Each study states its claims and pass bars in advance, and they are never re-scored afterwards. Tests must be able to fail, so we plant known bugs, attacks and broken controls to prove that a check really catches them.
Before we trust them with a sensor. Digital twins →
The people who simulate a design also build it and test it on real hardware. The team pairs AI engineers who know physics with an experimental physicist.
We will run our readout and calibration software on a partner laboratory’s own NV set-up, so that they can check the results themselves.
Gravity-gradiometer aiding in 1,408 whole 24-hour simulated missions across sea floors, depths, sea states and manoeuvres, and a replay of a published shipborne quantum-gravimeter trial.
The use we see is repeat monitoring of the same survey lines for 1.0–1.5 m voids on stable ground.
A design study of a servo-free trapped-ion optical clock using the readout, at model level (TRL 2–3).
Research only. The NV heart-sensor studies and real heart recordings checked against physical bounds are research evidence, not a medical device and not for diagnosis.
| Report | Title | In one line |
|---|---|---|
| Phy | Phy against OpenModelica and F# | 35 planted physics bugs, speed, size and cold start on real sensor firmware, plus four published results reproduced |
| Physics-checked AI | A physics-checked AI for instruments | Four pre-registered studies, 22 of 33 claims passed |
| Security | BlackPhantom Security Validation | 121 attacks and tests on seven components, 0 through; a proven handshake; 1,705 known-attack vectors |
| Paper | Phase Multiplexing for Non-Adaptive Dead-Spot Suppression in GHZ Parity Metrology | The readout’s theory (Zenodo, 2026) doi:10.5281/zenodo.21390434 |
Until the reports are hosted, they are sent on request through [email protected].