Micro-context health scorecard

microcontextdemo -health-scorecard grades the existing quality ledger rather than self-report. The score is deterministic: scale floor, reconciled terminal outcomes, independent verification, and useful-work throughput. Drift remains explicitly baseline-only until a second comparable ledger exists.

The controlled native-CUDA 1,000-context ledger grades A / 100: 1,000 successes, zero errors/refusals, verification rate 1.0, and 0.9564 independently verified nonempty results/s. Artifact: experiments/microcontext/s5-gcp-1000-cuda-health-scorecard-2026-08-07.json.

go run ./cmd/microcontextdemo -health-input experiments/microcontext/s5-gcp-1000-cuda-outcomes-2026-08-07.json -health-scorecard -
go run ./cmd/microcontextdemo -verify-health-scorecard experiments/microcontext/s5-gcp-1000-cuda-health-scorecard-2026-08-07.json