The earlier boundedness verdict below is incorrect. It was based on the finite snap/signature probe and its fixed alignment/scale search, not on a constructive classification of the inf family. The emitted representative is:
output/duoprisms/duoprism_4_6/duoprism_4_6_inf_family_a2-phi_b3phi-1.vZome
For {4} x {6} with kernel n=(a,b,0,0), the preserved hexagon plane can use
two blue edge axes at 60 degrees; the perpendicular square-height axis is
yellow. The blue/yellow length ratio contributes a sqrt(3) factor, so the
right snap condition is not the tesseract condition
sqrt(a^2+b^2) in Q(phi),
but rather
sqrt(3*(a^2+b^2)) in Q(phi).
This conic has Q(phi) points, for example
a = 2 - phi, b = -1 + 3 phi,
3*(a^2+b^2) = 45 = (3*sqrt(5))^2.
Using the blue hexagon edge frame scaled by sqrt(3*(a^2+b^2)) and the yellow
height axis scaled by 2a/2b gives all duoprism edges on zome axes. Since a
conic over Q(phi) with one point has infinitely many points, {4} x {6} has
an inf family, though it is a different arithmetic branch
from the tesseract support-2 family.
The old rng tables remain useful as diagnostics for that particular probe, but the “bounded at 3 shapes” interpretation is a false negative of the snap search. The old text is retained below only as historical context.
| rng | kernel-count budget | raw hits | dir-deduped | unique shapes (snap-eligible) | emitted | snap-failed |
|---|---|---|---|---|---|---|
| 2 | 332,800 | - | - | 85 | 1 | 84 |
| 3 | 2,883,150 | - | - | 343 | 3 | 340 |
| 4 | 21,523,360 | - | - | 743 | 3 | 740 |
| 5 | 107,179,440 | 14,928 | 2,604 | 1,037 | 3 | 1,034 |
Superseded: duoprism_4_6 was previously thought not to be an infinite
family.
The old finite probe found that its directional-zomeable subset of
(a,b,0,0) kernels grew quickly with rng, but its snap-emittable subset
stalled at 3 shapes. That measured a limitation of the probe’s snap/alignment
branch, not the mathematical family.
All 3 emitted kernels at every rng have form (a, b, 0, 0) — kernel lives
in the {4}-plane, hexagonal plane fully preserved. Geometrically these are
the 4 stacked regular hexagons at heights (±a, ±b).
The 3 emitted kernels are an independent low-norm k₁ plus a Galois pair
k₂ / k₃ (under σ: φ → 1-φ).
duoprism_4_10| rng | emit |
|---|---|
| 3 | 2 |
| 4 | 5 |
duoprism_4_10 is still growing at rng=4 (+3). It is now the
infinite-family candidate worth probing at rng=5 next.
ongoing_work/probes/rng5_4_6_summary.pyongoing_work/prismatic_rng4_progress.log (rng=4),
ongoing_work/duoprism_4_6_rng5_progress.log (rng=5)ongoing_work/prismatic_sweep_log.jsonl,
tagged with "rng": 5 for the rng=5 record.A separate ground-truth (a,b,0,0)-plane snap+signature probe
(ongoing_work/probes/duoprism_4q_snap_sig.py) was run at rng=4..8 to
independently confirm saturation. This uses the robust 5-decimal SHA-256
shape signature (the same one used by tools/dedup_corpus_by_shape.py),
which does not have the fingerprint-collapse defect at high rng.
| rng | n_dirs | aligned | snapped | distinct sigs | new at rng | elapsed |
|---|---|---|---|---|---|---|
| 4 | 3280 | 3280 | 104 | 3 | 3 | 110 s |
| 5 | 7320 | 7320 | 160 | 3 | 0 | 238 s |
| 6 | 14280 | 14280 | 224 | 3 | 0 | 465 s |
| 7 | 25312 | 25312 | 304 | 3 | 0 | 820 s |
| 8 | 41760 | 41760 | 384 | 3 | 0 | 1382 s |
The three SHA-256 hashes (top 16 hex chars) are identical at every rng:
7fcae3177a93bf9b
c3f32d1a4889bc9a
f0de356895a7e439
snap-count grows monotonically with rng (104 → 384) because more directions hit a valid kernel quantum, but every new kernel hash-collides with one of the three existing shapes.
ongoing_work/probe_4q_sigs_q6_ab00_rng{4,5,6,7,8}.jsonSuperseded: the rng probe saturated at 3 probe signatures, but this was not a proof of boundedness. The exact construction above gives a genuine infinite family outside the probe’s successful snap/alignment branch.