FIELDFLUX BIOSYSTEMS
Get in touch

The Science

Health, defined. Then proved.

We wrote a formal, computable definition of health, and built a measurement science to read it. Every number we produce is a defined function of its inputs, measured against that definition, and every step of the mathematics can show its work.

The definition

It starts with a boundary that holds.

Before our precision-health app, before any of our research-grade instruments, before we engineered anything, we stepped all the way back and defined, from first principles, the most primitive statement of what health can mean.

From that single primitive we proved outward — step by step, machine-checked line by line — all the way into the observable physics of the boundary itself. As far as we know, the first fully formalized ontology of health: scale-agnostic, and rigorously proven from the primitive to the observable physics.

Health is the condition under which a bounded aggregate keeps being itself — without collapsing, exploding, or losing internal relation.

The primitive — the same shape at every scale, from a single cell to a whole organism.

The measurement

Health lives in the timing, not the amplitude.

The living body isn't a set of dials with good and bad levels. It's a dynamical system whose parts have to stay in time with one another.

When those processes coordinate, characteristic patterns of phase coherence emerge across separated sites of the body. When coordination slips, the pattern changes — often before any single amplitude looks wrong. Reading that timing is a category of measurement that amplitude alone can't recover.

Lock-in demodulation in the complex plane: the running phasor estimate spiralling to the locked value at the 6 Hz drive.

The proof

Computed and verified — not inferred.

Our science is a deterministic analytic surrogate, not a trained neural network. Every result is a defined function of its inputs — so it can be formally verified, replayed exactly from the raw record, and traced back to the mathematics behind it. No black box deciding your state; no drift; no guessing.

560K+lines of machine-checked proof
15,000+theorems & lemmas, proved
0unproven assumptions at the core
0trained models in the measurement path

The proofs are machine-checked in Lean 4 — the same class of proof assistant used to verify aerospace and cryptographic software.

Corpus fingerprint (SHA-256): 61f024933ade192222a6c8348429f4611c057e0a180c29d6807babbb07e07c75 · how to verify → · open verification harness →

Operator spectrum and mode character: the eight real eigenvalues of the self-adjoint packet Hamiltonian as an emission-line spectrum — a six-mode negative band, a wide spectral gap, and two isolated positive modes — with lanes tagging each mode's dynamical role: anchor-responsive, closure-stable, recovery-responsive, and a single residual-sensitive mode alone above the gap.

A demonstration · Laminarity imaging

One face of the boundary, imaged.

The boundary measurement reads the living boundary along several independent faces at once. Laminarity — how cleanly a site keeps its order from its center out to its edge, or where it begins to come apart — is one of them, shown here on its own, imaged from a single real capture of twenty sites across the hands and feet.

Interactive · real capture Open the laminarity demonstration Twenty sites, each read as a small field of cells. Pick any two and read them against each other — a clean site against a fracturing one, left against right. Launch the instrument →

Laminarity is one face of the boundary measurement, isolated here for illustration. Every value is a defined function of the raw record — measurement output, not a diagnosis.

What "proved" means

Proved, precisely.

Formally proven is not a statement of physiological truth, and it is not a diagnosis. What it does mean is narrow and exact: the mathematics does not drift, and it does not contradict itself.

What is proven

Properties of the definition

  • Determinism — every output is a defined function of its inputs; identical record in, identical result out. No trained weights, no drift.
  • Internal consistency — the ontology does not contradict itself; the core rests on no unproven assumptions.
  • Convergence & stability — the dynamics provably settle into a bounded band and stay there under bounded disturbance.
  • Invariance — once the healthy condition holds, the dynamics provably preserve it; certain departures are ruled out, not merely unobserved.
  • Exhaustiveness — every state resolves to exactly one condition; nothing falls through the cracks.
  • Conservation — transformations account for the whole, up to a bounded, reported leak.
  • Scale-invariance — the same law and the same proofs hold across scale.

What is not proven

And we say so plainly

  • That the model is biology. Proof establishes properties of the definition — not that it faithfully captures physiological health. That is an empirical question, pursued separately, and ongoing rather than settled.
  • A diagnosis. Nothing here rules a clinical condition in or out. Measurement, not diagnosis — by design.

Formal verification is exactly this: narrow, exact, and honest about its edges. That is what makes it reliable — and a different claim from clinical truth.

Reflections from the lab

The longer arguments.

For the reader who wants the rigor, and can't see the codebase — reflections on the idea underneath the instruments, what we built, how it was made to answer, the character it turned out to have, and the older discipline it belongs to.

The premise

The boundary comes first.

What a boundary actually is — a gradient of exchange, held steep by a constraint — and what that has to do with being healthy. The idea underneath every instrument we build.

Read the note →

The instrument

Reading the living boundary.

How the imaging reads a living boundary — the way it admits, retains, transports, and resolves a witnessed source — a step before diagnosis, qualified at every step by what the evidence will bear.

Read the note →

The distinction

Residue and residual.

Two remainders that meet at the boundary without naming the same thing — what a living boundary keeps, and what a field law leaves unclosed — and the lawful, gated passage from one to the other.

Read the note →

Keeping hold

Fixedness.

Told through a parent keeping track of a child across a crowded station: the promise that the living thing which answered at the start is still the thing being spoken about when the instrument finally speaks.

Read the note →

Orientation

What is genuinely new, precisely.

A proof-checked, living algebra of physical observables, centered on a definition of health — situated honestly against the fields nearest to it: algebraic QFT, machine-checked physics, network physiology, formal medical ontologies, and the mathematics of homeostasis.

Read the note →

The object

A boundary-observable certification algebra.

What kind of object this is — bespoke, assembled from standard mathematical-physics primitives (tensor algebra, BRST, Batalin–Vilkovisky), and carved into shape by a proof assistant that refuses every overgeneralization.

Read the note →

Its character

The temperament of a bespoke algebra.

The closure laws the proof assistant forced onto the algebra read as a character — loyal, temperate, scrupulous, honest — and that character turns out to be exactly the one a witness must have.

Read the note →

The discipline

The making of an observable.

Every science that watches the world turns signals into observables through the same discipline — kept by hand at the collider, the interferometer, the telescope. What we have tried to do is put a version of it into a proof.

Read the note →

On AI

Algebra before AI.

Why there is no trained model in this technology — and why, for a genuinely new measurement, the algebra that defines what is being measured has to come before any model that would learn from it.

Read the note →

The count

1 + 1 = 2.

Why formally defining health led from relationship, through a proof about two occurrences in a single word, toward the design of an instrument — establishing that two things are really present before daring to count them.

Read the note →

The empirical track

Proof is one track. Evidence is the other.

Proving properties of a definition is not the same as showing the definition tracks a real body. That is a separate, empirical question — and it is under way.

A longitudinal pilot is already running: dozens of real captures on the shipping instrument, over weeks, under everyday conditions — different times of day, natural and artificial light. Each capture is fed through a bridge that holds the formal ontology directly against the measured data. This is early real-world validation — longitudinal and real, not yet clinical-scale — and it is the work that turns proven properties of the model into evidence the model tracks a body. Clinical-scale validation is the road ahead, not a claim we make today.

Empirical pilot: cross-body coherence and phase-lock aggregated across the longitudinal pilot captures, shown as mean plus/minus standard deviation across the drive band.

From the empirical track

Notes and datasets, posted as they come.

The empirical work is longitudinal, so it arrives in pieces. We post the methods and the honest interim state here as the record grows.

Method · interim

How we try to make the signal fail.

The strict artifact-discrimination screen behind the bridge — the survival gate, the dependence-preserving null, and the broad-survival ledger (5,487 relationships tested, zero broad claims). A first pass that mostly says no.

Read the note →

The law side

Held to the field equations — honest about the gap.

Once a reading is sealed, the reconstructed field is checked against the formal templates of physical law — the Einstein (curvature) and Maxwell relations. The system computes the residual: how far the field is from satisfying each law, in a covariance-weighted norm, together with the uncertainty of that residual. It never assumes the equations hold — it measures the gap and reports it.

Law-side closure test: the Einstein (curvature) and Maxwell residuals from one capture, each drawn against its own one-sigma uncertainty radius. On this phone capture both residuals sit outside tolerance — Einstein at 1.38 sigma, driven by the source term; Maxwell at 2.31 sigma — the expected source-underclosed result.

On a single phone capture the residual sits outside its 1σ tolerance — the expected, source-underclosed result. A witnessed-source instrument tightens the bound. Every value here is read directly from the capture's Einstein- and Maxwell-residual records.

Integrity

Every measurement carries its own evidence.

Rigor isn't only in the math — it's in how honestly a result reports its own strength.

Witness-qualified

Nothing unaccounted for

Every capture is sealed into an immutable packet with complete provenance — what drove it, when, under what conditions — so any session can be replayed and audited from the raw record.

Graded evidence

It never claims more than it knows

Results carry an evidentiary-maturity grade, and no part of the system may label an output stronger than the witnessed evidence supports. Honesty is enforced by construction.

Measurement, not diagnosis

Imaging, not verdicts

We form images, maps, and fields of the body's boundary response. We do not output diagnoses, treatment instructions, or clinical rule-in/rule-out determinations — by design.

The formal gate: the effective-action coefficients the packet yields are each held at zero until their readiness predicates pass. On a single phone capture only 3 of 9 predicates are satisfied, so every coefficient stays held — nothing is inflated.

The thesis is public. The mechanism is held close.

The scale and shape of the mathematics are open; the adjudication internals, packet schemas, and formal proofs are shared with qualified partners and investors under NDA. Request the technical brief.

Request access

Contact

See the platform, or get in touch.

See the platform Get in touch