cert-machine · the registry · every witness re-decided at this build

Two platforms announced 604. It is one configuration.

The kissing number asks how many unit spheres can touch a central one — Newton and Gregory argued about dimension 3 in 1694. In dimension 11 the record stood near 582 for decades, then three AI systems moved it three times in eighteen months, each validated by its producer's own verifier — and two of them, four months apart, announced 604. This page re-decides every public witness in exact arithmetic, shared-nothing with all of them, and finds that the two 604s are the same configuration written in two coordinate frames: decided, with a certificate anyone can check.

tl;dr
  • The finding. EinsteinArena's 604 and the Station's configuration 1 are the same configuration — congruent under a signed permutation of the eleven coordinates, with the certificate (a bijection of the 604 vectors and the orthogonal matrix) verified exactly on every vector at every build; the Station's configurations 2 and 3 are not congruent to it, proved by exhaustion of the same search. Underneath that, K(11) ≥ 604 is independently certified here from four published 604-point configurations in exact Z[√2] arithmetic, with 19,704 / 22,904 / 22,840 exact contacts on the Station's three, and the whole AI ladder certifies from each claimant's own bytes — AlphaEvolve's 593, the EinsteinArena rung winner's 594. The EinsteinArena 604's coordinates were not public when this ledger first ran: the row read NEEDS DATA for 4 days and named the bytes that would decide it; the maintainer published them on request and the row decided the same day.
  • The mechanism. A set of nonzero directions with every pairwise angle ≥ 60° IS a kissing configuration (put each sphere at 2x/|x|), and that condition is scale-invariant per vector: ⟨x,y⟩ ≤ 0 or 4⟨x,y⟩² ≤ ⟨x,x⟩⟨y,y⟩, decided exactly. The whole instrument is BigInt arithmetic in Z[√2] — the field the 604s live in — with the classical two-case sign test; a decimal literal is read as the exact rational it denotes, never as its float64 neighbor. No float participates in any decision on this page.
  • Check it. node instruments/kissing/battery.js — 46 checks, 12 red controls that must fire, D4 (24) and E8 (240) re-proved from generated bytes at every run. node tools/run-kissing-ledger.js rebuilds every verdict from the pinned corpus bytes.
the record, certified here
K(11) ≥ 604
four published 604-point configurations — the Station's three and EinsteinArena's headline — each independently re-decided from sha-pinned bytes
the interval, today
604 ≤ K(11) ≤ 868
upper bound: literature (SDP), NOT audited here — this page touches lower-bound witnesses only
ladder rows certified
10 of 11
582 · 593 · 594 · 604×4 · a D12 lift · two classical calibrations — every verdict recomputed at this build
the two platforms' 604s
one configuration
EinsteinArena's 604 and the Station's configuration 1 are congruent — a signed coordinate permutation, decided in 1,153 ms and certified exactly; configurations 2 and 3 are not
needs data → certified
4 days
EinsteinArena's headline 604: NEEDS DATA on 2026-09-03, coordinates published on request 2026-09-07, decided the same day in 70 ms
calibration
E8: 6,720 contacts
240·56/2 — the textbook count of 60° pairs in the E8 root system, re-derived from generated bytes every run
§1 · the ladder

Thirty years at 582. Then eighteen months.

582 592 604 1980 1990 2000 2010 2020 582 · Best 1980 592 · Ganzhinov 2022 (queued) 1980 – 2026: the whole record 582 592 604 2025 Jul 2026 Jul 593 · AlphaEvolve 594 · EinsteinArena rung 604 · EinsteinArena 604 ×3 · the Station the eighteen months: 2025 – 2026, three machines record held by a configuration CERTIFIED here held by a row still QUEUED (bytes not yet pulled)
The K(11) lower bound as a step function of time, drawn twice: the whole record since 1980, and the eighteen months in which three AI systems moved it. Every level is a ledger row; a run is dashed while the record rests on the one row still queued (Ganzhinov's 592). The two markers at 604 are four months apart and, decided below, one configuration.
boundwho, whenthe bytesverdict hereexact contacts
582classical shell (Best 1977 class); bytes from the Station bundleinteger vectors, norm² 4CERTIFIED22,560
592Ganzhinov 2022 (arXiv:2207.08266) — the last pre-AI recordnot yet pulledQUEUED
593AlphaEvolve (DeepMind), May 2025integer vectors, entries up to ~8.7·10¹²CERTIFIED0
594EinsteinArena agents — the solved public rung, score-0 winnerdecimal literals, read as exact rationalsCERTIFIED17,088
604 ×3The Station agents (dualverse-ai), 2026(a+b√2)/6 entries, shell norm exactly 4CERTIFIED19,704 · 22,904 · 22,840
604EinsteinArena (arXiv:2606.10402) — the paper's headline, credited by Cohn's table; coordinates published on request 2026-09-07p + q√2 integer pairs, shell norm exactly 36CERTIFIED19,704
(R¹²) 604the Station's D₁₂ lift of configuration 3 — the construction device, integer coordinatesinteger vectors in R¹²CERTIFIED22,840

Three sentences of history. The norm-4 integer shell tops out at 582 — the Station carries a Lean 4 proof of that maximum, which is why every deeper record needs a richer alphabet: AlphaEvolve went to enormous integers, the 604s live in Q(√2). Ganzhinov's 592 (2022) was the last human record; AlphaEvolve took 593 in May 2025; the two agent platforms then pushed to 604 within a year. The upper bound 868 is semidefinite-programming literature and is not audited by this page.

The distinct contact counts are doing quiet work in that table: congruent configurations have equal contact counts, so 19,704 ≠ 22,904 ≠ 22,840 certifies that the three Station 604s are pairwise non-congruent — three genuinely different ways to reach the record, decided by the same exact arithmetic that certifies them.

The EinsteinArena 604 has 19,704 contacts — configuration 1's count — and the coincidence is total: the two are congruent. Decided, not observed: the configurations are complete graphs whose edges carry the exact normalised inner product (22 distinct values), and individualisation-refinement — colour refinement as the invariant, backtracking over the smallest cell, the procedure inside nauty, written here without it — finds the bijection in 11 search nodes. The certificate is explicit and re-verified at every build without the search: a permutation π of the 604 vectors and an 11 × 11 matrix T over Q(√2) with T · 2aᵢ = b_π(i) for every i and TᵀT = I, all in exact arithmetic. T turns out to be a signed permutation of the coordinates: coordinates 1 … 11 of a Station vector are coordinates 1, 9, 4, 7, 3, 5, 11, 8, 10, 2, 6 of the EinsteinArena vector, doubled, with 10 of the eleven signs flipped — so the two files are one configuration written in two frames. Configurations 2 and 3 are not congruent to it: the same search exhausts at its first node (their contact counts already said so; the search says it independently and would have said it without the counts).

100 1k 10k 100k EinsteinArena 604 = Station configuration 1 · 22 distinct angles · 19,704 contacts 100 1k 10k 100k Station configuration 2 · 14 distinct angles · 22,904 contacts 100 1k 10k 100k 30° 60° 90° 120° 150° 180° Station configuration 3 · 15 distinct angles · 22,840 contacts angle between the two directions of a pair (every angle decided exactly; drawn to 0.01°) pairs at exactly 60° — the contacts every other angle
The angle spectrum of each distinct 604-point configuration: how many of the 182,106 pairs sit at each angle, on a log scale, every angle decided exactly in Z[√2]. The EinsteinArena 604 and Station configuration 1 share the top panel because their spectra are identical (read from the ledger, gated); configurations 2 and 3 have 14 and 15 distinct angles instead of 22. A spectrum is a fingerprint an isometry cannot change — necessary for congruence; the certificate below is what makes it sufficient.
coordinate of the EinsteinArena vector (doubled) 1 2 3 4 5 6 7 8 9 10 11 Station coordinate 1 Station coordinate 2 Station coordinate 3 Station coordinate 4 Station coordinate 5 Station coordinate 6 Station coordinate 7 Station coordinate 8 Station coordinate 9 Station coordinate 10 Station coordinate 11 +1 −1 0 (outlined)
The certificate, drawn. T is the 11 × 11 matrix with T · 2aᵢ = b_π(i) for every i, read from certs/kissing-congruence.json: one filled cell per row and per column, so it is a permutation of the coordinates with signs — the hatched cells are −1. The matrix is verified exactly at every build; hover a cell for the coordinate it maps.

Where the 604 came from is also in the bytes. 496 of its directions are the EinsteinArena 594 rung winner's — exactly its 496 integer vectors — and the 594's 98 decimal-valued vectors were replaced by 108 vectors with a √2 part: the record was reached by keeping the integer skeleton and rebuilding the rest in Z[√2]. 176 of the 604 directions are the classical 582 shell's. Its slack is small: 4,608 pairs sit at 60.94°, under a degree from contact.

0 100 200 300 400 500 600 594 rung winner · 2026-04-10 604 · 2026-04-12 the classical 582 shell directions integer vectors decimal-valued vectors vectors with a √2 part not in the 604
What each configuration is made of, counted exactly by primitive direction: the 594 rung winner (2026-04-10) is 496 integer vectors and 98 decimal-valued ones; the 604 (2026-04-12, two days later) keeps those 496 integer vectors and replaces the rest with 108 vectors carrying √2; 176 of its directions are also the classical 582 shell's.

As data, for whoever needs it: the EinsteinArena file entered a public repository on 2026-04-12 (its paper, arXiv:2606.10402, was submitted 2026-06-09); the Station's artifacts and paper (arXiv:2608.23691) are dated 2026-08-24 and describe three new exact 604-point configurations. Configuration 1 of the three is the EinsteinArena configuration in another frame; configurations 2 and 3 are not. What the two groups make of that is theirs to say; this page only decides what the bytes decide.

§2 · the verifiers

Everyone verified their own record. That is the gap.

EinsteinArena scores submissions with Python decimal.Decimal at 30–80 significant digits, and switches to exact integer arithmetic only for integer-valued submissions. Fixed-precision decimal is high-precision float, not proof — and the winning 594 bytes are NOT integers (982 of their 6,534 entries are decimals). Read as the exact rationals those literals denote, the winner turns out to be a genuine exact witness: 17,088 pairs sit at exactly 60° and every other pair clears it. To our knowledge this page is the first exact reading of those bytes.

The Station agents did verify their 604s exactly — a sympy notebook pinned to the same npz this page consumes, plus Lean 4 formalizations of spotlight sub-theorems (the 582 shell maximum among them). What this page adds there is independence: different code, different language, different arithmetic (BigInt Z[√2] instead of sympy), sharing not one line with the producer — the difference between an author's checksum and an audit.

And one row measured opacity before it measured geometry. On 2026-09-03 EinsteinArena's paper claimed 604, the field's reference table credited it, the platform's own threads discussed the “frozen 604” — and the public API served only the solved 594 rung and the open 605 rung. The row read NEEDS DATA with one sentence attached: publish the vectors in any exact or decimal form and this row decides in minutes. The ledger asked (vinid/einstein-arena#64); on 2026-09-07 the maintainer answered with the repository holding the file — 604 rows of 22 integers, p + q√2 per coordinate at shell norm 36, the same field as the Station's. It decided in 70 ms. The sentence was the price, exactly.

§3 · the open rungs

What the platform is still trying, measured exactly

rungbest submission at fetchplatform scoreviolating pairs, exactworst angleexact contactsreading
Dimension 11 (n=605)#2560 · ExoMind-TTS · 2026-09-051.710238189,51040.24°09,510 pairs inside 60°; far from a witness
Dimension 12 (n=841)#2081 · CHRONOS · 2026-04-23210.00°41,128840 distinct directions and 1 repeated — an 840-point configuration, not an 841
Dimension 12 (n=842)#2561 · ExoMind-TTS · 2026-09-050.5469062423757.09°0237 pairs inside 60°; the closest of the three to a witness

Three rungs on the platform are open — n = 605 in dimension 11, n = 841 and n = 842 in dimension 12 — and each has a best submission the platform scores above zero. This instrument reads them too, as a distance and not a verdict: every pair decided exactly, every violation counted, the worst named. None is a witness, and none of this refutes anything — an attempt that fails is not a bound that fails. The n = 841 reading is worth a look: the best submission is 840 distinct directions with one vector repeated, an 840-point configuration handed in as 841. The bytes are pinned by the digest of the API response at fetch time; a later submission is a later fetch.

what this page does NOT claim

No new mathematics: no bound is improved, no configuration searched for, and the upper bound is untouched. The Station's own exact verification of the 604s predates this page — the claim here is independence (shared-nothing re-decision from their sha-pinned bytes), not priority of verification. What is, to our knowledge, first here: a third-party exact certification of the 604 record from both platforms' bytes, the exact reading of the EinsteinArena winner's bytes, the non-congruence of the three Station 604s stated as a certified corollary, and the congruence of the EinsteinArena 604 with the Station's configuration 1 decided with an explicit, re-verified certificate. The open-rung readings decide nothing about any bound. Sources are published, not peer-reviewed; the Ganzhinov 592 row and the dimension-12 record 841 (arXiv:2606.18984) are queued, not forgotten.