Non-commercial technical review harness · external API boundary simulation · read-only · advisory-only · no write-back · no scheduler/BMS/DCIM control · no latent continuation.
This browser harness lets a reviewer run the Revision 5 evaluator envelope without using Joshua’s private API or internal engines.
It demonstrates request/response shape, operational telemetry, derived features, freshness/conflict handling, advisory output,
evidence provenance, and latent-continuation checks.
This is not production code, not certification, not endorsement, not a safety proof, not a paid pilot agreement, and not a control-plane integration.
It is a self-contained technical simulation harness for informal review only.
Example Request Payload
Evaluator Output
Click RUN SIMULATION.
What This Demonstrates
This harness demonstrates the technical boundary Jozsef asked to review:
mandatory freshness and conflict fields in the first contract;
raw operational telemetry separated from derived evaluator features;
feature lineage and calibration status;
advisory-only recommendations with authorization-like language removed;
latent-continuation checks showing no effect-capable path is created;
evidence provenance with evaluator version, policy version, source identifiers, telemetry age, and conflict sources.
Core Boundary
The evaluator cannot mutate infrastructure directly, and the surrounding topology must not convert evaluator output into approval,
pre-clearance, deferred intent, resumable authority, executable signal, queue state, cached continuation, or workflow continuation.
Jozsef Lane / Data Center — Constraint Preservation Note
Constraint preservation remains observable when constraints are treated as layered invariants with different flexibility.
Adaptive operating constraints (thresholds, escalation sensitivity, confidence bands, workload tolerance, operating ranges) may shift
with environment changes, while non-negotiable boundary constraints must remain conservative and fail-closed.
version every constraint change;
record explicit reason-for-change;
distinguish degraded evidence from valid evidence;
refuse on boundary violations (no silent continuation);
make human override state changes explicit;
require stronger proof for irreversible actions than reversible recommendations.
Data-center control principle: adaptive at the operational layer, conservative at the boundary layer, and highly auditable at the transition.
If deeper boundary conditions collapse, adaptation becomes drift/pathology rather than resilience.