BRM-loading · layering pigment…
BRM-loading · layering pigment…
Silicon pass · D5 intensity · captured 2026-05-06
Nineteen acceptance tests, drawn from classical and computational physics, run on an RTX 5070 Laptop in 12.91 seconds. The suite is reproducible; the silicon signature is not transferable. A pass on another machine is a different pass.
Three architectures · byte-identical · proof of capability
Physics acceptance suite
19 passed, 1 skipped in 12.91s
| test | asserts | status |
|---|---|---|
| Molecular dynamics · energy conservation | Lennard-Jones MD total-energy drift < 5% over 10,000 steps | PASS |
| Lattice-Boltzmann CFD · stability | D3Q19 stable, finite, inlet velocity boundary holds | PASS |
| Vortex-lattice aerodynamics · flat plate | High-AR wing Cl matches Helmbold 2pi*a / (1+2/AR) | PASS |
| DC circuit · voltage divider | Nodal analysis recovers analytic node voltages within 1e-3 | PASS |
| Pipe flow · friction factor | Colebrook f matches Moody chart at 3 (Re, e/D) points within 5% | PASS |
| Devulcanisation · unimolecular decay | A->B reaction matches exp(-kt) within 5% | PASS |
| Magnet · solenoid axial field | Biot-Savart B-axis matches m0NI/L within 5% | PASS |
| Beam optics · FODO cell | |trace/2| < 1 both planes; positive Twiss parameters | PASS |
| Cryogenics · Carnot bound | 1 W @ 1.9 K -> ~660 W shaft (Lebrun & Tavian) | PASS |
| Collider decomposition (x2) | Work-breakdown contains every subsystem; every leaf wired to a solver | PASS |
| CNC · CF-150 flange | G-code parses; 20 bolt holes; ISO 4781 cited | PASS |
| Fasteners · M8 torque | M8 / 8.8 VDI 2230 inside 18-30 Nm envelope | PASS |
| Materials · mission-matched | Space-qualified low-density returns Ti-6Al-4V / CFRP; source cited | PASS |
| PLC · pneumatic cycle | Full cycle returns to idle through all states | PASS |
| Tolerance stack (x2) | RSS range tighter than worst-case | PASS |
| Attestation chain · round-trip | Round-trip to substrate audit-chain server | SKIPPED |
What real silicon found
Three of the runs failed on first execution. Each failure was a real bug in code that had passed human review.
The patches are in the public commit history; the silicon signature on the post-patch run is the one stamped above. The materials database refuses entries without a source citation. There are no fabricated mission claims here.
