All defense POCs
POC 12DDeterminismBStability guaranteeRScale classificationCCompressionFlagship — full-stack substrateWave 4

JADC2 Reference Compute Substrate

One math, one bound, one classification, one bandwidth budget — across every node, every coalition partner, every cycle. The IEEE 754 of deterministic, bounded, magnitude-aware, compressed defense math.

1
Unique fusion hash across 5 nodes
0
Arbitration rounds for tier consensus
780 bps
Aggregate inter-node bandwidth (budget: 1 Mbps)
8.46e-05
Per-value error bound on inter-node reports

The scenario

Set the picture

Joint All-Domain Command and Control (JADC2) is the decade's largest DoD modernization program. It wants every sensor, every effector, every commander, every coalition partner reading the same operational picture, computing the same decisions, with auditable provenance, across heterogeneous hardware, over disadvantaged comms, with certifiable autonomy. Project Overmatch (Navy), ABMS (Air Force), Project Convergence (Army), and the OSD-level JADC2 program office are all building toward this.

The hard underpinning of JADC2 — the layer no architecture document quite addresses — is numerical interoperability across platforms, allies, and decades. Without it, every other JADC2 capability is built on sand.

What it costs today

JADC2 is an architecture-of-architectures stitched together with translation layers, bandwidth-hungry full-fidelity rebroadcast, multiple competing data fabrics, and per-system certification. Translation overhead in every cross-system data fabric — Link-16, Link-22, NATO STANAGs, CEC, TTNT, Mission Partner Environment — re-projects values into the receiver's numerical convention.

Distributed fusion can't be fielded because nodes don't agree on classifications without consensus rounds. So fusion stays centralized, which means JADC2 has the same single-point-of-failure pathology as the systems it was supposed to replace. Each platform's autonomy and fire-control stack is certified in isolation; coalition operation requires the certification to be re-done from scratch.

Bandwidth-vs-fidelity arguments at every link, with no portable framework for the answer. Forensic ambiguity at every replay. JADC2 program offices are aware of these gaps. The current strategy is to push them down the priority list and address them later. They cannot be addressed later.

What changes with SolvNum

All four capabilities together — the full substrate. This is the flagship pitch.

Dcross-platform determinism

Every node, every coalition partner, every cycle: identical math. The numerical-interoperability gap closes at the compute substrate, not at the data fabric. Translation layers stop translating numbers and start just routing them.

Bper-step excursion limit

Every command, every actuator, every fusion update: provable per-tick bound, certifiable once. The certification artifact for the autonomy stack becomes a single artifact that satisfies every national regime, every service safety board, every coalition partner.

RScale-Aware Classification

Every threat, every track, every sensor reading: shared magnitude classification without consensus protocol. Distributed fusion becomes deployable because the classification is identical-by-construction across nodes. The single-point-of-failure pathology dissolves.

Ccompression with explicit error bound

Every link, every report, every archive: bandwidth budget that fits the contested environment. The bandwidth-vs-fidelity argument has a portable answer — the k parameter and the error bound it implies.

Measurable outcome

What we'll claim — and how it survives review

Each line below maps to a captured number in the demo section. Every number is reproducible from the SolvNum validation suite.

  • Single SHA-256 every node produces independently for every cycle of the mission.
  • Single excursion-limit certification artifact satisfies every node's safety regime.
  • Distributed fusion converges in zero arbitration rounds via shared scale-tier classification.
  • Mission data fits inside a contested-comms link budget with documented per-channel error envelope.
  • One artifact — the SolvNum table file with its published hash — is the long-lived foundation under the entire architecture.

The demo

What was tested. How. What the script printed.

5-node JADC2 vignette across 4 distinct hardware classes: 2 ships (x86_64 servers, Navy CSG combat systems), 1 aircraft (CUDA GPU, airborne command node), 1 ground node (ARM SBC, JTAC / dismounted forward node), 1 coalition partner node (different x86 build, allied national system), and 1 satellite link with 600 ms latency and a 1 Mbps bandwidth budget.

A multi-vector mission cycle runs end-to-end: sense → classify → decide → engage → assess → archive. Every node computes the same fusion math (D — verified by hash), issues commands within bounded per-tick excursion (B — attested per regime), classifies threats consistently via scale field (R — verified by native-field equality of tier across nodes), and reports across the constrained satellite link within budget (C — verified by bandwidth-budget compliance and per-value error bound).

Illustration

In-browser diagram of what the demo proves. The numbers underneath are the captured demo output.

JADC2 vignette — 5 nodes, 4 hardware classes, one substrate

SolvNumsubstrateK=24 · TABLE_BITS=11ship_1x86_64Navy CSGship_2x86_64Navy CSGaircraftCUDAAirborne C2ground_nodeARMDismountedcoalitionx86 partnerAlliedDDeterminismBExcursion limitRBand / precisionCCompression

One canonical fusion hash 26c945819a49… produced independently by every node. Excursion-limited per-tick update (≤ 2.4623×) attested per regime. Band-tier classification identical by construction. Inter-node reports fit the 1 Mbps satellite budget at 780 bps with an 8.46e-05 per-value error bound.

Captured demo output

The numbers the script actually printed.

JADC2 vignette node roster — 4 distinct hardware classes
NodePlatform classRole
ship_1x86_64 serverNavy CSG combat system
ship_2x86_64 serverNavy CSG combat system
aircraftCUDA GPUAirborne command node
ground_nodeARM SBCJTAC / dismounted forward node
coalitionx86 partner buildAllied national system
Per-capability per-node verdict
NodeD fusion hashB max excursionR tier consensusC report size
ship_126c945819a49…1.6245× ✓all 0s ✓19.5 B
ship_226c945819a49…1.6245× ✓all 0s ✓19.5 B
aircraft26c945819a49…1.6245× ✓all 0s ✓19.5 B
ground_node26c945819a49…1.6245× ✓all 0s ✓19.5 B
coalition26c945819a49…1.6245× ✓all 0s ✓19.5 B

Aggregate inter-node bandwidth: 780 bps (1 Hz reports × 5 nodes × 19.5 B). Satellite budget: 1,000,000 bps. Decode hash agreement across nodes: ✓.

Joint All-Domain Command and Control — Numerical-Substrate Attestation

Mission cycle
sense → classify → decide → engage → assess → archive
Nodes
5 (ship_1, ship_2, aircraft, ground_node, coalition)
Hardware classes
4 distinct
D — cross-platform fusion identity
PASS
Canonical fusion hash
26c945819a492db53a4e3faf269c5a67d5f2151b8055c5003233d434016c4390
B — per-step excursion limit
PASS (max 1.6245× ≤ bound 2.4623×)
R — consistent threat classification
PASS (zero arbitration rounds; set union of integer tier tags)
C — bandwidth-budgeted inter-node
PASS (780 bps ≤ 1,000,000 bps; per-value bound 8.46e-05)
SolvNum table version
core.K=24, TABLE_BITS=11

Evidence pointers

Where the claims live in the repo

These are the files a reviewer should run, read, or grep to re-derive every number on this page.

  • SolvNum benchmark suite (quick mode) (and overnight comprehensive sweep)
  • SolvNum cross-platform attestation benchmark
  • SolvNum cross-platform determinism verification (x86, ARM, WASM, CUDA)
  • SolvNum documentation — defense autonomy strategic context

Want to see this in your environment?

Brief us on a program where this POC matters.

ITAR-aware. Air-gapped delivery available. Every claim above traces back to a script in the public repo.

Brief us