nx_evo_synth.nx

nx_evo_synth.nx -- EVOLUTIONARY PROGRAM SYNTHESIS (X-AUT-006L): SELF-BUILDING CODE. The leap past the emitter ladder. Rungs 006g..006k made the team able to EMIT a program from a hand-written DATA spec. This organ makes the team DISCOVER a program by EVOLUTIONARY SEARCH -- it is given only a TARGET (a desired output for a start input), NOT the answer, and SEARCHES op-list program-space until it finds code that hits the target, then MATERIALIZES that discovered code as a real sovereign compiled organ. Genotype (op-list genome) -> phenotype (its computed result) -> fitness (distance to target) -> selection+mutation -> next generation: this is genetic programming / an (1+lambda) evolution strategy, the substrate of "evolutionary computation, self-building code". nx_evo_synth <champion_basename> <start> <target> <seed> - deterministic (seeded LCG) so a run is reproducible/gateable. - genome = EVO_L ops (opcode add/mul/sub + operand 1..9); pop = EVO_P; gens = EVO_G; (1+lambda): elite best + lambda mutants of best each gen. - prints FOUND/NEAR gen=<g> dist=<d> result=<r> genome=<ops>. - WRITES runtime/_hdl_build/<champion_basename>.nx = the DISCOVERED program (an op-list organ) so the gate can COMPILE+RUN it and confirm it really produces the target -- the unfakeable proof the evolved code is real. no-false-green: different targets -> different DISCOVERED genomes, each verified by compiling+running the materialized champion to the target; the champion is NOT hand-written and (for non-trivial targets) NOT found at generation 0 -- the search must climb. Sovereign, no gcc/.sh. HONEST SCOPE: linear op-list genomes over one i64 (the 006g phenotype), distance fitness, mutation-only ES; NOT yet crossover / branch+loop genomes / multi-example fitness -- but it is GENUINE program synthesis by evolution, the seed of the research arc. license_tier: ORIGINAL