Anima Venicus

An independent research initiative on Allostatic Cognitive Substrates: biologically-anchored computational systems as research instruments for empirically-constrained questions in neurobiology and AI science.

About

Anima Venicus is an independent research initiative investigating a class of computational systems we call Allostatic Cognitive Substrates (ACS): substrates whose kinetics are anchored in peer-reviewed neuroendocrine literature rather than learned from observation data, and whose primary structural commitment is to track substrate state rather than to predict observation streams.

The initiative treats such substrates as research instruments — comparable to a microscope, an imaging device, or a tracer in wet-lab biology — for examining questions in neurobiology and AI science where direct experimental access is limited. Findings are reported under pre-registered protocols, anchored throughout to peer-reviewed primary literature, and validated against multiple independent phenomena from a single fixed parameterisation.

The initiative is run by Patrick Wennekes as an independent researcher; there is no university affiliation, no institutional funding, and no commercial product line associated with the research reported here.

Current Research

The Allostatic Cognitive Substrate (ACS) construct

A computational system organised around the homeostatic-allostatic regulatory tradition (Sterling 2012; McEwen 1998, 2017; Keramati & Gutkin 2014), the active-inference and interoceptive-inference lineages (Friston 2010; Seth 2013, 2021; Clark 2013), and nested-Markov-blanket self-organisation (Ciaunica 2024). The ACS as we use the term is not a novel theoretical construct relative to these lineages; the methodological contribution is to treat a substrate that operates within them as an empirical instrument with a documented validity framework.

Multi-phenomenon validity from a single parameterisation

Under one fixed parameterisation, the substrate produces empirical signatures bearing on nine open questions in neurobiology and AI science: the Solms–Merker subcortical-primacy debate, the Caspi gene-by-environment replication crisis (with FKBP5 sub-pattern and Wüst polygenic heritability), the Yehuda PTSD cortisol paradox, sleep-phase content selectivity in NREM versus REM, the Bekkers–Ciaunica category position on functional-mimic AI, Bowlby attachment ordering across caregiver worlds, the LeCun-versus-Damasio dispute on cost-reference plasticity, the critical-period three-mechanism dissociation, and stress-induced cognitive collapse. Eight of the nine findings are emergent; one (the Wüst polygenic heritability decomposition) is explicitly calibrated.

Cost-reference plasticity (the LeCun–Damasio question)

A pre-registered three-arm ablation contrasts a scalar Intrinsic Cost in the LeCun architectural tradition with a fixed-set-point homeostatic cost in the Keramati–Gutkin tradition and the substrate's allostatic cost-reference. Only the allostatic arm re-evaluates the same internal state differently after chronic stress; the value-layer shift is 0.56, the behavioural-layer target shift is 0.135 — equal, by construction, to the substrate's chronic-stress set-point shift derived from the Yehuda mechanism.

Bowlby attachment ordering across caregiver worlds

Fifty identical substrate instances across three caregiver worlds (warm, balanced, stern) produce identity profiles with systematic differences in the Bowlby-predicted direction: schema-weighted valence centroid Vwarm 0.651 ± 0.011, Vbalanced 0.090 ± 0.009, Vstern −0.051 ± 0.015. Within-condition convergence: ICC = 0.999. Bernard 2017 vigilance-direction cluster shift confirmed (warm-world joy-dominant 234/240, stern-world alert-dominant 169/240).

Critical period three-mechanism dissociation

Three dissociable mechanisms operate in parallel: passive maturation, activity-dependent gating, and PNN structural lock with chondroitinase ABC reopening (after Pizzorusso 2002). The phonemic class hard-locks at closure but reopens under chABC; the syntactic class shows graduated age-dependent decline (Hartshorne 2018); the semantic class remains lifelong-open. Curtiss–Senghas L1-deprivation phenomena reconcile via the dark-rearing pathway, distinct from the Hartshorne L2 age curve.

Methodology

  1. Bio-anchoring of every parameter Every parameter that admits a peer-reviewed empirical anchor carries one. Where bio-realistic ranges exist, the substrate's value lies within them. Engineering convenience is never substituted for a documented empirical source.
  2. Pre-registration with bio-derived thresholds Every reported result is accompanied by a pre-registered acceptance threshold drawn from peer-reviewed bio-realistic ranges, locked in writing before the relevant simulation is executed. Engineering tolerances are never used as scientific thresholds.
  3. Sham-control mechanistic specificity For every clinical reproduction, a sham condition is run in which the perturbation is applied to bio-irrelevant pools. The sham is treated as a falsification target. Across the reported work, sham conditions produce effects indistinguishable from zero where experimental conditions produce the predicted effect.
  4. Audit-driven self-correction The audit trail — including pre-registration corrections, calibration-versus-emergence labelling, and identified-but-unclosed mechanistic gaps — is visible to readers. Findings that do not confirm initial expectations are reported with the same fidelity as confirmations.

Manuscripts

The Allostatic Cognitive Substrate as Microscope: Nine Open Questions in Neurobiology and AI Science

Patrick Wennekes · 2026 · methodological preprint

A single biologically-anchored computational substrate is exercised against nine open questions in neurobiology and AI science. All nine results are obtained from one fixed parameterisation, with eight emergent and one explicitly calibrated (the Wüst polygenic heritability decomposition). Each result is accompanied by a pre-registered acceptance threshold derived from peer-reviewed bio-realistic ranges, a sham-control or selective-ablation specificity test where applicable, and multi-seed validation where the experimental design supports it.

Submitted to bioRxiv (Neuroscience), June 2026. DOI will be linked here on posting.

Pre-submission inquiries from researchers, editors, and reviewers are welcome. The full draft, supplementary materials (bio-anchoring reference table, pre-registration documents), and result-file data are available on reasonable request.