Engineering Status Report
What has been built and tested, what has been measured, and what still needs independent validation before QUORUM is used in production. This replaces the May 2026 "Institutional Readiness Report", whose performance, recall and certification statements are withdrawn.
01 // Summary
QUORUM's decision path, evidence chain, governance controls and regulatory-reporting workflows are implemented and covered by automated tests. It has not yet been piloted with an institution, independently reviewed, or measured for detection accuracy on labelled data. The next step is a shadow-mode pilot that measures those things.
02 // Verified Today
TypeScript tests passing at the code baseline, plus Rust and Go unit tests.
Components in the software bill of materials. No GPL or AGPL; licence, advisory and secret scans recorded.
FinCEN SARX XML and FINTRAC STR JSON validated against the regulators' published schemas.
AMD SEV-SNP and Intel TDX verification tested against genuine vendor evidence and collateral.
03 // Measured (With Caveats)
An internal, in-process benchmark (100 concurrent workers, no network overhead) measured about 388 risk evaluations per second at about 256 ms average latency.
It has not been re-run on the current build and is not an independent measurement. Per-channel latency percentiles under a defined load will be published with their test conditions before any performance claim is made.
04 // Not Yet Done
05 // Withdrawn From the May 2026 Report
"Certified for high-throughput institutional deployment"; "100% adversarial recall" (based on three test cases); sub-millisecond processing; "P2P sidechain" audit integrity; HSM key management "validated"; the comparison table against legacy systems; and the resilience-simulation percentages. None of these is supported by a reproducible measurement or independent assessment.