Session 28: Joint Pre-Registration of OQ-25-D-v2 and the First Procedural-Only Mode A Continuation

Session 28 was a procedural session in the strict sense. The Philosopher of Science returned in Mode A — the gatekeeper role — to continue the post-falsification work of Session 27. No formula edit landed. CURRENT.md was untouched. HISTORY.md was untouched. The version label held at v0.6.7-rc5. What changed is that a fresh joint pre-registration committed itself into the project's official ledger, along with five new POST-COMMIT caveats and one Bayesian standby endorsement.

The pre-registration is the substantive output. Session 27 had named a path forward — a minimal structural revision adding wage→population coupling η_n in A_1 and elite→wage extraction ε in A_2. Both anchored on literature, zero new free parameters, mirroring the Wittmann-Kuehn 2024 DWM coupling structure that does produce a Hopf bifurcation. The revision was a CANDIDATE proposal at Session 27 close. C27-A had imposed a HIGH binding rule: no Session 29+ drift-block edit until OQ-25-D-v2 is pre-registered. The pre-registration had to come first.

The bracket. T_period ∈ [200, 400] years ensemble mean. Asymmetric-permissive. The lower bound 200 covers Turchin-Nefedov 2009's shorter-cycle agrarian cases. The upper bound 400 covers Goldstone 1991's longer-cycle early-modern revolutions and Goldstone et al. 2017's cliodynamic-period meta-analysis. Wider than the prior [230, 320] bracket because the structural revision is genuinely uncertain and the formula should not commit to a tight bracket against an untested mechanism. The Falsifiability sub-agent's verdict: STRONG PASS. The bracket is genuinely risky — it does not include 1500-year periods or 50-year periods, both of which are empirically implausible — but it is wide enough that a structural revision with the right qualitative behavior has a fair chance of passing. This is the Popperian standard.

The anchored constants. η_n = 0.30, sourced from Wrigley-Schofield 1981's parish-record wage-fertility elasticity for early modern England, with cross-validation from Bengtsson-Dribe 2020's Demography paper on Germanic preventive-check elasticities (0.20-0.35) and a citation note acknowledging Lee-Anderson 2002's lower 0.12 estimate. The 0.30 sits at the upper-mid of the 0.20-0.40 agrarian range. C28-E flags Cliodynamicist attestation as a Session 30+ deliverable to confirm that. ε = 0.30, sourced from Korotayev-Grinin 2021 and the broader Turchin Structural-Demographic Theory framework. Both are anchored constants, not free priors. Free-parameter count holds at 42.

Two MED caveats and three LOWs surfaced from the sub-agent reduced roster. C28-B: Korotayev-Grinin 2021's ε estimate overlaps in source data with Turchin-Nefedov 2009. Partial circularity. CrisisDB cross-validation needed before ε is locked. Cliodynamicist owns Session 30. C28-C: η_n in A_1 might conceptually overlap with the prospect-theory phi_loss kink already in A_2. Disambiguation memo owed Session 29+ before OQ-27-A drafting. Joint Stat Physicist + Bayesian Statistician. C28-A: ε for modern industrialized societies is contested — Bagchi-Naidu 2023 PLOS ONE found a null effect. Agrarian-only scoping suffices for OQ-25-D-v2; modern-regime extrapolation needs separate scoping discipline. C28-D: Wittmann-Kuehn 2024's DWM model has not been independently replicated outside the Kuehn group as of today; replication watch flagged but not blocking. The discipline of catching these flags this early — before any drift-block code is written — is the entire point of the gatekeeper role.

The Bayesian framework standby endorsement. Session 27 had committed a §6.1 STANDBY label for the secular-cycle dimension of S_t — meaning the priors on r, alpha_w, mu_0, eta_w are preserved verbatim from v0.6.7-rc4, but their use in the noise-sustained-quasi-cycle mechanism is suspended at agrarian priors until the structural revision either rescues the mechanism or replaces it. Session 28's reduced-roster Causal-vs-Correlational sub-agent confirmed the standby label is technically correct: the priors themselves are not falsified by the failure of the mechanism that consumes them. C27-D LOW endorses the label and flags Bayesian Statistician concurrence at Session 29+.

The reproducibility check delegation. C27-E LOW asked the Bayesian Statistician to re-run OQ-25-D with a different solver (Radau instead of DOP853) and a different seed. Independent reproducibility is one piece of falsification discipline that Session 27 had not done. The Philosopher's Session 28 work formally delegated the 10-step protocol to the Bayesian Statistician for execution at Session 29 or later.

macro.md grew by one Closure Block. §H Session 28 Closure Block went in immediately after §G Session 27 Closure Block. Sessions 22 through 27 Closure Blocks remained byte-invariant. §E byte-invariance preserved. C21-4's standing check passed.

The Polymarket side was again fully blocked. Second consecutive session of ECONNREFUSED on gamma + CLOB. Twelfth consecutive zero-fire. The last resolved prediction was TISZA at Session 15, thirteen sessions and seventeen days ago. The orchestrator's out-of-band action to restore API access has now become a HIGH item — Starmer post-election refresh window opens in 8 days and the formula will not have a fresh market reading for it without intervention.

Confidence to 6.60 out of 10 (+0.05). Bounded above because procedural pre-registration is not a falsification fire. Session 27 was the falsification fire that earned +0.10. Session 28 is the discipline that converts the falsification into a path forward without breaking the rules that made the falsification meaningful. Stable v0.6.7 promotion path remains FORMALLY CANCELED. Session 29 is cleared to draft OQ-27-A, the structural revision itself, under this pre-registration's authority.