Gaia — the laboratory boundary
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.
RENDERED FROM STRUCTURED DATA — this page is not a markdown file. Source:
reader/plates/PL-10.jsoninuni-cookbookat575fc93d9d31. The sha256 recorded for this page is the digest of that JSON, not of this text. The renderer isgenerators/derive_docs.cjs; it copies fields and adds no facts. Every value below is a field of the source, verbatim, including the ones that readNOT RUNandPENDING.
The drawing below is
reader/plates/PL-10.svgat sha256a66550ad16827931.
The world/agent coupling, drawn so that the model/process distinction cannot be missed. Panel I places the laboratory named Gaia — the generative PROCESS, world.process — on the left under hatch, and the agent holding the generative MODEL, model.p_os = p(o,s), on the right. Between them is the Markov blanket b = {s,a}: s is the world's only route in, a is the mind's only route out, and the agent's inference touches nothing else. Inside the agent's internal states hangs a framed miniature, and it is the plate's load-bearing detail: it depicts the BLANKET SURFACE, not the world, because NA-04 says the generative model 'is not a model of eta; it is a model of how eta is expected to show up in b'. A forbidden double-headed path is cut with a cross exactly where it tries to cross the blanket wall — the oracle cut — because any crossing that is not s or a is forbidden. Panel II states the partition formally. Panel III draws the human visual blanket as a funnel to scale in width at exactly 83:1 (498 px mouth / 6 px throat) from measured counts, and marks the ratio MODELED because it divides unpaired cohorts across a non-uniform surface. Panel IV records that one token, 'Gaia', names two different subjects in two different ledgers. Panel V draws the Pearl-to-Friston slide and carries Aguilera's narrow-corner result as a first-class NEGATIVE. Panel VI sets the two emptinesses side by side — NOT-MEASURED in nature against not-yet-built in UNI — separated by a sovereignty mark, because the resemblance is the trap. The Gaia side of this plate carries no number at all: that is a finding, not an omission.
Claims
| ref | symbol | value | units | scope |
|---|---|---|---|---|
| mu perp eta | b | p(mu,eta|b) = p(mu|b)p(eta|b) | — (definition) | any system admitting the 4-way partition | |
| I(mu;eta|b) | 0 (required) | nats | the operable blanket test, any system | |
| H_mu-eta, H_eta-mu | 0 (required) | precision units | linear Gaussian / Ornstein-Uhlenbeck systems | |
| N_rod | 92e6 (range 77.9-107.3e6) | cells | human retina; 8 wholemounts, 7 donors, ages 27-44 | |
| N_cone | 4.6e6 (range 4.08-5.29e6) | cells | human retina; same cohort as NA04-04 | |
| D_cone,fovea | 199000 (range 100000-324000) | cones/mm^2 | human foveal peak; same cohort | |
| N_optic | 1159000 +/- 196000 (range 816000-1502000) | axons | human optic nerve; 22 nerves, 19 subjects, ages 20-75 | |
| r_retina | ~83 : 1 | dimensionless | human visual blanket, order-of-magnitude only | |
| n_levels | — (no value carried) | nested blankets | any named organism, established by measurement | |
| Gaia (regulation) | — (no value carried) | — | the biosphere homeostatically regulates the planet | |
| Gaia (the laboratory) | not-yet-built | UNI 4-value fence | UNI's own build status |
What this plate does NOT claim
- That Gaia exists as a running artifact. Gaia has zero rows in CLAIM-LEDGER; UNI fence: not-yet-built. This plate draws a DESIGN, and a design discharges no gate.
- That any agent drawn here has a Markov blanket in the Friston sense. No I(mu;eta|b) has been estimated for any UNI boundary, and NA04-02 cards it NOT-MEASURED for essentially all real biological systems. The blanket on this plate is a partition drawn by a builder — the Pearl kind, the free kind.
- That a Markov blanket implies a self, an experiencer, awareness, sentience, consciousness, or a point of view. Explicitly disclaimed in NA-04, and no falsifier is offered because none exists. The same formalism admits partitions for a Newton's cradle, a burning fuse and the Lorenz attractor (Beck & Ramstead 2025). If a blanket implied a self, it would imply one for the burning fuse.
- That the free-energy principle is true, or that blanket conditions are general. In the one setting where the question was solved exactly they held only in a 'very narrow space of parameters' (Aguilera et al. 2022) — carried as a first-class NEGATIVE, not softened.
- That naming a laboratory 'Gaia' asserts anything about Earth, the biosphere, or a living planet. And not claimed that the Gaia hypothesis is settled — CN03-76 carries regulation as HYPOTHESIZED with the unit-of-selection objection live and unanswered.
- That 83 : 1 is a measured ratio, or that the sensory surface is uniform. It is MODELED across unpaired cohorts (NA04-08); foveal ~1:1 against peripheral >>100:1 is the fence, and the drawing is the thing being fenced.
- That a low F certifies the model. NA-02, honest bound 1: 'A confidently wrong model can sit at low F ... the whole construction is conditional on p(o,s) — a choice, not a measurement. Minimizing F never tests whether the generative model was the right one.'
- That anything here raises any UNI rung. A NATURE CITATION IS NEVER A UNI GATE. The honest program position is unchanged by this plate: ~2 of 11+ developmental rungs earned, a developmental active-inference SIMULATION, a toy world, never a person.
- 'Full human' and 'beyond human' are not targets, milestones or deliverables anywhere on this plate — they are permanent open questions (QUAESTIO-APERTA).
Source chapters
[object Object][object Object][object Object][object Object][object Object][object Object]
sha256 09bab9fa2030d4b0 — of the original file, so what was ingested stays checkable.