Skip to content

Publications

Work produced by the godon project.

2026

The Impulse Protocol: Non-Destructive Coupling Discovery in Autonomous Agents

The method behind coupling discovery, with its full validation: the push/pause/hold protocol, CFAR block-step detection, the 21-cell sensitivity sweep across coupling shapes and noise levels, chain topology recovery, and the measured boundaries of the method — reported as an honest boundary map, not a pass/fail bar.

DOI: 10.5281/zenodo.21962956 · Zenodo

From Edges to Curves: Calibration of Autonomous Coupling Characterization in Live Systems

The calibration of the characterization walk: measured response curves per coupling edge, with error bars validated against planted truth across a 32-cell grid — strengths 0.0–0.9, noise to σ=0.10, four response shapes, six wiring topologies, and non-stationary regimes. Every certified point within its own reported uncertainty; zero invented structure across all sentinels; the working envelope and the refusal conditions measured, not asserted.

DOI: 10.5281/zenodo.22127166 · Zenodo

From Curves to Cascades: Composed Coupling Prediction Among Autonomous Agents

The composition of the measured curves: pairwise response curves, measured autonomously and referenced by total level, chained through nonlinear switching elements and summed at converging junctions to predict far-end response — judged by disjoint direct walks, within error bars propagated from each piece. 95 of 97 referee points across chains, a junction, and a live threshold relay; the propagated bars validated as honest metrology; and the composition horizon measured at roughly two nonlinear hops at σ=0.02 — beyond it the far end falls below the detection floor, and composing the pieces is the only remaining door.

DOI: 10.5281/zenodo.22401439 · Zenodo

The Steerwish: Intent Kept on Live Systems

Intent, made keepable: a promise to a live system — keep this reading inside this band — declared by its holder, kept against drift, and refused, not faked, when the measured range says no. Four claims, each carried by receipts from a live measurement campaign: the map steers a multi-line promise through measured couplings, its lines judged live and held, repeated across seed runs with the boundary stated as measured — a limit of map density, not of principle; a target beyond the measured range is declined by name with zero actuation; a declared drift event is survived end to end — the misses stamp, the map re-opens, the reading re-lands; and a three-flight envelope prices the keeping between what luck alone keeps and what perfect hindsight could have kept.

DOI: 10.5281/zenodo.23948170 · Zenodo

See Also