Research preprint v1.5.0

Research preprint v1.5.0

This revision adds a separate 256-state mathematical model with two ordered one-bit individuals, supplied private caches and ongoing symmetric policy switching at 1/16. The primary recurrence effect on stationary historical-policy frequency is +0.458222476 percentage points. Active-minus-neutral population accuracy is adverse under ZERO (−0.259357994 points) and positive under HALF (+0.702715312 points). These small effects have exact residual-based numerical certificates. They are not statistical confidence intervals, independent empirical replication, stationary evidence for the published 32-individual model, or evidence of spontaneous memory origin.

The 15,284 statistical estimates through study 060 remain separately registered from the 14 certified mathematical quantities of 061. Includes 18 figures, four saved matrices and vectors, source, certificates and a saved-certificate checker that does not perform a stationary solve or repeat the raw audit. The unresolved 059/060 primaries, frequent founder loss and all earlier adverse outcomes remain. The four-work guide/catalog is retained and updated; a version is not a new paper.

Manuscript/data: CC BY 4.0; original code: MIT. Not externally peer reviewed; AI assistance disclosed. No new scientific execution was run for publication.

Version DOI: https://doi.org/10.5281/zenodo.22950710
Previous v1.4.0 and all earlier versions remain unchanged: https://doi.org/10.5281/zenodo.22949032

Research preprint v1.4.0

Research preprint v1.4.0

This revision integrates the fixed observation-error challenge (060) into the existing private-memory paper. The primary noise-by-recurrence founder interaction remains unresolved: +1.635742 [-2.604370, +5.655924] percentage points. Exact and noisy founder recurrence effects remain positive above one expected descendant. Noise reduces absolute mean true population accuracy in both arms under both future laws; terminal effects remain unresolved. Founder transmission and population utility remain separate outcomes.

The release retains the unresolved pulse primary, frequent founder loss, all adverse outcomes, one prespecified noise level and the reuse of exact-observation 057 preparations. It does not establish noise equivalence or independent training replication.

Includes 17 figures and 15,284 saved-table estimates through 060, compact evidence and scientific source, 207 registered manuscript statistics, and preserved audit/portability records. Manuscript/data: CC BY 4.0; original code: MIT. Not externally peer reviewed; AI assistance disclosed. Large raw trajectories and random tapes are excluded. No new scientific experiment was run for this publication.

Version DOI: https://doi.org/10.5281/zenodo.22949032
Previous v1.3.0 DOI and all prior tags remain unchanged: https://doi.org/10.5281/zenodo.22939403

Research preprint v1.3.0

Research preprint v1.3.0

Adds the fixed one-update fresh-expression pulse study (059) to the existing private-memory paper. Its primary terminal population-accuracy contrast is unresolved: +0.037 [-0.247, +0.336] percentage points relative to unchanged historical retrieval. The interval also overlaps the separate one-expected-matching-bit benchmark. This is not evidence of equivalence or absence of effect.

Restoring historical retrieval after the common first update improves terminal founder representation relative to continued fresh expression: +2.995 [1.465, 4.688] points. This is a secondary ancestry outcome, not demonstrated population benefit or superiority over unchanged historical retrieval. Negative first-update ancestry, frequent founder loss, two continuous-F survivors, unresolved comparisons and exact first-update coupling identities remain in the record. New transfer inputs reuse all 192 selected 057 preparations; no independent training cohort is added.

Includes revised manuscript/supplement, 16 figures, all 15,203 saved-table estimates, scientific source through 059, seed provenance, compact time series/fates/adverse outcomes and scoped audits. Raw candidate trajectories and random tapes remain outside the compact release. No new simulations are needed for its saved-table and figure checks.

This is a revision of the same paper. Earlier tags and DOI versions are preserved. Manuscript/data: CC BY 4.0; original code: MIT. AI assistance is disclosed; no external peer review is claimed.

Research preprint v1.2.0

Research preprint v1.2.0

Adds a prospectively fixed paired Hamming/four-bit-trap objective challenge (058) to the existing private-memory paper. The trap future-recurrence contribution is +23.74 [17.26, 30.23] percentage points and exceeds the separately specified one-expected-descendant benchmark. The direct trap-minus-Hamming comparison is +5.92 [-1.06, 13.13] and unresolved. Most introductions disappear; founder transmission and objective-specific population utility remain distinct.

Includes the revised manuscript/supplement, 15 figures, all 15,167 saved-table estimates, scientific source through 058, seed provenance, compact fates/adverse outcomes and scoped audit records. Figures and arithmetic can be rebuilt without new population trajectories. Raw candidate trajectories and random tapes remain outside this compact release. No claim of a general ruggedness advantage, uniquely identified epistasis mechanism, biological result or independent software implementation is made.

This is a revision of the same paper, not a new paper. Earlier tags and DOI versions remain preserved. Manuscript/data: CC BY 4.0; original code: MIT. AI assistance is disclosed; no external peer review is claimed.

Research preprint v1.1.0

Research preprint v1.1.0

Adds the crossed preparation/future-recurrence comparison (056) and its prospectively specified independent cohort (057) to the existing private-memory manuscript. This is a revision of the same paper, not a second paper or independent software implementation.

The fixed future-recurrence direction reproduces: +22.76 [17.12, 28.95] percentage points in the reused cohort and +16.20 [11.23, 21.02] in the new cohort, using separate pointwise 95% intervals. No cohort pooling or parameter search is added. Unresolved preparation interactions, sampled extinctions and the negative-founder/positive-population-accuracy combination are retained.

The package includes the revised manuscript and supplement, 13 figures, all 15,053 saved-table estimates, scientific source through 057, seed provenance, full compact fate/negative-outcome tables and scoped audit receipts. The reproducibility scripts generate no population paths. Raw candidate trajectories and random tapes remain outside the compact release.

Version 1.0.0 and its DOI remain preserved. Manuscript/data: CC BY 4.0. Original code: MIT. AI assistance is disclosed; this preprint has not undergone external peer review.

Research preprint v1.0.0

Research preprint v1.0.0

This first public release integrates accepted studies 045–055 into one focused manuscript about environmental recurrence, private-memory use and strategy transmission in an abstract finite-budget population model.

The release includes a 19-page manuscript, supporting information, eleven figures, archived scientific source, and complete saved block-level outcome grids. The statistics script reproduces 14,729 means and approximate pointwise intervals with no new population paths or random draws. Reused cohorts, negative and unresolved contrasts, frequent introduction loss, technical corrections and the distinction between transmission and population performance are retained.

The final study finds a partial-recurrence matched F-to-H founder-frequency effect of +21.18 percentage points (pointwise 95% interval 16.36–26.35), and an H-to-F effect of −2.79 points (−4.75 to −1.03). Both terminal population-accuracy effects remain unresolved. These are conditional computational findings, not measured biological effects or an equilibrium-invasion result.

This is a compact evidence release, not a full raw-trajectory archive. Manuscript, figures and original data are CC BY 4.0; original code is MIT. The work has not undergone external peer review.

Archived publication: DOI 10.5281/zenodo.22907237. Zenodo contains the same manuscript, supplement and compact evidence ZIP as these release downloads.