Local orientation survives the interface, so the frozen core can distinguish which way a discrepancy points.
LIVE THESIS MODEL
Research atlas ↗DIRECTIONALITY · RESONANCE · MEMORY · REFUSAL
UDI Live Lab.
Stress the relational interface. Change what the system can observe, break phase organization, reverse a learned direction, remove a spire, and see whether Candidate-0003 earns a commitment—or refuses.
12 RELATIONAL MEMBERS / THREE-AXIS READOUTCOMMIT
GLOBAL COHERENCE0.896
DIRECTION ALIGNMENT0.984
SUPPORT0.671
HEADING ERROR0.0°
Explanatory reduced model. The live values interpolate relationships found in the thesis; they do not rerun the original research code or predict NASA hardware.
READ THE RESULT
The evidence supports action.
The phase relationship is organized and the distributed state can compress coherently.
Negative and neutral consequences can reduce support for an obsolete action, although the thesis did not establish universal superiority.
Nominal conditions are only the baseline. NASA value begins when the body, environment, or evidence changes.
THESIS LANDMARKS
Data behind the controls.
Every number below is bounded to its declared simulation. The lab uses them as landmarks, not as universal constants.
Recovery after adding locally derived orientation-sensitive relations while keeping the frozen core unchanged.
Candidate-0003 pressure-direction alignment; rank-3 gain, condition number 1.626, and rotation error near numerical precision.
Mean global coherence across 360 hidden runs; all 15 gates passed in the declared simulation topology.
Matched-arrow target response over its inverse. Direction was learned, but autonomous reversal hardened and failed.
Mean active pressure versus positive-only evidence. The overall gate still failed because passive damping did better.
Coherent maintenance was zero in the 928-run transport-replacement matrix: extra phase data destroyed closure when it replaced geometry.
PROPOSED NASA EXPERIMENT
Move from explanatory lab to hardware falsification.
- 01
Freeze the interface, baselines, protected envelope, and refusal threshold before testing.
- 02
Remove a real wheel, joint, sensor, or support relation without announcing its identity.
- 03
Reverse a previously useful direction and measure adaptation without erasing history.
- 04
Compare fixed control, state-of-practice adaptation, UDI, and UDI-without-refusal.
- 05
Report mission value retained, unsafe commitments, calibrated refusals, and recovery time.