S-L3 - Organ: the Karaaslan Heart Lab
How to read this page
Three ways to read this page. Precise is the document itself, exactly as it is written in the repository. Plain and Clear were written for this website to help you meet that document — they are about it. They are not it, and they are not evidence.
The Encyclopedia is the UNI method written out as a reference work: 39 pages, arranged in wings, setting out what the programme is attempting and why it is built the way it is. This is where the ideas are explained in order and in prose, rather than as code, as runbooks, or as dated receipts.
Every chapter is authored against two ledgers and never ahead of them. One records what UNI has built, and the evidence class of each claim. The other records nature's own regularities, kept separate on purpose. That way a fact about biology is never quietly reused as a fact about the software. Where a chapter and a ledger disagree, the chapter is the thing that is wrong. Every chapter closes with an invitation to falsify it, and a recorded negative is published beside the result it qualifies rather than after it.
Read "How to read this work" first. It is the evidence constitution: the classes, the four ledger states, and the rule that a finished chapter is not the same as a working system. Then the calibration ledger, which carries the figures every other chapter is required to use.
What it is not: a description of a person or of a mind. The programme calls itself a developmental active-inference simulation, a bounded peek into a toy world, and its own index prints how much of the developmental ladder has actually been earned — roughly two rungs out of eleven or more. It is also not a report of what is running today. For what ran, and when, go to the evidence record.
Your browser cannot switch reading levels, so the document itself is shown.
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.
Honest position. This chapter describes the Heart Lab, the L3 ("organ / physiological control") rung of the UNI developmental ladder. It is a browser-native teaching simulation that re-reads a published clinical cardio-renal model in active-inference language. It is a developmental SIMULATION: a toy world, not the real world. It is not a clinical tool and not a diagnostic instrument, and nothing in it should be acted on as medical fact. What is earned here is narrow and engineering-grade: a faithful, parity-tested re-expression of one reference model, gated through the program's evidence discipline. The honest program position holds: roughly 2 of 11+ developmental rungs are earned, and L3 is one of the modest, proven rungs, fenced exactly to what it shows.
What the Heart Lab is
The Heart Lab (heart-lab.html) is one of a family of self-contained interactive "labs" that each run the same active-inference / predictive-processing machinery in a different domain: Precision (a POMDP maze agent), Echo (the same engine with an echolocation observation model), Loop (a dynamical-systems bifurcation), Cell (an autopoietic service cell), and Heart. The thesis the family is built to teach is "one organism, one math": same math, many scales, from cellular viability up through physiological homeostasis to cognitive precision-weighting. The Heart Lab is the physiological/clinical face of that family, built to teach doctors across disciplines.
Its scientific basis is the Karaaslan et al. long-term cardio-renal model, a published mathematical physiology reference (a reduced RSNA -> MAP -> sodium/volume loop: renal sympathetic nerve activity driving mean arterial pressure, coupled to sodium and fluid-volume regulation). The Heart Lab re-expresses that reference loop in active-inference terms, framing clinical homeostasis as prediction-error minimization and a heart attack as a failure of the prediction loop: maladaptive priors, miscalibrated precision, or a damaged generative process. Crucially, the underlying Karaaslan mathematics is reproduced; the program's own patent-level UNI math is not reproduced here. The framing stays at textbook level.
Architecturally the Heart Lab follows the family's reusable spine. The page carries an inline JavaScript engine, delimited by explicit parity markers; the canonical model lives in TypeScript under api/_lib/worlds/cardio_renal.ts; and a tsx parity test pins the inline engine to the canonical TS engine so the page and the "real" model cannot drift. Determinism is enforced with a seeded PRNG and committed result artifacts, so any displayed result is reproducible from its seed.
The exact ledger claim (L3.1) and what it does not say
The single ledger row for this rung is L3.1, and the chapter is authored against it verbatim.
- Claim (L3.1). The Karaaslan cardio-renal Heart Lab models clinical homeostasis as prediction-loop failure on the same one active-inference engine (the reduced RSNA -> MAP -> sodium/volume loop, re-expressed in active-inference language).
- Evidence class. C (dev-gate / held-out eval) for the lab gate, and Class E (test-covered) for the Heart-Lab engine ticket OAS-710-T3, recorded at 15/15 tests passing.
- Status. Proven (toy / clinical-model), at the classes above and no higher.
- Source.
uni-precisionarchive.
It is worth stating exactly what the Class-E engine ticket earned, because the temptation is to read it as more than it is. OAS-710-T3 added parameter clamping and a labeled safe-state to the engine: CARDIO_BOUNDS, a clampState() that clamps all ten state variables every tick (finite-safe), and a labelCardioState() that returns a clinical regime label (hypertensive crisis / decompensated / hypertensive / hypotensive / normal) instead of a raw clamp or a NaN. The full TDD cycle ran, 15/15 tests pass, the typecheck is clean, and there was no sibling regression. The companion viewer ticket OAS-710-T2 (the inline-JS Canvas viewer, with a fixed-timestep render loop and a pause-on-hidden behavior) was built test-first as a later ticket; the program's own discipline refused to mark a viewer criterion satisfied before the viewer ticket actually existed, rather than fabricating a verdict. That refusal is part of the evidence: the 15/15 is an engine-stability and labeling result, not a clinical validation.
The class is the ceiling, and here the ceiling is low and exact. Class C means a pre-registered, gated lab artifact; Class E means tests pass. Neither class licenses any claim about real hearts, real patients, or real clinical outcomes. A passing test (Class E) is not a runtime observation of clinical fact (Class A), and DONE here means test-covered, not "true of medicine."
The travelling fences (the negative that rides with the claim)
L3.1 carries no separate NEGATIVE row in the ledger; its load-bearing negative is the fence on interpretation itself, and that fence travels with the claim and must never be stripped. The same honesty discipline is coded into the lab artifacts, not merely the prose: sibling labs in this family tag interpretive claims [HYPOTHESIS], state that "the resemblance is an interpretive act, not a measurement", and list explicitly: not a clinical tool, not a diagnostic instrument, not evidence that active inference is the correct theory of mental health (or, here, of cardiology). The Heart Lab is a computational hypothesis sandbox.
So the claim and its fence are inseparable. The Heart Lab reproduces a published model's behavior and labels its regimes; it does not predict a patient, diagnose a condition, or establish that the active-inference reading of homeostasis is correct. The "heart attack as prediction-loop failure" framing is a statement about the toy model, an interpretive lens laid over the Karaaslan equations, and it does not transfer to clinical reality. Where the program's preprint (Polzin et al. 2026, Zenodo DOI 10.5281/zenodo.19785799, MIT) underwrites the shared mathematical foundation, it is the foundation only, and always fenced unrefereed: its Layer-1 AI-executable audit is complete, but Layer-2 human expert review is pending. The preprint is never proof that active inference is the correct theory.
What is NOT claimed in S-L3 (the Heart Lab)
- Ceiling. It is NOT shown that UNI predicts, diagnoses, or models any real heart, patient, or clinical outcome, and it is NOT shown that active inference is the correct theory of cardiovascular disease. The most we claim is exactly L3.1: a reduced Karaaslan RSNA -> MAP -> sodium/volume loop, re-expressed in active-inference language as a toy "same math, many scales" demonstration, proven as a Class-C lab gate with a Class-E engine ticket (OAS-710-T3, 15/15 tests), and nothing above those classes.
- Fences engaged. Red line 7 (never raise a claim above its source evidence class: this stays at C/E, never A, never "clinically validated"). Red line 10 (no patent-level UNI math: only textbook-level Karaaslan physiology is reproduced). Red line 11 (the preprint is the mathematical foundation only, fenced unrefereed, Layer-2 review pending). Plus the chapter's load-bearing fence: not a clinical tool, not a diagnostic instrument; the prediction-loop-failure reading is about the toy model, not clinical reality.
- Negatives that travel with this claim (cite alongside, never strip). The interpretation fence is the travelling negative: "the resemblance is an interpretive act, not a measurement", and the coded artifact disclaimers "not a clinical tool / not a diagnostic instrument / not evidence that active inference is the correct theory". The Class-E 15/15 is engine stability + regime labeling (OAS-710-T3), NOT clinical validation; the viewer was a later ticket and its criterion was deliberately left unmarked rather than fabricated.
- Parked / owed. No sealed sign-to-park and no Class-A observation are owed for L3 specifically (it is a proven toy/clinical-model rung at C/E, not a parked frontier). The standing program owe remains the preprint's Layer-2 human expert review, which is pending and gates how strongly the shared mathematical foundation may ever be leaned on.
- One-line honest summary a skeptic could not dispute. UNI faithfully re-expressed one published cardio-renal model as a parity-tested, regime-labeling teaching simulation (Class C lab gate, Class E engine ticket, 15/15 tests), and that is all: it is a toy world, not a clinical instrument.
Falsify this
The lead falsifier is operable and pre-registered: if the Heart Lab's predictions diverge from the Karaaslan reference model beyond the pre-registered tolerance, or fail the parity tests against the canonical TypeScript engine, the claim fails. Concretely, run the committed seed through the inline page engine and through api/_lib/worlds/cardio_renal.ts; if the tsx parity test does not hold the two in agreement, or if the engine ticket's 15/15 suite drops below 15/15 on a clean run, L3.1 is falsified. (Note the falsifier is purely about fidelity to the reference model and to its own engine; there is deliberately no falsifier offered for any clinical claim, because no clinical claim is made.)
Sources
- Digest:
curated/uni-precision-digest.md(the Heart Lab roster entry; the Karaaslan cardio-renal basis; OAS-710-T3 engine ticket at 15/15 and OAS-710-T2 viewer; the inline-engine + canonical-TS + parity-test triad; the refusal-to-fabricate-a-Y-verdict discipline). - Digest:
curated/worldmodels-digest.md(the shared active-inference engine and precision-knob framing; the coded honesty fences "not a clinical tool / not a diagnostic instrument / not evidence that active inference is the correct theory"; the preprint Polzin et al. 2026, Zenodo DOI 10.5281/zenodo.19785799, MIT, fenced unrefereed with Layer-2 review pending). - Ledger:
encyclopedia/CLAIM-LEDGER.md, row L3.1 (Section 1, the L0-L12 developmental ladder). - Archive pointer (no PII):
...-Precision.
sha256 d6d16d96c741a549 — of the original file, so what was ingested stays checkable.
Plain — written for this website, not the source document
A published cardio-renal model from the physiology literature is re-read here in the vocabulary of prediction and inference, and it runs in a browser as a teaching simulation. This is not a clinical tool, it is not a diagnostic instrument, and nothing it displays should be treated as medical fact. What the chapter does claim is small and engineering-shaped: one reference model faithfully re-expressed, pinned by a parity test, and graded through the program's evidence discipline. Homeostasis is described as prediction error being driven down, and a heart attack as the prediction loop failing. That description belongs to the toy model and does not carry over to a real patient. The chapter's own phrase for the likeness is that it is an interpretive act rather than a measurement.
Plain · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is d6d16d96c741a549
Clear — written for this website, not the source document
The lab is one of a family of self-contained interactive pages that each run the same machinery in a different domain, and the family exists to teach one thesis: same math, many scales. This one is the physiological face of it, and it is a teaching simulation — a toy world, not a patient and not a person.
Its scientific basis is a published long-term cardio-renal model, a reduced loop in which sympathetic nerve activity drives mean arterial pressure, coupled to sodium and fluid-volume regulation. The lab re-expresses that loop in inference language. The underlying reference mathematics is reproduced; the program's own private mathematics is not. Architecturally it follows the family's spine, with an inline engine on the page, a single master copy of the same model held separately, and a parity test pinning the two so the page and the real model cannot drift. A seeded generator and committed result artifacts make any displayed result reproducible from its seed.
The single row in the ledger, a record added to and never edited, is graded at two classes at once: a gated lab artifact, plus a test-covered engine ticket recorded with a full passing suite. The chapter is careful about what that ticket earned. It added bounds and a safe state to the engine, clamping the state variables every tick and returning a labelled clinical regime instead of a raw clamp or an undefined value. That is an engine-stability and labelling result, not a clinical check. A companion viewer was built as a later ticket, and the program's own discipline refused to mark a viewer criterion satisfied before that ticket existed, rather than fabricating a verdict.
Because there is no separate negative row, the travelling negative is the limit on interpretation itself, and it is coded into the artifacts rather than left in the prose. Interpretive claims are tagged as hypotheses, the resemblance is called an interpretive act rather than a measurement, and the disclaimers list what the lab is not.
What would show it wrong is deliberately about fidelity and nothing else. Run the committed seed through both engines and check that the parity test holds them in agreement, or watch the engine suite for a drop. Nothing at all is offered that could overturn a clinical claim, because no clinical claim is made.
Clear · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is d6d16d96c741a549