How hard to fake — 3D model

Left: one epoch — independent rooms, one leaf each, fused into one Merkle root inside one 30-second window. Right: difficulty in bits — the room stack rising with room count, against the post-quantum reference lines.

Room stack bound
276 bits
≈ 9.6 × 10⁻⁸⁴ at n = 1,000 · Internal estimate
With β (illustrative)
276 bits
β is unmeasured. This point is sensitivity only.
PQ reference
128 · 192 · 256
NIST categories 1 · 3 · 5 (AES-128 / 192 / 256 cost)
Classes per epoch
5,000
5 classes × n leaves, same window, one root

Epoch — 1,000 rooms, one root

Each point is an enrolled room (10 × 10 × 10 shown). Lines are leaves converging through the Merkle tree to the root for this epoch. Drag to rotate. Scroll to zoom. A red room failed a class this epoch and its leaf drops out.

Difficulty — bits vs rooms and β

Surface height is bits. Flat planes are the PQ references at 128, 192, 256 bits. The back edge (β = 1.00) is the room stack we claim. Everything in front of it (β < 1) is how much lower the true value could sit — unmeasured.

Physical Cryptographic Logistical Statistical Temporal Failed leaf (dropped)

How to read this. Security bits mean the attacker wins about once in 2bits tries. The room stack gives about 0.276 bits per room from the measured stacking input α ≈ 0.826, so the bound crosses the 256-bit Category 5 line near 928 rooms and reaches about 276 bits at 1,000 rooms. That line is the claim, labeled Internal estimate: arithmetic under independence across rooms, not a measured 1,000-room fleet. The five-class AND can only push the true difficulty above that line. The β slider shows how much — for illustration only. β has not been measured, so no point with β < 1 is a claim. Short of fooling the measurement, the only known way to mint a genuine leaf is a witness physically present in the enrolled room, in that window, with the enrolled key. The cryptographic class alone is Category 5 post-quantum cryptography (ML-DSA-87 signatures; ML-KEM-1024 key agreement on the record path).