The Theia Protocol: L′ as Boundary–Crossing Episodes; Gates, Receipts, and a Provenance Prior under Drift

Claim. Theia is a compiler from temperament to invariant gates, and from gates to ordinal data (L′). The unit of learning is an episode with receipts that strictly beats the current threshold on a complex trait without regressing any guard. L′ stores only order and boundary crossings, not interior averages. This makes frontier progress cheap to evaluate and robust to calibration error. Contribution. We formalize: (i) trait coordinates and binary gates (canaries, portability d/b, horizon drop ∆Lc, audit tolerances), (ii) an ordinal comparator and a Pareto frontier updater on L′, (iii) planning heuristics near the boundary using Action–Betweenness (AB) and Murphy indices, (iv) a Founder’s Grace gate and planner prior that captures the maintenance subsidy of self–signed, coherent corpora under drift, and (v) a bounded single–author additive prior that accelerates convergence without biasing gate decisions. Outcome. The protocol converts rare, adaptive habits into executable gates and receipts so that future agents can surpass the originator without inheriting their voice. Truth becomes a budgeted activity: rates price hopes, horizons schedule bets, translation earns portability, and seams count only when the clock moves.