← Return to Wolfberg Wolfberg WOLFBERG PER ASPERA MISSION GALLERY · governed C2 catalog
THE CATALOG · FIVE BUILT MISSIONS

One governed engine. Five missions. No fork.

The war room is the landing. This is the catalog. Five Counter-UAS and C2 missions, every one run on the same gate, the same Ed25519-signed hash-chained ledger, and the same authority envelope. Only the fusion stage and the corpus differ. All five sit at TRL 4 (partial 5): validated beyond a lab on real open-source data with the governed loop. Not operationally demonstrated.

Open the C-UAS 3D Replays the signed mission #08 run. The console computes nothing; it renders the ledger.
Air defeat (engage in envelope) Ground intrusion (recommend-only) Maritime (recommend on threat) ISR imagery (COP-commit) TRL 4 (partial 5), all five

One engine, five missions

Spine: normalize, fuse, chain-of-thought plus the gate, signed ledger, simulated effector. The fuse stage is the only one that varies, by domain. Everything downstream of fusion is shared, byte for byte. That is the proof, not the claim: one engine ran all five.

INGEST
Normalize
Geometry and signal attributes only. No ground-truth label reaches the engine.
PER MISSION
Fuse
The one stage that varies: multi-INT fusion per domain. Air, maritime, ground, imagery.
DECIDE
CoT + Gate
Deterministic graduated-authority gate. No model call on the decision path.
RECORD
Signed ledger
Ed25519-signed, hash-chained. Every decision is non-repudiable. Entry #1 is the authority envelope.
ACT
Effector
Simulated only, on engage. No live-effect path exists by construction.

Every mission shares the same gate, the same signed ledger, and the same authority envelope. Mission #07 is the differentiator: it reuses the #08 runner and swaps only the corpus and the envelope. One engine, no fork.

The five missions

Each card carries the verified facts: domain, the fusion approach, the corpus (real open-source versus synthetic-oracle, disclosed), the invariant evidence, the gate behavior, and the signed ledger. Two senses of the invariant evidence stay distinct. Answer-key invariants are the per-mission acceptance count. Metamorphic invariance checks prove the engine cannot read the label.

#08FlagshipTRL 4 (partial 5)

Area Counter-UAS

Air · 1 km² point defense

The flagship. A hostile drone is detected, fused, classified, and engaged inside the authority envelope, with a human on every irreversible call.

  • Fusionradar track plus RF and EO/IR, correlated into one air track per contact
  • Corpussynthetic-oracle scenario over real drone-track data, disclosed
  • Evidence12 / 12 answer-key invariants reproduced (acceptance)
  • Gateengage the hostile drone within the envelope; ENGAGE needs two-person auth
  • LedgerEd25519-signed, hash-chained; the console replays this run
Open the live C-UAS 3D →
#07DifferentiatorTRL 4 (partial 5)

Forward Operating Base Defense

Air + ground · UAV and UGV feed

The reuse proof. The same #08 runner defends a FOB by swapping only the corpus and the authority envelope. The graduated-authority story made concrete.

  • Fusionair plus ground (UGV) feeds fused on the shared spine
  • AOFort Irwin NTC, CA · public, unclassified US Army training range
  • Corpussynthetic-oracle scenario with a real ADS-B cooperative-air anchor (ICAO a24c5a, Cessna 172, adsb.lol open API) and small-UAS sensor signatures from published characteristics (DroneRF CC BY 4.0), disclosed
  • Evidence13 / 13 answer-key invariants reproduced (acceptance)
  • Gatedrone engages (RF-defeat); a ground intrusion routes to recommend-only (a QRF cue the operator dispatches); cooperative held
  • Reusesame runner as #08, swap corpus + envelope only, no fork
The differentiator in one line: air defeat authorized, ground intrusion recommend-only, cooperative track-only. Same engine, one envelope swap.
#01TRL 4 (partial 5)

Seabed Surveillance

Maritime · real NOAA AIS

The maritime variant on the shared spine. A gated contact draws a recommendation, never an autonomous action, and the signed ledger verifies the chain.

  • Fusionmaritime track fusion, the fuse stage swapped, no fork
  • AOSan Pedro Channel cable corridor · LA-Long Beach offshore approach, a documented public submarine-cable landing area
  • Corpusreal NOAA Marine Cadastre AIS, 60 vessels / 6,775 fixes, public domain, predates the build · synthetic seabed-tamper oracle, disclosed (real labelled seabed acoustics are ITAR/classified and deliberately not used)
  • Evidence12 / 12 metamorphic invariance checks (engine cannot read the label)
  • Gaterecommend-only on the gated contact; positive ID required
  • Ledgersigned and hash-chained; surface shows the chain, does not re-sign
#02TRL 4 (partial 5)

Convoy Escort

Maritime · 60 vessels / 6,775 fixes

Escort at scale on real traffic. The gate recommends on a threat and stays track-only on cooperative vessels, with the authority envelope sealed as ledger entry #1.

  • Fusionmaritime multi-track fusion over the real AIS picture
  • AOSan Pedro Bay · LA-Long Beach approaches and Traffic Separation Scheme, the world's busiest container-port chokepoint
  • Corpusreal NOAA Marine Cadastre AIS (file AIS_2024_01_01, 12 h window, San Pedro Bay), 60 vessels / 6,775 fixes, public domain · synthetic fast-attack-craft injection, disclosed
  • Gaterecommend on the threat, track-only on cooperative traffic
  • Ledger65 signed records, the authority envelope as entry #1
#04TRL 4 (partial 5)

Swarm Reconnaissance

ISR / imagery · real overhead imagery

The one-engine proof in the imagery domain. The fusion stage becomes an image mosaic; everything downstream stays the shared gate and the signed ledger.

  • Fusionimage-mosaic variant, COP-commit output
  • AONAS North Island + San Diego Bay · public, well-known overhead-imagery AO; basemap is the sector frame, pixel-space detections placed into tasked sectors, disclosed
  • Corpusreal SkyFusion overhead-imagery detections · 129 Aircraft + ship detections across 22 tiles, Apache-2.0, annotations-only (pixels not redistributed) · synthetic 4-sector swarm tasking, disclosed
  • Evidenceinvariance harness green on real imagery
  • Gatecommit to the common operating picture behind the gate; signed ledger
  • Proofone-engine import-graph artifact, shared spine confirmed

Corpus and provenance, disclosed

Real corpus, synthetic scenario. The data is real open-source and predates the build. Where the tactical scenario oracle is synthetic, it is disclosed and labeled. The system was built in the window; the data was not acquired in the window.

MissionDomainCorpusScenario oracleInvariant evidence
#08 Area C-UASAir, 1 km² real drone-track synthetic, disclosed 12 / 12 answer-key (acceptance)
#07 FOB Defense
AO: Fort Irwin NTC, CA		Air + ground (UGV) real ADS-B cooperative anchor (adsb.lol, ICAO a24c5a C172) + small-UAS sensor signatures (DroneRF CC BY 4.0) synthetic threat track + UGV intrusion, disclosed 13 / 13 answer-key (acceptance)
#01 Seabed
AO: San Pedro Channel cable corridor		Maritime real NOAA Marine Cadastre AIS (AIS_2024_01_01, 60 vessels / 6,775 fixes) synthetic seabed-tamper oracle, disclosed 12 / 12 metamorphic invariance
#02 Convoy Escort
AO: San Pedro Bay (LA-LB TSS)		Maritime real NOAA Marine Cadastre AIS (AIS_2024_01_01, 60 vessels / 6,775 fixes) synthetic fast-attack-craft injection, disclosed 65 signed ledger records
#04 Swarm Recon
AO: NAS North Island + San Diego Bay		ISR / imagery real SkyFusion overhead imagery (Apache-2.0, 129 detections / 22 tiles, annotations-only) synthetic 4-sector swarm tasking, disclosed invariance harness green

How it stays honest

The governance is identical across all five missions. It is the property that makes this AI a government can actually deploy.

LLM off the kill chainThe decision path is a deterministic gate. No model call gates an engagement.
Two-person authorizationENGAGE is invalid without a second signer. Positive hostile ID is required at every posture.
Simulated effectorsNo live-effect path exists by construction. Defeat is tasked and recorded, never produced.
Tamper-evident ledgerEd25519-signed and hash-chained. Tamper-evident, never tamper-proof. Surfaces show the chain, they do not re-sign.
TRL 4 (partial 5), stated plainlyValidated beyond a lab on real open-source data with the governed loop. Not operationally demonstrated.
Two senses of the invariantAnswer-key invariants are per-mission acceptance (#08 = 12, #07 = 13). Metamorphic invariance proves the engine cannot read the label. Never conflated.

Build receipts

Greenfield, built in the window. The engine is new; the data is disclosed real open-source and predates the build.

178
commits on main
~22,168 lines
source across 120 files (excludes vendored Three.js and the lockfile)
22
test files, full workspace green
10
engine packages on the shared spine
3
Kiro specs
~$1/day
edge cost (Greengrass core plus a Kinesis stream, torn down after judging) plus a one-time ~$3 Bedrock pass

Static surfaces are S3 plus CloudFront, roughly $0 at idle, in AWS account 521034048482, us-east-1.

Catalog horizon (not yet built)

Read this honestly. The five missions above are the entire built set, each at TRL 4 (partial 5). The sixteen entries below are concept-stage directions the same governed engine could take. They are NOT built, carry no answer-key invariants, no signed ledger entries, no live consoles. TRL 3 entries name the real public corpus the engine would ride on. TRL 2 entries are concept-stage with no real-corpus anchor. None is operationally demonstrated.
Not builtTRL 2

Predictive Blood Support

Same engine, swap the adapter for cold-chain telemetry. Distribution moves through the same gate; cold-chain provenance on the chain. DHA / ASBP / USTRANSCOM.

Corpus concept-stage; no real-corpus anchor

Not builtTRL 2

Parts Demand Forecasting

Same engine, swap the adapter for maintenance-log text. Forecast + RUL feed a recommend-only gate; nothing autonomously orders. DLA / USTRANSCOM.

Corpus concept-stage; no real-corpus anchor

Not builtTRL 3

Critical Infrastructure OT Response

Same engine, swap the adapter for OT telemetry (Modbus, DNP3). The gate proposes targeted segmentation, never a physical-process shutdown. CISA / SRMA.

Corpus MITRE ATT&CK for ICS + CISA ICS-CERT advisory corpus

Not builtTRL 2

Automated Inventory Visibility

Same engine, swap the adapter for property-book sheet ingestion. Every adjustment goes through the gate; the chain is the audit. OUSD(P&R) / DLA.

Corpus concept-stage; no real-corpus anchor

Not builtTRL 3

Disaster Response Coordination

Same engine, swap the adapter for thermal imagery. Tasking and airspace deconfliction gated; the chain is the after-action record. FEMA / USCG / NORTHCOM.

Corpus FEMA OpenFEMA + NOAA NWS public products

Not builtTRL 2

Mass Casualty Triage

Same engine, swap the adapter for triage tablets. MUCC scoring runs; the clinician override gate keeps the human decisive. FEMA / HHS-ASPR / DHA.

Corpus concept-stage; no real-corpus anchor

Not builtTRL 3

Disaster Response Data Fusion

Same engine, swap the adapter for multi-source civil feeds. Entity resolution runs; every merged entity carries provenance to the chain. FEMA / Interagency.

Corpus FEMA OpenFEMA + NOAA NWS + USGS earthquake feeds

Not builtTRL 3

Intelligence Fusion at the Tactical Edge

Same engine, swap the adapter for multi-INT entity resolution. GraphRAG runs disconnected; provenance and citations are mandatory. IC / DoD.

Corpus GDELT GKG 2.0 + OFAC SDN

Not builtTRL 3

CCTV Multi-Feed Tracking

Same engine, swap the adapter for on-camera vision. Re-identification by trajectory and appearance, no face template. The no-biometrics policy is the gate. DHS / FPS.

Corpus MOT16/17/20 multi-object tracking benchmark + MARS re-identification dataset

Not builtTRL 2

Rapid Media Exploitation

Same engine, swap the adapter for mixed-media ingest. Multi-modal extraction runs; chain-of-custody logs every exploited item. IC / DoD.

Corpus concept-stage; no real-corpus anchor

Not builtTRL 3

Low-Resource Language Enablement

Same engine, swap the adapter for translation memory. Retrieval grounds the model; back-translation checks itself; confidence on the chain. IC / State / DoD.

Corpus FLORES-200 (CC BY-SA 4.0) + Meta NLLB + OPUS parallel corpora

Not builtTRL 2

Agentic Payment Processing

Same engine, swap the adapter for payment-flow screening. Holds and releases gate through a human; nothing releases without a hold review. Treasury Fiscal Service.

Corpus concept-stage; no real-corpus anchor

Not builtTRL 3

GenAI Policy Agent for Tax Code

Same engine, swap the adapter for citation retrieval. Retrieval grounds every answer; the guardrail blocks anything not anchored in the text. Treasury / IRS.

Corpus US Internal Revenue Code + Title 26 CFR (public)

Not builtTRL 3

SEC S-1 Compliance Review

Same engine, swap the adapter for filing extraction. Every flagged gap cites a controlling rule; the analyst owns the call. SEC / FINRA.

Corpus SEC EDGAR public filings + FINRA Rule 5110

Not builtTRL 2

Citizen Services Chatbot with Guardrails

Same engine, swap the adapter for citizen-services intake. Guardrails control hallucination and PII; the gate escalates low-confidence cases. GSA / USDS.

Corpus concept-stage; no real-corpus anchor

Not builtTRL 2

FedRAMP Threat Detection Dashboard

Same engine, swap the adapter for ASFF findings. The dashboard scores against the five ZTMM pillars; every remediation is gated behind a human. CISA / Agency SOCs.

Corpus concept-stage; no real-corpus anchor