# Spectral Basin Allocation MVP

`run_experiment.py` implements a small delayed Kuramoto associative-memory network with shared adjacency and phase delays, phase-gradient prototypes, composite-Laplacian spectral rates, direct ODE integration, and gauge-invariant attractor classification.

Run:

```bash
/home/maxwelhelp/main/bin/python3 run_experiment.py
```

The script writes `results.json` and prints the core measurements. The fixed-seed experiment uses 12 oscillators, two ring distances, 3,000 random initial states for the baseline and shaped networks, and 20 additional random graphs with 700 initial states each.

Observed in the run:

- For the math check, formula rate = `0.5070508`, Jacobian rate = `0.5070508`, and measured log-decay slope = `-0.5083407`.
- Manually reallocating coupling weights from `[0.5, 0.5]` to `[0.9, 0.1]` increased the q=1 basin from `0.0777` to `0.1677`.
- This reallocation did **not** increase q=1's local rate: it changed that rate from `0.3660` to `0.2588`. Thus the experiment supports controllable basin asymmetry, but does not establish the stronger claim that larger local contraction rate universally implies a larger basin.

The basin fraction is estimated empirically; no classifier or nearest-neighbor lookup is used during dynamics or integration.