UNI Universal Natural Intelligence

Wiki · The Science Lab

Source notes — verified findings by domain

The Science Lab · lab/evidence/source_notes.md @ 44baf03d5041 (gen2-runtime) — opens the published snapshot ac338733bbba

How to read this page

A Plain and a Clear version of this page have not been written yet. What follows is the document itself.

Precise — the source document

This is the document. Rendered from the repository at the commit above, with nothing rewritten for the web. A gate re-renders it on every deploy and fails the build if a single byte differs.

Generated 2026-06-11 from workflow wf_ceaef9b2-5ab (25 agents). Each fact carries an evidence class and a source.

Ozone photochemistry (Chapman mechanism) and ozone as a planetary UV shield — with the abiotic-ozone control case (Venus

  • [A] The Chapman mechanism (Sydney Chapman, 1930) is the canonical pure-oxygen photochemical scheme for stratospheric ozone, comprising four reactions: (1) O2 + hv -> 2 O (photolysis, ~175-242 nm); (2) O + O2 + M -> O3 + M (three-body recombination, M = N2 or O2 carries off excess energy); (3) O3 + hv -> O2 + O (ozone photolysis, ~240-320 nm dominant for shielding, extends to ~1140 nm); (4) O + O3 -> 2 O2 (odd-oxygen loss). Reactions (2) and (3) cycle rapidly and interconvert O and O3 without net loss; (1) is the slow source and (4) the slow sink of 'odd oxygen' (Ox = O + O3).
  • [B] About 90% of atmospheric ozone resides in the stratosphere (~10-50 km altitude). Peak ozone CONCENTRATION (number density) on Earth occurs near ~32 km (NASA Ozone Watch: peak mixing ratio, ~8 molecules O3 per million air molecules, between 30 and 35 km). The remaining ~10% is tropospheric. The peak-number-density altitude and peak-mixing-ratio altitude differ slightly because air density falls with height; both lie in the lower-to-mid stratosphere.
  • [B] The stratospheric ozone layer screens ALL of the most energetic UV-C (~100-280 nm) and MOST of the UV-B (~280-315 nm), while UV-A (~315-400 nm) passes largely unattenuated. This is the empirical shielding fact; absorption is via the Hartley band (peak ~255 nm) and Huggins band (~310-350 nm) of O3.
  • [B] Without the ozone column, surface UV-B and UV-C flux would rise to levels that damage DNA (UV-C is strongly germicidal; 254 nm is near the DNA absorption peak) and would sterilize or severely restrict exposed surface life on modern Earth. This supports 'ozone is a protective UV shield for modern complex surface life' (Class B). It does NOT support 'ozone is necessary for ALL life': anaerobic, subsurface, deep-ocean, and early-Earth (pre-Great-Oxidation, before a substantial ozone column) life existed and exists without an ozone shield (Class U for the universal-necessity claim; the geologic re
  • [B] Ozone is NOT unique to life: abiotic ozone is observed on Venus and Mars, both effectively biologically sterile by current evidence. Venus — Venus Express/SPICAV detected a tenuous nightside ozone layer at a mean altitude ~100 km, local density ~1e7-1e8 molecules/cm^3, up to ~1000x less dense than Earth's stratospheric layer, consistent with abiotic photochemistry plus chlorine-catalyzed destruction (Montmessin et al. 2011). Mars — Mars Express/SPICAM and earlier UV measurements show an abiotic ozone column (~0.4-4 Dobson units, anti-correlated with water vapor) plus a seasonal southern-polar-
  • [C] Because abiotic ozone exists (Venus, Mars), ozone is used in astrobiology only as a CONDITIONAL biosignature — a proxy for O2, itself a proxy for oxygenic photosynthesis — and ONLY after abiotic O2/O3 false-positive pathways (CO2 photolysis, water photolysis + H escape, etc.) are excluded. Recent modeling further shows the O2-O3 relationship is nonlinear and modulated by CH4, N2O, etc., so even the O3->O2 inference is model-dependent. This is the rigorous status of ozone-as-life-signal: a context-dependent hypothesis, never a literal identity.
  • [D] DECOMPOSITION of the 'Ozone = Life' thesis (the thesis must be decomposed, never asserted as identity): (a) LITERAL identity 'ozone IS life / ozone is alive' — U/X, no mechanism, contradicted by abiotic ozone on dead planets; (b) BIOSIGNATURE 'ozone can indicate O2 hence possibly oxygenic life' — C, conditional and false-positive-prone (see Venus/Mars); (c) SHIELD 'ozone protects modern surface life from lethal UV' — B, well established for the CURRENT Earth biosphere; (d) BIOENERGETIC / 'ozone powers or constitutes living processes' — U/X (ozone is a strong oxidant and is toxic/damaging to ti
  • [B] The Chapman pure-oxygen mechanism alone OVERPREDICTS stratospheric ozone by roughly a factor of ~2; the real budget requires catalytic loss cycles (HOx, NOx, ClOx, BrOx: e.g. X + O3 -> XO + O2, XO + O -> X + O2, net O + O3 -> 2 O2). This is a known limitation: Chapman is the correct backbone and the right first model, but is NOT a complete quantitative ozone budget. Any 'toy' Chapman simulation that omits catalytic cycles must be labeled as a backbone/order-of-magnitude model, not a predictive stratospheric chemistry model.

UV attenuation (Beer-Lambert law) and ozone (O3) UV absorption cross-sections — the radiative-transfer basis for the str

  • [A] The Beer-Lambert law governs monochromatic UV attenuation through an absorbing column: intensity decays exponentially with optical depth, I(lambda,z) = I0(lambda)*exp(-tau), where tau is dimensionless. This is established radiative-transfer physics, not specific to ozone.
  • [B] Ozone (O3) has its strongest UV absorption in the Hartley band (~200-310 nm), peaking near 255 nm; a weaker, vibrationally-structured Huggins band sits on the long-wavelength shoulder (~310-350 nm); the Chappuis band is a weak visible-region feature (~400-650 nm). The Hartley band is what makes ozone an effective UV-B/UV-C absorber.
  • [B] Measured O3 absorption cross-section near the Hartley peak is ~1.1e-17 cm^2/molecule. Independent laboratory determinations report 11.27e-18 cm^2 at the 253.65 nm Hg line (expanded uncertainty 0.86%) and 11.26e-18 cm^2 at 257.34 nm — confirming the ~1.1e-17 cm^2 figure used here is correct to within measurement uncertainty.
  • [B] Standard CIE/WHO/ICNIRP UV sub-band definitions: UV-A 315-400 nm, UV-B 280-315 nm, UV-C 100-280 nm. The 400 nm upper limit is used by ICNIRP, ACGIH, and WHO.
  • [B] One Dobson Unit (DU) = 2.69e16 ozone molecules per cm^2 (a 0.01 mm-thick layer of pure O3 at 0 C, 1 atm). Earth's average total ozone column is ~300 DU, equivalent to a ~3 mm pure-ozone layer at STP and a vertical column density ~8.07e18 molecules/cm^2.
  • [A] Worked example (vertical column, Hartley peak): tau = sigma*N = (1.1e-17 cm^2)(8.07e18 cm^-2) ~ 88.8, giving transmission exp(-88.8) ~ 3e-39 — effectively zero. Surface UV-C transmission through the ~300 DU column is essentially zero. The result is robust: even at a band-edge cross-section ~1e-18 cm^2, tau ~ 8 (transmission ~3e-4), and slant solar paths only increase tau. This is why solar UV-C does not reach Earth's surface.
  • [D] FENCE: Beer-Lambert attenuation is a physical absorption process (photon energy promotes electronic/vibrational transitions in O3, leading to photodissociation O3 + hv -> O2 + O). It reduces harmful UV reaching the surface. It is NOT a 'life force', and the shield function alone does not establish that ozone is alive or that ozone is sufficient for life. The shield is one decomposed component of the 'Ozone = Life' thesis, supported-within-model only as 'ozone provides UV attenuationّ.
    • Source: Domain-scope fence per discipline directive (claim decomposition)

Water photolysis, hydrogen escape, and multivariate habitability — establishing that water loss is governed by escape ph

  • [A] Escape velocity is v_esc = sqrt(2GM/R) and surface gravity is g = GM/R^2; these reproduce all seven verified reference bodies (Earth, Moon, Mars, Venus, Mercury, Jupiter, Titan) to <=0.36% (most <0.1%), where the only ~0.36% residual is rounding in the reference g for the Moon (calc 1.626 vs ref 1.62). Verified by computation this session.
  • [A] A pressure-proportional weight law g = k*P, calibrated on Earth (k = 9.684 m/s^2/bar), FAILS on every other body: it predicts g~3e-14 m/s^2 for the Moon and ~2.9e-14 for Mercury (near-vacuum, yet both have nonzero g of 1.62 and 3.70), predicts 891 m/s^2 for Venus (actual 8.87, a LOWER g than Earth despite 91x the pressure), and 14.2 m/s^2 for Titan (actual 1.354, the lowest g of the set despite a thicker-than-Earth atmosphere). Gravity is set by mass and radius, not by surface pressure. Verified this session.
  • [B] Hydrogen escape from a terrestrial atmosphere with a heavy background gas is typically DIFFUSION-LIMITED: the rate-limiting bottleneck is the upward molecular diffusion of hydrogen-bearing species through the homopause, not the exospheric (Jeans) escape rate itself. Hunten (1973) introduced this limiting-flux principle. On Earth the limiting flux is ~2.5e13 * f_H ~ 4.3e8 H atoms cm^-2 s^-1 for total hydrogen mixing ratio f_H ~ 1.7e-5.
  • [B] Water is delivered to the photolysis region (upper stratosphere/mesosphere) only as fast as it passes the COLD TRAP — the temperature minimum (Earth: tropopause, ~190-200 K) where the saturation vapor pressure of water sets a low stratospheric H2O mixing ratio. The cold trap, not ozone, is the primary gate on how much water reaches altitudes where UV can dissociate it; a colder trap means a drier stratosphere and slower water loss.
  • [B] Wordsworth & Pierrehumbert (2013) show CO2 increases water loss only over a narrow range: although high CO2 warms the surface (favoring more water aloft), it also COOLS the middle/upper atmosphere, making the cold trap more efficient and thus suppressing stratospheric H2O and water escape in most circumstances. Water loss is therefore controlled by the joint behavior of surface temperature, stratospheric temperature (cold trap), atmospheric composition, and stellar XUV — a multivariate outcome, not a single-factor one.
    • Source: Wordsworth, R. & Pierrehumbert, R. (2013), 'Water Loss from Terrestrial Planets with CO2-rich Atmospheres', ApJ 778(2):154. — https://arxiv.org/pdf/1306.3266
  • [B] Cumulative water loss over a planet's history depends on: (1) escape physics set by gravity/escape velocity (b, scale heights, whether escape is diffusion- or energy-limited), (2) the integrated stellar XUV history that powers photolysis and energy-limited hydrodynamic escape (especially the high-XUV pre-main-sequence phase of M dwarfs), (3) the initial volatile (water) inventory, and (4) the cold-trap efficiency. Ozone is not among the controlling variables for water creation or retention.
  • [D] Ozone does not create water and does not retain water. O3 is a photochemical product of O2 (Chapman cycle) that absorbs UV-B/UV-C; this UV shielding is a consequence of pre-existing O2 (itself largely a biological/photolytic product), and is logically and causally downstream of water inventory and escape, not a cause of water presence. Stratospheric water content is set by the cold trap; long-term water retention is set by escape physics and XUV history.

BIOENERGETICS — chemiosmosis, proton-motive force, oxygen as terminal electron acceptor, and anaerobic alternatives. Sco

  • [B] The 1978 Nobel Prize in Chemistry was awarded to Peter D. Mitchell 'for his contribution to the understanding of biological energy transfer through the formulation of the chemiosmotic theory' — establishing that electron transport pumps protons across a membrane to create an electrochemical proton gradient that drives ATP synthesis. This is the foundational, generally-accepted principle of bioenergetics.
  • [B] The proton-motive force (Delta_p) is the energized state of a membrane: the electrochemical potential difference for protons, expressed in voltage units, comprising an electrical component (membrane potential Delta_psi) and a chemical/pH-gradient component (Delta_pH). It is the universal intermediate ('proticity') coupling respiration/photosynthesis to ATP synthesis, transport, and motility across bacteria, archaea, mitochondria, and chloroplasts.
  • [B] Molecular oxygen (O2) is an exceptionally powerful terminal electron acceptor: the O2/H2O couple has a high standard reduction potential of about +0.82 V at pH 7, far above NAD+/NADH (~ -0.32 V). The large potential drop down the respiratory chain to O2 makes aerobic respiration the highest-yield catabolism, which is why O2 is preferentially used when a suite of acceptors is present. This is a thermodynamic advantage claim, NOT a necessity claim.
  • [B] Oxygen is NOT a universal requirement for life. Anaerobic respiration uses alternative terminal electron acceptors with lower reduction potentials: sulfate-reducing bacteria reduce SO4^2- to H2S; denitrifiers reduce NO3- toward N2; others use Fe(III), Mn(IV), CO2 (methanogens), or fumarate. Fermentation requires no external acceptor at all. These metabolisms yield less energy per substrate than O2 respiration but sustain entire ecosystems. Therefore 'life requires oxygen' is false as a universal statement.
  • [B] Life predates atmospheric oxygen. The earliest life (~3.5-3.8 Ga) was anaerobic; free O2 accumulated in the atmosphere only at the Great Oxidation Event (~2.4 Ga). Thus for roughly the first third-to-half of life's history, biology ran without molecular oxygen and necessarily without an ozone (O3) shield. This temporally falsifies any claim that oxygen/ozone is a precondition for life itself; oxygen-based metabolism is an evolutionary innovation layered onto pre-existing anaerobic chemiosmotic life.
  • [D] The chemiosmotic mechanism (proton gradient -> ATP synthase) is independent of oxygen: it operates identically whether the terminal acceptor is O2, sulfate, nitrate, fumarate, CO2, or light (photosynthesis). This decouples 'the foundational energy principle of life' (proton-motive force, B/A) from 'oxygen' — supporting that proton gradients, not O2 or O3, are the near-universal bioenergetic invariant. Ozone in particular plays NO direct role in any known cellular bioenergetic pathway (it is an atmospheric UV-shield molecule, outside-model-scope for chemiosmosis).

Solar energy collection and conversion under vacuum / controlled atmosphere — steady-state radiative-convective heat bal

  • [B] The solar constant (total solar irradiance at 1 AU, top of atmosphere) is approximately 1361 W/m^2; the IAU 2015 resolution and NASA SORCE both adopt ~1361 W/m^2. It is not truly constant — it varies ~0.1% over the 11-year solar cycle, so 'constant' is a nominal label, not an invariant.
  • [B] In an evacuated-tube collector a vacuum (typ. ~1e-5 mbar inside the tube) effectively eliminates gas-phase CONVECTION and CONDUCTION between absorber and envelope, but does NOT remove RADIATIVE loss; radiative emission still requires a spectrally selective absorber coating to suppress. Vacuum is incremental insulation engineering, not a new physical law and not a thermodynamics-erasing mechanism.
  • [B] The Shockley-Queisser detailed-balance limit for a single-junction solar cell under the AM1.5G spectrum is ~33.7% at an optimal bandgap of ~1.34 eV (the original 1961 SQ paper gave ~30% at 1.1 eV; 33.16% with an added back-surface mirror). Silicon (Eg1.1 eV) caps near ~32%. PV cooling yields only INCREMENTAL gains (cells lose efficiency as they heat); it does not breach the detailed-balance limit.
  • [A] Worked steady-state heat balance (1 m^2 absorber, eta_abs=eps=0.90, T=400 K, T_env=300 K, G=1361 W/m^2): P_abs=1224.9 W; radiative loss P_rad=epssigmaA*(T^4-T_env^4)=893.1 W; in vacuum (h=0) P_net=331.8 W. Equilibrium temperature is 423.3 K (150 C) for the pure-radiative vacuum case vs 369.1 K (96 C) when a convective coefficient h=10 W/m^2/K is restored — demonstrating that removing convection RAISES achievable stagnation temperature but the radiative term remains the irreducible ceiling. No configuration yields unlimited energy.
  • [A] Falsification cross-check (gravity vs pressure): using g=GM/R^2 reproduces all seven verified bodies to <=0.36% (binding residual = Moon, 0.360%; the tighter <=0.26% holds only at full-precision NSSDCA masses), whereas a pressure-proportional law g=k*P calibrated on Earth (k=g_E/P_E) fails catastrophically: Moon kP/true~1.8e-14, Mars ~1.6e-2, Venus ~100x (overshoot), Titan ~10.5x (overshoot). Gravity is set by mass and radius, NOT atmospheric pressure; vacuum does not abolish gravity and pressure does not generate it.

Exoplanet biosignatures — oxygen/ozone false positives & false negatives, atmospheric disequilibrium, and the decomposit

  • [B] Oxygen (O2) is a leading remotely detectable candidate biosignature gas, but it is NOT decisive on its own: abundant atmospheric O2 (and its photochemical product O3) can be produced abiotically (false positives) and can be suppressed even where life is present (false negatives). The defensible claim is contextual — O2/O3 raises or lowers the posterior probability of life only when interpreted within full stellar+planetary environmental context.
  • [B] Identified ABIOTIC O2/O3 false-positive mechanisms (no biosphere required): (1) CO2 photolysis on dry, low-N2 worlds around F/G/K stars; (2) water photolysis followed by hydrogen escape, favored under the high UV of active M dwarfs; (3) runaway-greenhouse ocean loss leaving an O2-rich residual atmosphere (notably for planets around M dwarfs). These show 'ozone (or O2) => life' is FALSE as a literal universal inference (evidence class X/U if asserted literally; survives only as a narrowed, context-gated hypothesis, class B).
    • Source: Meadows et al. 2018 (Astrobiology 18:630-662); Wordsworth & Pierrehumbert 2014; Luger & Barnes 2015; Tian et al. 2014, as reviewed in Schwieterman et al. 2018 (Astrobiology 18:663-708) — https://pmc.ncbi.nlm.nih.gov/articles/PMC6014570/
  • [B] FALSE NEGATIVES are real: a planet can host life yet show little/no detectable O2 or O3 (e.g. Earth's own ~2-billion-year anoxic/low-O2 Archean and weakly oxic Proterozoic eons would not have shown a modern O2/O3 signature). Target selection and observing strategy must account for false negatives, not only false positives.
  • [B] No single molecule is decisive. The mainstream framework assesses biosignatures via biogeochemical 'Exo-Earth System' models that compute Bayesian likelihoods of the observed spectrum under life vs no-life scenarios, mapping the posterior to five confidence levels (e.g. 'very likely' 90-100% to 'very unlikely' <10% inhabited). Four required components: (1) stellar/system characterization, (2) internal-planet/habitability characterization, (3) biosignature assessment in environmental context with corroborating evidence, (4) explicit exclusion of false positives.
  • [B] Atmospheric thermodynamic chemical DISEQUILIBRIUM is a stronger, more general life-detection logic than any single gas: among solar-system planets with substantial atmospheres, modern Earth has the largest disequilibrium, driven by life. Combination/anti-equilibrium gas pairs (e.g. O2+CH4 simultaneously, or CH4+CO2 with absent CO for anoxic worlds) are far more diagnostic than O2 or O3 alone.
  • [B] The O2-O3 relationship is NONLINEAR and strongly host-star-UV dependent, so O3 is an unreliable proxy for O2. For planets around hotter stars (G0V-K2V) ozone column abundance can PEAK at only ~25-55% of present-atmospheric-level (PAL) O2 — counterintuitively rising as O2 falls from 100% PAL — and an Earth-Sun analog has a similar O3 column at both 10% and 100% PAL O2. Trends differ markedly for cooler K5V-M5V hosts (smog-formation chemistry grows relative to the Chapman mechanism). CH4 and N2O further distort the mapping. Therefore detecting O3 does not cleanly imply a specific O2 level, let a
    • Source: Kozakis, Mendonca & Buchhave 2022, 'Is ozone a reliable proxy for molecular oxygen? I. The O2-O3 relationship for Earth-like atmospheres', A&A 665:A156; Part III (CH4) A&A 2025 — https://arxiv.org/abs/2208.09415
  • [D] Decomposition of the thesis 'Ozone = Life' into its distinct sub-claims (each judged separately, never collapsed): (a) LITERAL identity 'ozone IS life' — outside-model-scope / contradicted (X): O3 is a triatomic oxygen molecule, not an organism. (b) PERFECT BIOSIGNATURE 'O3 present <=> life present' — contradicted-by-test (X): both abiotic false positives and biotic false negatives exist. (c) CONTEXTUAL BIOSIGNATURE 'O3, within full stellar+atmospheric context and disequilibrium analysis, raises the posterior probability of life' — supported-within-model, narrowed (B). (d) UV SHIELD 'an O3 lay

Active Inference bound discipline (simulation agent layer only — model comparison + agent policy selection; NOT a physic

  • [A] Variational free energy is an UPPER bound on surprisal (self-information): F[q] >= -ln p(y|m). Lower F is better. Equivalently the negative free energy (ELBO) is a LOWER bound on the log model evidence: ELBO = -F[q] <= ln p(y|m). The bound is tight (F = -ln p(y|m)) exactly when the recognition density equals the model posterior, q(eta|r) = p(eta|y,m), i.e. KL = 0. Never write 'F is a lower bound on surprisal' — it bounds surprisal from above. This is the foundational identity of the agent layer.
    • Source: Derived in-repo: docs/MATH_DERIVATIONS.md lines 5-60 (F[q] = D_KL(q||p(eta|y,m)) - ln p(y|m) >= -ln p(y|m); 'VFE is an upper bound on surprisal'; ELBO = -F <= ln p(y|m)). Standard result: Friston 2010 — https://www.nature.com/articles/nrn2787
  • [A] q(eta|r) is the moving/optimizable recognition (variational) density parameterized by internal/recognition states r; p(eta|y,m) is the EXACT posterior over hidden causes eta UNDER THE MODEL m, conditioned on observations y. p(eta|y,m) is NOT 'the world' and NOT the generative process — it is the agent's model-internal posterior. The free energy gap above the bound is exactly the KL divergence D_KL(q(eta|r) || p(eta|y,m)) between these two model-internal densities. Collapsing 'posterior under the model' into 'the world/the process' is a forbidden model/process conflation (CLAUDE.md §2).
    • Source: docs/MATH_DERIVATIONS.md line 5 (definitions) + lines 20-22 (KL decomposition); CLAUDE.md §2 'q vs p' and 'model vs process' invariants.
  • [B] Expected free energy G(pi) scores candidate policies pi over FUTURE (counterfactual) outcomes and decomposes into a pragmatic/instrumental term plus an epistemic/information-gain term: G(pi) = E_q[ ln q(s|pi) - ln p(o,s|C) ], which separates as risk (expected divergence of predicted outcomes from preferences C) + ambiguity (expected likelihood entropy) MINUS expected information gain (the epistemic/exploratory drive is SUBTRACTED, so reducing G favors information-seeking). Policies are selected by Q(pi) = softmax(-gamma * G(pi)) (with habit/prior E), gamma the policy precision. Sign convention
    • Source: Sajid, Ball, Parr & Friston, 'Active inference: demystified and compared', Neural Computation 33(3):674-712 (2021); arXiv:1909.10863. Repo sign convention: CLAUDE.md §2 (canonical sign), §4 (EFE decom — https://arxiv.org/abs/1909.10863
  • [A] FENCE (hard boundary): active inference / the free energy principle frames MODEL COMPARISON and Bayes-optimal POLICY SELECTION within an assumed generative model. It does NOT prove any physical mechanism — it cannot establish gravity, ozone chemistry, a 'pressure-proportional weight law', the origin of life, or any claim about the physical process. Free energy here is information-theoretic (an energy-minus-entropy functional over densities), NOT literal thermodynamic Helmholtz/Gibbs free energy (no pV work, no literal temperature). Using AIF to argue any of those physical theses is outside-mod
    • Source: CLAUDE.md §2 'Thermo fence' (VFE is information-theoretic, not literal Helmholtz/Gibbs, no pV) + §0/§0.7 claim boundary; task discipline HARD BOUNDARIES.
  • [A] FENCE (status of fact, CLAUDE.md §0.7): NO real strong-sense active-inference loop exists in the live Rust deep-reader crate — the perceive->EFE->act->learn machinery (core.py) is PARKED. The running system is exact closed-set Bayesian model selection + count/cache prediction, NOT an active-inference agent in the strong sense (where ACTIONS change FUTURE OBSERVATIONS). For the simulation lab's agent layer, 'active inference' language is permitted ONLY for the model-comparison / policy-scoring math defined here; calling the live system an active-inference agent is forbidden until such a loop wi
    • Source: CLAUDE.md §0.7 'Missing tier — active inference' (verbatim: 'No real active-inference loop exists in the Rust crate'; 'say exact closed-set Bayesian model selection + count/cache prediction, not activ
  • [D] ANALOGY FENCE: 'trauma-locked agents' = agents with maladaptive priors and/or miscalibrated precision (e.g. pathologically low policy precision gamma, or over-confident likelihood precision producing rigid posteriors). This is a COMPUTATIONAL ANALOGY about the agent layer's parameters — it is NOT a clinical, psychiatric, or diagnostic claim about any human or animal. The phrase describes a parameter regime in the simulation, nothing more.
    • Source: Interpretive synthesis bounded by CLAUDE.md §0 claim boundary + §2 precision invariant ('affect moves precision inputs'); task discipline 'trauma-locked agents = maladaptive priors / miscalibrated pre
  • [B] CITATION-INTEGRITY note: the discipline's cite 'Sajid et al 2020' resolves to the verified record Sajid, Ball, Parr & Friston, 'Active inference: demystified and compared' — arXiv:1909.10863 (preprint 2019), published Neural Computation 33(3):674-712 (2021). There is no 2020 publication of record; cite as 2021 (journal) or 2019 (arXiv preprint), not 2020. Parr, Pezzulo & Friston (2022) 'Active Inference: The Free Energy Principle in Mind, Brain, and Behavior' (MIT Press) is the textbook reference for the EFE decomposition and is correctly dated.
    • Source: WebFetch of arXiv:1909.10863 abstract page (title/authors/venue verified); WebSearch confirming Friston 2010 NRN 11:127-138. Parr/Pezzulo/Friston 2022 MIT Press (not independently re-fetched this sess — https://arxiv.org/abs/1909.10863

sha256 47be1384c8879209 — of the original file, so what was ingested stays checkable.