Fold Go: Exact Counted Legality, Certified Solves, and Zero-Parameter Competitive Play

Fold Go derives legality, solved values and competitive decisions from the Smithian fold with zero trained policy, zero value network and zero learned weights. The exact engine reproduces the legal-position census 1, 57, 12,675 and 24,318,165 through 4x4 and secures exact empty-board values through 2x2. Applied play includes two independently replayed two-game 9x9 sweeps against GNU Go at depth ceiling three and the preserved 19x19 harness measurement. The current engine binds complete augmented state, typed transposition bounds, universal pass, complete empty-board orbits, complete active fronts and immutable source/opponent receipts. Typed-table and no-cache values agree on a sealed 134-state surface with 24 pass-pass terminals. The calculation ledger now preserves every SFT decision, completed depth and candidate fraction. Benchmark victory remains an explicit objective under Maria Smith's greater-than-50-percent criterion; development games direct construction and do not give agents authority to declare rank, findings, losses or endpoints. Scientific author and publication authority: Maria Smith, Ernos Labs. Open source: Fold Go.