Skip to content
02

Prototype evidence

Measured awareness

A chip that models itself can prove it. Here is what the V5 prototype recovers, and the cost of running without it.

Unaware hardware pays the Stability Tax.

Four costs, paid on every chip that cannot feel itself drift.

Embedded tuning hardware

Control ASICs, DACs, heaters, taps, and monitor photodetectors: silicon spent holding the optics in place.

Power and thermal overhead

Continuous tuning power that gets harder as density rises.

Calibration and drift labor

Long bring-up, expert intervention, and re-tuning that never ends.

The evidence gap

Tuned states that can't be proven, transferred, or audited, so every chip and every team starts over.

The V5 prototype runs a causal residual-observer flow with strict train, validation, and stream separation.

Lower streaming error than a static baseline
38.5%
V5 prototype, brassboard emulator
Validation gain from adaptive self-modeling
41.9%
V5 prototype, brassboard emulator
Stream anomalies above a robust threshold
0
V5 prototype, brassboard emulator

Brassboard results on an emulator platform, reported as prototype evidence, not production hardware claims.

Customer pilots are designed to test this transfer to real hardware.

Pilot goals are proposals we validate against measured baselines before claiming them.

[CLIENT TO CONFIRM: pilot partners, timeline, and what a pilot includes.]

The full picture: three measured, three factual.

Lower streaming error than a static baseline
38.5%
V5 prototype, brassboard emulator
Validation gain from adaptive self-modeling
41.9%
V5 prototype, brassboard emulator
Stream anomalies above a robust threshold
0
V5 prototype, brassboard emulator
Authority at install
Zero
Shadow mode first, earned in four stages
Hardware bypassed
None
Connects to existing controllers and test benches
Parent patent
US 12,450,407 B2
Issued October 2025, continuations filed

Brassboard results on an emulator platform, reported as prototype evidence, not production hardware claims.