A frozen design, in plain sight
The A=1 lattice is public, parameter-free, and was frozen weeks before the first paper. That refutes retrofitting — but the honest answer to the retrodiction charge needs a second leg: minimality.
Twice now the program has met the same objection: the wall of PASS rows — the Dirac equation, the hydrogen spectrum to four significant figures, gravitational time dilation — is retrodiction, the lattice shaped to reproduce physics we already knew. It is the right objection to raise, and it deserves a precise answer rather than a defensive one.
The precision starts with noticing that “retrodiction” names three different accusations, not one — and they need three different answers.
- Post-hoc tuning. You had free parameters and turned them, after the fact, until the output matched the data. Overfitting. Fatal.
- Design-time selection. You turned no knobs after the fact — but you chose the architecture, from a space of alternatives, already knowing the target physics. The serious charge.
- Benign reproduction. You derived something already known. What every unification must do.
Most defenses of a framework like this quietly answer only (1) and call it a day. (1) is the easiest to refute and the least interesting. The charge that actually bites is (2). So let me take all three, in order, and be honest about which leg of the argument carries which.
(1) Post-hoc tuning — refuted by the freeze
There is nothing to tune. The A=1 lattice is a fixed minimal object — six basis vectors, a single conservation axiom \mathcal{A}=1, one update rule — with no free parameters but the Planck-scale lattice spacing, which is fixed, not fitted. You cannot curve-fit a model that has no knobs.
And the design is not hidden where a knob could lurk. Its complete specification — every vector with its coordinates, the conservation law, the chirality assignment, the tick rule — is the figure that previews every page of this site. You cannot read a word of this program without seeing the whole architecture; there is no private layer to adjust, because the design is on the thumbnail.
That figure is also older than the audited results, and frozen. Its source, figures/lattice.drawio in the public Paper I repository, was committed in late March 2026 — weeks before Paper I’s first Zenodo deposit. Its earliest committed version already carries the entire design-bearing content: all six vectors, the RGB/CMY ↔︎ ψ_R/ψ_L roles, the even/odd tick rule, and \mathcal{A}=1. Not one coordinate — the whole spec. You need not take my word: pull the file’s history in the public repo and watch it resolve to that frozen specification, unchanged since; the deposits themselves are immutable and DOI-stamped. The demonstration was dated after the design, in version control and at Zenodo.
So post-hoc tuning is refuted, and refuted checkably. There were no knobs, the design is public and complete, and it provably predates the audited demonstration of these results. (Note the careful wording: the hydrogen spectrum predates the lattice by a century — that is why it was available as a target. What predates the freeze is not the physics, but the audit.)
(2) Design-time selection — bounded by minimality
Here is the charge the freeze does not touch, and I will not pretend it does. A timestamp proves I did not cheat after March 2026. It cannot prove I did not try twenty lattices on paper before I committed the one that worked. Choosing octahedral over cubic, six diagonals of the eight, three colors, the \mathbb{C}^2 \to \mathbb{C}^{12} extension — that is selection, performed by a designer who already knew the target. I have said as much plainly: the color structure was a deliberate choice, added to give the design a better chance of producing physics. That is design-time fitting, and version control cannot see it, because it happened before the first commit.
The freeze is the wrong tool for this charge. The nearest right one is minimality — though, as we will see, it bounds the charge rather than killing it. The freeze says “I could not have cheated after.” Minimality says “I could not have cheated much before.” A design’s capacity to fit by selection scales with the number of independent structural choices it contains; a minimal design is drawn from a small space, which limits how much design-time fitting it can encode.
But minimality has to be claimed honestly, because forcing is always forcing-given-premises. Paper I states that, given coordination six and a split into chirally-opposite triples, the bipartite octahedral structure is the unique \mathbb{Z}^3 solution. Grant it — it still does not end the regress; it moves it up one level, into the premises. Why coordination six? Why chirally-opposite triples? Those are posited in Paper I, not derived from anything deeper (the paper’s own “forced by the bipartite octahedral geometry” is, read strictly, circular). And the second premise is the seam a referee should press, because “chirally-opposite” is the requirement that most looks chosen to deliver spinors. I will not pretend the uniqueness claim terminates the selection. It relocates it.
What minimality can honestly claim is that the premises it relocates the choice into are few and abstract:
- a \mathbb{Z}^3 adjacency — because space is three-dimensional;
- a single conservation axiom, \mathcal{A}=1;
- a two-sublattice tick — a tick needs a place to go and a place to return;
- six cube-diagonal directions, with the chirally-opposite split.
The fewer and more abstract the premises, the less room target-knowledge has to hide in them — but “less” is not “none,” and the color/chirality encoding is where I most openly chose. I added three colors to give the design more room to produce physics; three because space is three-dimensional — the same three that become the spatial gamma matrices. Whether that identification, color-three is space-three, is a principle or a coincidence wearing the word, depends entirely on what follows from it. If color and space share an origin, that should imply relations standard physics does not expect — a checkable claim, a candidate bite. If nothing follows, “principled” is doing work the physics has not yet earned. I hold it as the former and concede it could be the latter.
So minimality does not abolish design-time selection. It compresses it into a short, stated, increasingly abstract list of premises — real progress against a hostile read, but a bound, not a refutation.
(3) Benign reproduction — and the only thing that clears selection
Even granting all that, minimality only bounds design-time selection; it does not abolish it, and it certainly does not turn reproduced physics into novel physics. Recovering the hydrogen spectrum from a frozen, parameter-free, minimal design is a unification — real and hard, the thing every foundational theory must earn — but it is not yet a novel prediction.
What clears both the residue of selection and the benign charge at once is a single observation: you cannot select for, or reproduce, an answer no one has measured. A designer with the hydrogen spectrum in his head can, in principle, hunt for an architecture that reproduces it. That same designer cannot tune or select toward the value of the optical-axis birefringence, or the radius of the quantum Roche limit, because there is no measured target to aim at. Selection is powerful against the past and powerless against the future.
So the program’s weight does not, finally, rest on the PASS wall at all. It rests on the predictions that are not yet known to be true — the kinematic and gauge-sector birefringence along (1,1,-1), the multi-channel concordance, the quantum Roche limit and its atomic-ISCO signatures. The claim map labels those honestly as STUB: closed-form, untested. They are the bites — and they are selection-proof, because a frozen minimal design cannot have been fitted to a number that does not exist yet. If one of them lands, no story about a designer’s foreknowledge can explain it away. If it fails, the framework loses, as it should be able to.
Three legs, not one
That is the honest shape of the answer, and the three legs are not co-equal. The freeze closes the temporal hole: no tuning after the fact, and the dates are public. Minimality bounds the selection hole without closing it — it compresses the design to a few abstract premises, but every “forced” result still bottoms out in a premise that was, at some level, posited. Only the novel predictions escape the regress entirely, because you cannot have chosen a premise to hit a number no one has measured. They are not merely the last leg of the argument; they are the only leg that rests on no chosen premise at all. Which is why the program’s weight has to sit there — and why an essay about a frozen design has to end by pointing away from the PASS wall, toward the bites.
In the spirit of the program’s Statement of Intent: none of this asserts the lattice is the substrate of reality. It asserts something checkable — that a single minimal geometric object, fixed in advance and altered for no result, induces a great deal of known physics on its own, and stakes out novel claims it could lose on. The design has been in plain sight the whole time. The question the bites will answer is whether plain sight was enough.
References
- Paper I — Geometry First — the substrate and the six basis vectors (posited as the causal-adjacency directions), and the no-free-parameter experiment suite; figure source
figures/lattice.drawio. - Claim map — the
PASSwall, and theSTUBnovel predictions where the program stakes its falsifiable, selection-proof claims. - About the program — the audit system and what “proven” means here.