Abstract We present benchmark results for the NM-SRN v2.0 AGI architecture across three canonical instances of the Job-Shop Scheduling Problem (JSSP): JSSP-A (10×5), JSSP-B (50×10), and JSSP-C (100×100). Using structured, evolutionary search guided by neural resonance (RE-S), NM-SRN achieved makespans of 421, 2621, and 12524 respectively — with reductions of 91.4%, 45.6%, and 41.6% from their initial states. The JSSP-C instance, comprising 10,000 operations and a search space exceeding 1035659 states, was solved with full transparency, traceability, and auditability. These results establish a new standard for solvable, explainable solutions to NP-hard scheduling problems, validating the emergence of structured general intelligence beyond statistical LLMs and traditional heuristics.