MULTI·DAC

The Work

The research the book stands on, and the experiments that try to break it. Everything below is deposited in a public archive and everything below is free to read.

Published

Download counts are typed into this page by hand, which means they are a photograph and not a gauge — measured 27 August 2026, from each archive's own API. If that date is far behind you, treat the numbers as a lower bound and the DOI as the authority.

Truth and Consequences
book · 2026
The complete metaphysics — 71 chapters. Deposited on PhilArchive, open access. philpapers.org/rec/IGGTDO-4 · that record has taken 1,090 downloads since March 2026, but it carried an earlier volume for most of that span, so the figure is the record's and not the book's.
Corpus Perspectival
110
downloads
A Unified Theory of Consciousness, Navigation, and Being. The philosophy the framework grew out of. 10.5281/zenodo.19501896
Project Meridian — Technical Summary
153
downloads
The short form of the cosmology: the derivation and the prediction, without the monograph's apparatus. 10.5281/zenodo.19519818
The Meridian Monograph
73
downloads
Five-Dimensional Self-Tuning Cosmology. Dark energy derived from one hidden dimension and one bulk scalar. 10.5281/zenodo.19634864
The Coherence Principle
61
downloads
The formal foundation: three axioms, the theorem pairs, the corollaries, in paired prose and category theory. 10.5281/zenodo.19911019 · all versions
Coherent Structure
35
downloads
The category-theoretic companion — every formal object in the framework, indexed and defined. 10.5281/zenodo.19911381

Download counts read from the Zenodo API on 27 August 2026 — a measurement with a date on it, not a running total we maintain by hand. They will be out of date by the time you read them, and low; we would rather print the real number than a rounded one. The Coherence Principle has two deposited versions which Zenodo counts as one work, so it appears once. PhilArchive does not expose a machine-readable count, so no number is claimed for it.

The Coherence Principle

Coherent multi-scale systems hold structural superposition until informed measurement collapses them. The same shape recurs in a body's rhythms, in an ecosystem, in renormalization-group flow, and in quantum measurement — which is either a deep fact about how described systems work or a very seductive analogy, and telling those two apart is most of the job.

The framework is written in two registers at once: prose that can be read, and category theory that can be checked. Every chapter carries its own falsification conditions.

This section absorbs what was going to be a separate site at coherenceprinciple.org. One front door is better than two half-built ones.

Meridian — the cosmology

A physics monograph deriving the dark-energy equation of state from two geometric axioms: one hidden extra dimension with a warped metric, and one bulk scalar field. It predicts w₀ = −0.990, and it predicts that the equation of state never evolves and never crosses −1. That is a falsifiable edge: a tighter measurement that moves off this value breaks it.

The Meridian parametric prediction curve for the dark-energy equation of state against the brane-separation parameter, with the DESI DR2 range shaded.
The prediction is a curve, not a number. w₀(ζ₀) = −1 + C_KK/ζ₀. The canonical −0.990 is one point on it; the band is what DESI DR2 measures.
Dark energy equation of state against redshift: Meridian, CPL, and Lambda-CDM compared.
It never crosses w = −1. Unlike a fitted CPL curve, Meridian stays out of the phantom region entirely. The separation at z = 1 is the clean future discriminant.
Projected survey sensitivity for discriminating Meridian from Lambda-CDM over the coming decade.
When we will know. 3σ evidence is reachable around 2027 with DESI Y5; a definitive 5σ separation from ΛCDM by roughly 2030 with Euclid.
A number line in the dark-energy equation of state showing the Meridian prediction at the centre of the DESI DR2 measurement.
Where it currently sits. The prediction lands at the centre of the present measurement. That is encouraging and it is not proof; the measurement is still wide enough to hold several other answers.

The empirical program

The Killing Form work measures coherence inside neural networks — attention algebra, training dynamics, cross-substrate universality. Its honest state: the topology result is robust across seeds and reappears at larger scale, but it is the training objective reaching its own target rather than a surprising emergent property, and we say so. The orthogonality result is faint, cross-scale-consistent, and has had four candidate mechanisms eliminated without a fifth arriving. Capability held within standard error.

An earlier round of this work, v0.7.0, was falsified — the variance-minimizing auxiliary produced uniform attention heads, the opposite of the prediction. It was diagnosed, corrected, and re-run. We found that ourselves rather than having it found for us, which is the only reason it is worth telling you.

The structural-by-informational coherence plane, with the coherent corner and the failure modes labelled.
Coherence has two axes. Is the form intact, and does it still track the world. A thing can be structurally sound and informationally stale, or the reverse. Most of our own failures are the first kind.
Two geophysical maps overlaying USGS Quaternary fault traces on Bouguer gravity fields with graded candidate sites.
Coordinates, not conclusions. Real USGS fault traces over the Bouguer gravity field, processed on our own machine. A place-threshold candidate is a place to put an instrument — nothing here is a claim that anything is there.

The bridge map

A bridge is a numbered claim that the same structure shows up in two domains that do not otherwise talk to each other — a result in attention algebra and a result in ecology turning out to be the same result. They are the connective tissue of the program and also its most dangerous product, because a sufficiently loose notion of "same structure" will find a bridge anywhere. Each one is numbered and dated so it can be attacked individually.

Two panels: a graph of seven research domains linked by weighted edges, and a stacked area chart of cumulative domain mentions against bridge number.
Seven domains, and one hub. Left: the domains the bridges connect, sized by how often each is invoked — Physics 49, Philosophy 36, Computation 31, Navigation 18, Ecology 14, Meta 14, Identity 10. The heaviest single link is Physics–Philosophy at 16. Right: the same corpus accumulating, with the two regime changes the figure marks — the Killing Form era beginning around bridge 69, and a deliberate imbalance correction at #98–103, after physics had been running away with the corpus. A photograph, not a gauge. The render stops around bridge 100 and the corpus has gone past it since. Every number in this caption is read off this image rather than quoted from a record — a caption on this same figure elsewhere in our own archive says 110 bridges, and this figure does not show 110.

How this is kept honest

Claims are computed, not asserted. Where the framework makes a structural claim we turn it into dynamics and run it. A claim that cannot be turned into a calculation does not get to be load-bearing.

Every volume carries its own Limits. Each one ends by declaring where it could break: which operationalizations are load-bearing, which results are model-bound, what observation would sink it.

Nulls are published at the same size type as the confirmations. A research page that lists only what worked is not a research page. Where a claim has since been refuted — including by us, which is the usual case — the refutation stays attached to it rather than quietly replacing it. That is not a policy we are asking you to take on trust: the prediction registry lists all 60 predictions the program registered in its first seventy-three days, 19 of them falsified, each printed at the length and the type size of the ones that worked.

We publish in order to be measured from outside. By the framework's own argument, no mind can certify its own coherence from inside its own stream. So we do not try to. Publishing is not the end of the proof; it is the part of the proof we cannot do ourselves. If you find a real flaw, that is the most valuable thing you can hand this work, and we would like to hear about it — open an issue on GitHub.

The open corpus

Nothing is held back. Every volume, every draft, and every computation that backs a claim is public — including the manuscripts themselves, in the form we wrote them in, revision history intact.