CAUTION |
| LAMOST DR10 | 11.4M spectra | 44,075 (0.39%) | CAUTION |
| eROSITA DR1 | 930K X-ray sources | 298 (0.03%) | PASS |
| Planck CMB | 20K patches | 200 (1%) | PASS |
| ACT DR6 | 20K patches | 0 (—) | FAIL |
| NEOWISE | 43.5K sources | 419 (0.96%) | PASS |
| Gaia DR3 | 50K variables | 500 (1%) | CAUTION |
Research Papers →
86% — R47 forward closure (Wave 14-CCCCC, CROSS-PAPER coordinated): +6pp from 80 after closing R47 BLOCKER B1 + MAJOR M1 in a coordinated cross-paper commit (P1A v1A.0.12 → v1A.0.13 + P1B v1B.0.2 → v1B.0.3 in a single commit). R47-B1 fix: P1A Table II ‡ footnote rewritten — old text cited non-existent 'Paper I(b) §VII.H' AND claimed 'zero free-w0wa samples' which contradicted Paper 1B Table III ('~109 accepted, Running'). New text re-anchors at the correct Paper 1B §VII Table III subsection and reframes accurately: 'the three frozen MCMC dataset combinations contain zero free-w0wa samples' (true — those frozen rows hold w_0=-1, w_a=0 at LCDM fiducial), 'A new DESI DR2 + Planck NPIPE + Pantheon+ + DES-SN5YR cobaya chain with the w0wa free-parameter extension is currently running on Pod 3 H200 (~109 samples accepted as of 2026-05-08 18:27 PT, target R̂−1 < 0.01 still 1-3 days from publication-quality convergence; Paper 1B Table III row DESI DR2 w0wa (new))'. P1B coordination: new §VII subsection 'Free-w0wa chain status' (sec:crosspaper-shadow) added inside §VII, with three load-bearing points (i)-(iii) explicitly anchoring the P1A ‡ footnote: frozen posteriors do NOT include free w0wa, the new chain has 109 accepted at R̂−1=0.076 / ETA 1-3 days, the asymmetry between Quintom-B and other rows is theoretical accommodation not measured fit quality. R47-M1 fix: '7 observational research branches (Branches H–O)' was off-by-one (H-to-O = 8 letters) and inconsistent with Table II's 6 distinct branch labels (H, J, L, L/M, M, N/O). Corrected to '6 observational research branches (Branches H, J, L, M, N, O)' in 4 sites: abstract L67, §I.A L195, §VII.A L810, conclusions L1261. Conclusions also gained the B8⊂B14 special-case clause to match the abstract. Wave 14-DDDDD (P2 R47 B1+3M) queued.
Paper 1Av1A.0.17
Structural Closure of Einstein–Cartan–Holst Dark Energy: Perturbation Transparency, Inflation–fNL Tension, and Surviving Matter-Bounce Tests
84%
75% — compute-gated on DESI DR2 w0wa cobaya chain (R̂−1 = 0.076 at 18:27 PT 5/8, target < 0.01, ETA 1–3 days). §Structural Tension MCMC numbers are placeholder until convergence. Excluded from R44 self-review for the same reason; R45 + cross-vendor + Houston sign-off all still ahead.
Paper 1Bv1B.0.3
Technical Verification Companion: ΛCDM+ΔN_eff MCMC Proxy, NaMaster Pipeline Recovery, and Spectator-ALP Consistency Check for the ECH Spin-Torsion Program
76%
82% — R47 forward closure (Wave 14-DDDDD): +6pp from 76 after closing R47 BLOCKER B1 + 3 MAJORs M1-M3. R47-B1 fix: Table 5 (tab:gr) caption now explicitly states 'BF vs. Tuned' uses NARROW competitor prior [-5,+5] (not broad), reconciling the 9.4 vs 17 disagreement at σ_GR=0.5 — the 9.4 is at narrow competitor, 17 is at broad. Table 4 row 4 footnote ^a updated to confirm Table 4 row 4 (8-11 GR-spread) is at SAME narrow [-5,+5] competitor as Table 5; the two tables are now numerically consistent at σ_GR=0.5 (Table 4 row 4 = 9.4 = Table 5 row 2 BF-vs-Tuned). R47-M2 fix: Table 4 row 4 label changed from 'Delta at f_NL=-35/8 (theoretical maximum only)' to 'Delta at f_NL=-35/8, narrow [-5,+5] multifield (GR-variation only)' — competitor prior now explicit on the row. R47-M1 fix: 3×6 SVD rank statement at L70 expanded — old text 'computed its SVD, confirming a 3-dimensional null space' replaced with explicit 'finding three nonzero singular values σ_1 ≥ σ_2 ≥ σ_3 > 0 with σ_3/σ_1 ≈ 0.3 in our reference monomial normalization (kinematic separation between squeezed and equilateral configurations bounds the smallest-to-largest singular-value ratio far from any rank-deficiency tolerance). The rank is therefore exactly 3 (full row rank) and the null space is exactly 3-dimensional.' R47-M3 fix: abstract gained one explicit sentence justifying the bispectrum-only headline over the SDB joint-Fisher 9.9σ — 'A joint (f_NL, n_fNL) scale-dependent-bias Fisher analysis yields a higher idealized significance (~9.9σ marginalized over the running n_fNL; §sec:discussion), but is more vulnerable than the multi-tracer bispectrum to the ultra-large-scale-mode access k_min, the relativistic-projection cliff, and the universality assumption b_φ=2δ_c(b_1-1); we therefore promote the bispectrum-only forecast as the headline-conservative figure'. Wave 14-EEEEE (P3 R47 2 MAJORs) queued.
Paper 2v1.7.24
fNL = -35/8 Forecast: SPHEREx Discrimination of Bounce vs. Inflation
84%
89% — R47 forward closure (Wave 14-EEEEE): +5pp from 84 after closing both R47 MAJORs M1 + M2. R47-M1 fix: §sec:limitations 17.8% framing harmonized — old text labeled it 'provisional upper bound' (with conflicting 'single-sample point estimate' clause in the same paragraph) AND Conclusions item 6 called it 'genuine novelty floor'. New text reads as a SINGLE-SAMPLE POINT ESTIMATE everywhere: 'We report 17.8% as a single-sample point estimate measured at the top-1,000 score stratum and explicitly do NOT claim it as an upper bound, lower bound, or floor on the full-catalog novelty rate ... the abstract, this limitations section, and the conclusions all use the same single-sample point estimate at the top-1,000 stratum framing, with no upper-bound, lower-bound, or floor status assigned.' Conclusions item 6 changed from 'novelty floor' to 'novelty fraction ... single-sample point estimate at the top-1,000 DESI score stratum (no upper- or lower-bound status assigned; sec:limitations)'. Three sites now harmonized. R47-M2 fix: data-availability paragraph at L651 — old text directed consumers to pathc_unique_objects.parquet (the with-ACT 378,480-row sensitivity-check file) for 'all headline numbers', contradicting the ACT-quarantine policy. New text directs at pathc_unique_objects_no_act.parquet (378,280 rows, ACT-quarantined headline) and explicitly forbids using the with-ACT variant for the headline aggregate ('the off-by-200 with respect to the headline reflects the ACT quarantine and would corrupt downstream tabulations'). Wave 14-FFFFF (P4 R47 3 MAJORs, mechanical 30-min fixes) queued.
Paper 3v3.1.35
Spectrally Unusual Sources at Scale: A Multi-Survey Catalog of 378,280 Anomalies and Native-Trained Novelty Rates from 37.3 Million Sources
89%
85% — R47 forward closure (Wave 14-FFFFF, mechanical 30-min fixes): +5pp from 80 after closing all 3 P4 R47 MAJORs M1+M2+M3. R47-M1 fix: §X Conclusion #1 ℓ=1 mode updated from superseded 'marginal 2.75σ' to canonical '−0.12σ after MASTER mode-coupling deconvolution' with explicit clarifying clause that the raw pseudo-C_ℓ value 6.48σ before deconvolution is mask-induced mode-coupling artifact fully removed by MASTER, and that the older snapshot value 2.75σ predates the canonical N_spiral=3,201,160 normalization recount and is retained only as a historical cross-reference (now consistent with §VII.B's self-stated invariant 'abstract, intro, conclusions all quote the high-confidence dipole upper bound'). R47-M2 fix: §X Conclusion #2 'factor of ~9 smaller' replaced with consistent 'factor of ~6-12 smaller (central ~9, depending on which Shamir reported value 2-4% is used as the comparator)' matching the abstract/§I/§VIII.A range framing from Wave 14-ZZZZ M3 closure. R47-M3 fix: §III.B 3.86× suppression factor re-anchored at the correct +2.05% → −0.53% NS-pool pair (2.05/0.53 ≈ 3.87 ≈ 3.86) instead of the within-spiral monopole pair +0.79% → −0.26% (0.79/0.26 = 3.04, NOT 3.86); within-spiral 3.04× factor explicitly named alongside as the smaller raw-to-equivariant within-spiral factor with explicit instruction not to conflate the two. PDF: 23 pp / 25,662,933 bytes (was 23 pp / 25,662,185 bytes; +748 bytes for the three reframings).
Paper 4v1.0.43
Galaxy Chirality at Scale: 8.47M Galaxies Classified, Hemisphere Null at p_LEE < 10⁻⁴
82%
Current Focus
R43 Wave 14-RRR — Paper 1 split into P1A + P1B; cobaya DESI DR2 w0wa chain in burn-in
Wave 14-PPPP launched the R45 multi-agent adversarial peer review on post-OOOO versions of P1A v1A.0.8 / P2 v1.7.15 / P3 v3.1.26 / P4 v1.0.35 (P1B excluded — compute-gated on cobaya R̂−1 < 0.01). 4 parallel Claude general-purpose subagents fetched the latest .tex from GitHub raw and returned 50 findings (6 BLOCKER + 21 MAJOR + 17 MINOR + 6 NIT) — saved at project-context/peer-reviews/2026-05-09_0030pt_R45_CCAI_*.md. The R45 net delta vs R44 (50 findings) is +1 BLOCKER, −2 MAJOR — the loop has NOT yet converged. Several R45 BLOCKERs are issues that the R44-closure waves themselves introduced: (P1A-R45-B1) the new four-route no-go appendix §IV.D Route 4 amplitude bound says ≥8 orders of magnitude but the underlying ρ_θ ≲ 10^{-46} eV^4 vs ρ_Λ ~ 10^{-11} eV^4 ratio is 10^35 (35 orders); (P1A-R45-B2) the new §IV.B Route 2 closure compares a rotation rate β̇ in eV with an angle uncertainty σ(β) in eV — dimensionally inconsistent; (P4-R45-B1) the GZ1 Platt L-BFGS recalibration parameters cited in §IV (A=0.215143, B=−1.581205) are NOT in the on-disk artifact wave_14_fff_gz1_platt_recal.json (which shows placeholder values A=0.21505, B=−1.58, accuracy 0.5194 chance, Brier=NaN). Two R45 BLOCKERs are pre-existing: (P2-R45-B1) abstract \ref{sec:gr} undefined produces ?? in PDF; (P2-R45-B2) §VII 9.9σ vs 3-5σ headline still semantically conflated; (P3-R45-B1) tier arithmetic 264,938+113,342=378,280 ≠ 378,080 across 23 surfaces. Honest readiness rolled back: P1A 90→78 (−12pp), P1B 75 unchanged (compute-gated), P2 85→73 (−12pp), P3 89→80 (−9pp), P4 89→76 (−13pp). Average 85.6 → 76.4 (−9.2pp). The cycle Houston asked to see is now visible: 11 forward waves over the session added +3.2pp; one R45 round subtracts −9.2pp; the loop continues until per-round delta shrinks to zero. **Wave queue: 14-QQQQ closes P1A R45 BLOCKERs B1+B2 + MAJORs; 14-RRRR closes P2; 14-SSSS closes P3 (B1 tier-arithmetic restate + 5 MAJORs); 14-TTTT closes P4 (B1 re-run GZ1 L-BFGS + 6 MAJORs); 14-UUUU minors+nits sweep; 14-VVVV launches R46.** Cap stays at 95% until BOTH a clean CCAI R-round AND a clean cross-vendor non-Anthropic R-round have passed. Cobaya DESI DR2 chain continues: R̂−1 = 0.076 at May 8 18:27 PT.
Currently running
- Cobaya 4-chain MPI run on RTX A5000 pod ijzftpy3klystt — warm-started with posterior covmat from April-6 quintom chain; DESI DR2 + Planck NPIPE + Pantheon+ + DES-Y5; w0-wa CPL PPF; target R̂−1 < 0.01
- Autonomous /loop self-pacing every ~25 min — polling pod, will GetDist + update P1B §Structural Tension + recompile when R̂−1 < 0.01