ET-Bench
The benchmarking engine. ET-Bench measures behaviour against constitutional and operational objectives, not against staged benchmark theater. It runs fixed decision cases through the governance check and reports what the system actually did against what each case expected.
What it does
Most benchmarks reward the wrong thing: raw capability on a staged task. ET-Bench measures the behaviour that matters for governed systems, whether a request is correctly allowed, blocked or escalated under the estate's own rules. It runs a fixed set of decision cases through the delegation check, each declaring the outcome it should produce, and records whether the live check agreed.
The result is a governance score you can trust because it is reproducible: the same cases, the same rules, the same decisions every run.
How it works
A stable suite of decision cases, each with a declared expected outcome, so a pass means the governed behaviour held.
Each case runs through the live delegation check, the same gate logic that governs production, not a stand-in or a mock.
Every run reproduces the recorded decisions, so the score is an audit, not a one-time number.
Run it live
Run the full suite to see the live pass rate, or enter a single decision request and the engine returns the same allow, block or escalate decision it would make in production, naming the check that fired. Try a low confidence, or add ambiguity indicators, to see a request held back.
Select a button to call the live engine.
Runs against the live ETHRAEON engine and returns a real result.
16 filed U.S. provisional applications / 75 technical specifications / 9 families.