OSH: Physical Validation of an Autonomous Homeostatic Cognitive Architecture — From CPU-Resident Cognition to Causal Parity, External Timing, and Closed-Loop Execution on ESP32-S3

OSH (Organismo Semántico Homeostático / Homeostatic Semantic Organism) is a proposed autonomous homeostatic cognitive architecture designed to operate as a resident dynamical mechanism rather than as a token-predictive language model. This work presents a cumulative experimental validation spanning CPU-resident benchmarking, cross-target compilation portability, native execution on physical ESP32-S3 silicon, cross-device reproduction, five-level PC–MCU causal parity, 100,000-cycle endurance, independent artifact revalidation, external oscilloscope instrumentation, and a final physical closed-loop experiment. The final closed-loop validation completed 100,000 cycles with INC=5,160, HOLD=93,358, DEC=1,482, ADC_FAIL=0, STATE_FAIL=0, PHYSICAL_100K_GATE=PASS, and FINAL_LATCH=PASS. The internally measured route mean was 29.525 µs, while external oscilloscope evidence includes a representative 23.0 µs route pulse and a separately instrumented 660 ns action/selection interval. The validated OSH route executes locally on microcontroller silicon without requiring cloud inference. OSH is not presented as an LLM, Transformer, prompt wrapper, remote cognitive API, or agent whose cognitive authority is delegated to a foundation model. SHA-256 fingerprints are provided for the validated native modules, final physical harness, firmware BIN/ELF, portability assay, and evidence set, enabling integrity verification while preserving proprietary equations, coefficients, internal geometry, source code, and raw knowledge-store contents. The reported evidence establishes native, repeatable, instrumentable execution of the tested OSH computational mechanism. It does not claim consciousness, biological life, general intelligence, or superiority over LLMs in reasoning or language quality.