SENTRIK · CHORD · DAF26BZ01-DV007

Prototype Evidence & Scope Statement

Simulated · Unclassified · IRAD prototype data — not operational mission telemetry

Sentrik CHORD is a self-funded IRAD prototype built to demonstrate how runtime assurance, fused human–autonomy timelines, and tamper-evident proof records can support post-mission debriefing and collaborative review. This evaluator console uses simulated, unclassified mission narratives — not deployed operational data, not a fielded DoD program of record, and not a mature production after-action review system.

The purpose of this submission slice is to show technical feasibility and a credible Phase II maturation path, not to claim current military deployment or full BAA closure on every topic.

CHORD requirement mapping

RequirementStatusCurrent capability (prototype)Current limitation
Debriefing / after-action review toolPrototypeFused mission timeline, operator log, client-side AAR modalNot a legacy AAR replacement; limited export
Data playbackPrototypeDecision replay, proof ledger, mission map event playback (simulated)No full-motion video or live telemetry ingest
Digital map displayPrototypeSVG mission map with zones, routes, ACP markers (this build)No GIS/COP integration; simulated coordinates
Fused mission timelineDemonstratedHuman + autonomous events with causal links (demo seed)Causal graph not automated from all sources
Event logsDemonstratedAction feed, operator log, proof recordsNo single unified enterprise log viewer
Vehicle fuel, health, statusPrototypeTelemetryPanel with simulated vitals (fuel, health, comms, sensor, payload)Not tied to live platform telemetry
Multi-vehicle / multi-ACP simulationDemonstratedSimulation module + scripted profilesNot physics-based; scripted events
Planned vs actual executionPrototypeDedicated comparison page with deviation narrativesStatic demo deviations; not automated plan ingest
Autonomous decision-chain explanationDemonstratedPolicy engine constraints, gate, replay, policy auditNo full planner DAG visualization
Human-in-the-loop reviewDemonstratedHITL console with proof updatesNo external C2 notification fabric
What-if replayPrototypeWhat-If Lite dry-run on replay pageDoes not persist counterfactual proofs
AI-enabled AAR summaryPrototypeAfter-Action Report with OpenAI-assisted summarization when OPENAI_API_KEY is configured; deterministic generated summary fallback; policy engine and proof records remain source of truthLLM summarizes structured evidence only; not validated for classified use
Mission policy / rule evaluationDemonstratedRuntime gate + policy engine + auditSix constraint families only (demo scope)
Trust calibration / human-machine teamingPlannedHMI Evaluation Plan page + AAR trust notes (advisor-led Phase II)No formal trust metrics or calibration UI yet

What evaluators can run today

  • Fused Mission Timeline — human + autonomous events (simulated fuse)
  • Digital Mission Map — SVG tactical picture with planned/actual routes
  • Planned vs Actual — structured deviation cards for debrief
  • Human-in-the-Loop Review — escalated autonomy decisions
  • Mission Replay + What-If Lite — dry-run policy re-evaluation
  • Multi-ACP Simulation — scripted event stream through runtime gate
  • Policy Truth Audit & proof ledger — constraint-level traceability
  • After-Action Report — OpenAI-assisted summary of structured evidence (deterministic fallback)

Explicit non-claims

  • Not fielded on operational weapon systems or C2 networks
  • Not certified for classified processing in this demo build
  • Not a replacement for service-owned debrief / AAR tools
  • AI-assisted AAR narrative is available when OpenAI is configured; otherwise the system falls back to a deterministic generated summary
  • The LLM summarizes structured simulated mission evidence only and does not reclassify decisions, override policy outcomes, or perform autonomous intelligence analysis

Phase II maturation plan, indicative

  1. Ingest live or recorded mission plans and COP tracks through API adapters, retaining simulated fallback for demos
  2. Expand trust-calibration metrics tied to HITL outcomes and operator override rates
  3. Mature LLM-assisted AAR drafts into a governed debriefing assistant grounded only in proof-linked events, telemetry, policy evaluations, and human-reviewed mission records
  4. Add GIS map layer or DIS/HLA bridge for multi-ACP exercises beyond the current SVG tactical sketch
  5. Harden for security accreditation path, separate from this unclassified evaluator slice