A0 synthetic experiment
Synthetic mathematical demonstration; no H* verdict or organism validation.
Numerical demonstration gates: 18/18 passed.
| Measurement | Result |
|---|---|
| Branch-constrained endpoint vs direct development (max absolute state error) | 1.96e-10 |
| Neutral directions in full equilibrium Jacobian | 3 |
| Maximum adult conductance change between rearing histories | 13.16% |
| Adult activity difference between those histories | 5.67e-11 |
| Implicit gradient vs direct developmental finite differences | 1.77e-08 |
| Transient tangent vs finite differences | 5.47e-08 |
| Equilibrium / transient sensitivity ranks (four parameters) | 3 / 4 |
| Noise seed | Rate relative error | Held-out noiseless RMSE / injected noise sigma |
|---|---|---|
| 7 | 0.14% | 0.213 |
| 19 | 0.67% | 0.097 |
| 41 | 0.02% | 0.121 |
The history-aware endpoint check uses terminal branch information from direct simulation. It does not demonstrate a shortcut for computing that history. Recovery fits know the wiring, body, regulatory gain ratios, initial adult branch and intervention strength. Ranks are local; three synthetic noise seeds do not establish global identifiability or uncertainty calibration.
Full configuration, versions, source hashes and numerical gates: report.json. Sampled trajectories: trajectories.npz.