SPY research and synthetic execution runtime

Completed C++20 SPY research and offline execution study.

• Three intraday strategy families and a separate daily-risk experiment use inspected 2022–2023 development data. No candidate advances under the registered screens.
• Causal strategy adapters connect decisions to exact order accounting, synthetic broker lifecycles and durable restart/reconciliation.
• Execution measurements cover 144 synthetic account cases; Release benchmarks contain 120 trial rows per host.

The application code and compiler driver are unchanged from the native acceptance revision 2e566bb. Windows Release/Debug and Linux Release acceptance includes independent demo, execution, benchmark and publication verification. The earlier 29-group regression run applies to revision ee0b72e. Current verification commands are manual; no continuous workflow is configured. PDF page layout was reviewed separately.

Downloads:

• report.pdf: the 17-page technical report with eight evidence figures.
• report.html: the self-contained report and interactive saved-event replay; download and open locally.
• windows-runtime-benchmark.zip: the original 101-file Windows reference run, including raw timing samples, journals, source inventory and host metadata. Extract and run python -B research/verify_runtime_benchmark.py <extracted-runtime_benchmark_0-directory>.
• SHA256SUMS: SHA-256 checksums for the three assets above.

The Windows benchmark summary SHA-256 is 94585a586a04b7568a13a5a50674498edeef550f327f04faa270ad81d6d4b501, matching the distributed evidence manifest. Timings are host-specific observations.

A fresh clone supports the offline demo, synthetic execution measurements, benchmarks and publication reproduction. Historical strategy studies require the original frozen acquisition evidence, which is not distributed. See the versioned README and runbook for build commands and verifier dependencies.