The Lattice We Build

Hand "what matters is the dimensionless ratio" to a machine and it becomes a linear-algebra problem: the integer kernel of the dimension matrix. We enumerate all 162,931 dimensionless quantities you can build from the fundamental constants, pick out the ones whose value lands on 1, and compare that against a randomized universe. The result is indistinguishable from coincidence. But precisely because we struck out, we got to measure where the wall of O(1) coefficients — the wall dimensional analysis cannot reach past — actually stands. p = 0.67

A lattice as a computational tool works flawlessly ── the breaking of rotational symmetry is held down to O(a²) for group-theoretic reasons, and that has been measured. But the moment you claim "spacetime itself is a lattice," radiative corrections cancel M_P against the loop cutoff and hand you a Lorentz violation 18 orders of magnitude larger than experiment. The culprit is not discreteness; it is regularity. predicted 10⁻² ／ observed 6×10⁻²⁰

If we do not need the spacing, do we need the warping? "Flat spacetime plus a scalar variable speed of light" is checkmated by the Weyl tensor — but what was missing was a single word: direction-dependent. Put a tensor field on a flat stage and the whole of general relativity stands up, with the equivalence principle coming out as a conclusion. And because the lattice never warps, you can solve it by finite differences ── which is exactly what numerical relativity does. g = η + h

After a gravitational wave has passed, LIGO's mirrors do not return to where they started. The symmetry of flat spacetime is not Poincaré but BMS ── infinite-dimensional, and ordinary translations were only the first four rungs, \(\ell=0,1\) of the spherical harmonics. And the geometry of 1962, the scattering amplitudes of 1965, and the gravitational waves of 1974 turn out to be three faces of one and the same structure. Ending with holography for flat spacetime. permanent displacement ≈ 1/3600 of a proton radius

The "four-point function breaks" problem found in the previous bonus episode. The real culprit was codimension ── the Mellin transform integrates the energy away, so the boundary had become two-dimensional. Keep \(\omega\) as \(u\) and the boundary becomes the three-dimensional \(\mathscr{I}\cong\mathbb{R}_u\times S^2\), putting the codimension back to 1. And that boundary carries a geometry with zero speed of light ── Carroll's country, where nothing can go anywhere. BMS = the 3d conformal Carroll group

The opposite episode, in a series that has published nothing but negative results. Soft theorems come in an infinite tower, and that tower closes into an algebra ── \(w_{1+\infty}\). What is more, this algebra had already shown up in area-preserving geometry, self-dual gravity, and twistor theory. Unlike Episode 1's brute force, the target was not chosen afterwards ── this is discovery, not fitting. the structure constant is just ad − bc

The previous episode's "geometry with zero speed of light" does not belong to \(\mathscr{I}\) alone ── because any null hypersurface is necessarily a Carroll manifold. The universe has two such surfaces. Approaching a horizon is literally taking \(c\to0\) (\(dr/dt=c(1-r_s/r)\)), which gives the membrane paradigm — a "figure of speech" for forty years — a geometric pedigree: the Damour equation was the conservation law of a Carrollian fluid. a fit became a derivation

Six times now I have written "there is no independent definition." That is a label, not a diagnosis ── so this episode writes the specification. Of the four roads to defining a theory (action, constructive definition, bootstrap, identification), what is happening on each right now? The action problem is that too many can be written to choose between them, and for the bootstrap the tool's own premises have collapsed. Even so one corner is already closed, and once again the way it closed was "a two-valued target." the real blanks number only two

Two essential unsolved items were left over ── the number of horizon states and de Sitter. But the de Sitter horizon is also a null surface, so Episode 7's tools apply unchanged. If they are the same, the answer is one sentence: a Carrollian theory living on a null surface of area A has \(e^{A/4G}\) states. For our universe, \(S=2.25\times10^{122}\). And it becomes a candidate for the very object Episode 1 stopped at when it wrote that "filling in the \(1/4\) requires something to count." dim H = 10^(9.76×10¹²¹)

How would you ever prove the previous episode's conjecture? Looking it up: the template already exists, and in three dimensions it is finished ── asymptotic symmetry → central charge → Cardy. The decisive point is that this template identifies not a single microstate. So the dead end of Episode 8, "there is no definition of the theory," may not be an obstacle for this particular problem. That Cardy really works we check by counting partitions. ratio 0.99991

Going after the leading candidate of the previous episode, the central charge, brings two pieces of bad news ── it is reported to be \(c=0\), and the central extension is not even a constant. And yet in three dimensions \(c_L=0\) as well. BMS is a semidirect product, so it has two central charges, and the one carrying the entropy was the supertranslation side, \(c_M=3/G\). The question rewrites itself: "does the supertranslation sector of BMS₄ have a \(c_M\) analogue?" no conclusion. the shape of the question changes

If the theory is ultralocal the partition function becomes a product over points, and a Carrollian theory lives on a null surface ── so \(S\propto A\) falls out of the structure automatically. No counting of microstates, no Cardy. All that remains is the coefficient, and that is one number: 2.773 Planck areas per bit. And for the third time we return to Episode 1's "you get as far as \(S\propto A\), and only the \(1/4\) refuses to come out." 1.381×10⁶⁹ bit/m²

I actually ran the procedure table from Episode 11. The positive control failed once, spitting out a bug in my own formula. Corrected, passed, and then on to the real run ── put the central term in as unknowns and solve the Jacobi identity exactly over the rationals, and for BMS₄ you get \(\dim H^2=0\). It has not merely gone unfound; there is no way for it to enter. And the reason came out too: central charges only stand at integer weights, and four dimensions misses by exactly \(1/2\). dim H² = 0

On to the remaining route B. It does not work in four dimensions ── the leading term is not a universal quantity (\(\mathcal{N}=4\) SYM picks up a factor of \(3/4\) when you change the coupling). But pinning down why it fails shows that Episode 12 had set the question up wrong. Add matter and the entanglement grows — but the same matter grows \(1/G\) by exactly as much ── the \(\delta(1/G)\) obtained independently from the heat kernel agrees with \(S_{EE}\) down to the factor of \(A/4\). A/48πε²

I went to dig into the remaining "where does \(G\) come from," and stopped before digging ── that question is not well posed. A quantity with dimensions cannot be predicted; I had verified that myself back in Episode 1. Fix it and you get \(\alpha_G=Gm_p^2/\hbar c=5.9\times10^{-39}\), and this one has a mechanism. From a single measured value, a hierarchy of twenty orders of magnitude drops to the unremarkable number \(1/56.6\). e⁻⁴⁶·³⁹

Escape route ② from Episode 2, followed to the end. At strong coupling the Lorentz violation dies off as a power, and moreover chiral fermions come out of SO(3) spinors (both have two components). Measuring the usable window gives \(10\) TeV \(\le\Lambda_{\rm conf}\le3\times10^9\) GeV. But every one of the authors puts gravity on a fixed background ── and violation in gravity leaks into matter through radiative corrections. 1.27 ≤ ε ≤ 2

Episode 3 wrote that "what was missing was a single word, direction-dependent," and never did the calculation. This time we do ── without warping the coordinates even once, put two functions on a Euclidean lattice and integrate the geodesic, and you get 42.9807 arcsec/century, ratio to GR 1.0000. Drop that one word and you get exactly 1/3. And the finite differences turned out to have one condition you cannot skip. 42.9807″/century

Is the spacing real, or can it be taken to zero ── this is a question observation answers. Entropy is blind (the divergence is absorbed into \(1/G\)), interferometers have already struck out, and time-of-flight needs 183 EeV, which is beyond the GZK cutoff and therefore impossible. What is left is photon decay, and there the formula has no free coefficient. A single photon of about 112 PeV decides it. 112 PeV

Taking stock of eighteen episodes. No new physics came out ── what came out was four closed doors, one calculation that went through, one falsifiable number, and a ledger of the ways I got things wrong. Laid out, the ledger is skewed: the bugs I could not catch myself cluster in the formalism-heavy episodes, while the episode built on four ordinary differential equations produced not one physics bug. Three tools actually helped, and all three were matters of discipline. |H+½| = 2×10⁻¹⁶

To give the conclusion first: it is already written simply ── all of known physics is generated from six terms, or rather from a four-line procedure. Those six terms can be opened one at a time with a button so you can check what is inside. And what is complicated is not the laws but three other places: ① the 25 constants (config), ② the effort of solving (runtime), ③ the translation between scales (the build pipeline). only two terms carry dimensions

If you pin the speed of light times time to a constant, surely space need not warp ── a naive but good question, and it reads three different ways. Read as "c varies," it dies; read as "a choice of coordinates," it genuinely does make things simpler (in conformal time light travels at exactly 45° and the expansion drops out of the electromagnetic field equations); read as "a symmetry," it reaches the core of the theory. And the wall has a name ── the Weyl tensor. FRW can be written. A star cannot

An honest status report on the two developments introduced in Episode 4. The Mellin transform of a plane wave becomes a conformal primary in one line ── but try to build a four-point function and it breaks. Compute the cross ratio and it comes out real to machine precision, and moreover equal to −s/t. The cause was elementary ── 2→2 scattering is planar. Along the way: why soft hair is called a "wig." cross ratio = −s/t, and real

Bonus ② ended with "one function (conformally flat) cannot write a star; the wall is the Weyl tensor," and it pressed the point: "that light bending comes out right is trivial and carries zero information. What distinguishes them is clocks, matter orbits, and tidal forces." This time we go through all of that ── and it went through, just by going from one function to two (② was not wrong; the requirement is simply different).
Perihelion 42.9806″/century (ratio 1.000000), redshift 0.999996, GPS 38.610 μs/day, and the shadow reproducing \(\min_\rho \rho/c(\rho)=\sqrt{27}\,m\) to twelve digits ── with no geometry, no curvature, and not one Christoffel symbol anywhere. Out of the photon ring came \(e^{2\pi}=535.49\), which we never taught it. Frame dragging comes out as well (space is not twisted; the lattice flows).
And then the horizon ── past \(c\to0\) is not the interior but another exterior, and covering the interior requires flowing the lattice. This turned out to be ADM's 3+1, which is to say a spreadsheet. Numerical relativity puts the coordinate light speed \(\alpha\) into the CFL condition and deliberately crushes it to 0 near the hole.
And still the physical speed of light has never moved ── the evidence being that two different lattices produced the same shadow, agreeing to ten digits. The remaining freedom is fixed by hand at five digits, and the wall's name is rotation (Kerr). min ρ/c = √27 m, 12 digits