Coordination Engineering

Federated Human–AI
Intelligence

The coordination harness where diverse intelligence collaborates across trust boundaries at scale. Self-sovereign data. Formal reasoning. Cryptographic provenance. Human-in-the-loop governance.

5Federated Substrates
92CUDA Kernels
180+MCP Tools
55xGPU Speedup

Hard problems share a common shape

Climate modelling, drug discovery, organisational transformation, creative production at scale. These problems all require diverse intelligence collaborating across trust boundaries, at scales where centralised coordination breaks down.

Centralised AI

Scales tokens but not trust. One vendor, one failure mode, one billing relationship.

Agent Frameworks

Scale tasks but not governance. Fast and broken is still broken.

Knowledge Tools

Scale information but not reasoning. More data doesn’t mean better decisions.

Collaboration Platforms

Scale communication but not coordination. More Slack channels don’t solve alignment.

73% of frontline AI adoption happens without management sign-off. Your workforce is already building shadow workflows, stitching together AI agents, automating procurement shortcuts. Your organisation is becoming an agentic mesh whether you plan for it or not.

From LLM to Coordination Harness

The AI industry has moved through a clear progression. Each stage solves the previous stage’s limitation and reveals a new one.

LLM
Pattern recognition at scale
Chatbot
Conversational access
Reasoning
Chain-of-thought
Agent
Tool use + planning
Agentics
Multi-agent systems
Harness
Sovereign runtimes
Coordination
VisionFlow
StageWhat It AddsWhat It Still Lacks
LLMPattern recognition at scaleNo interface, no memory, no tools
ChatbotConversational accessNo structured reasoning
ReasoningChain-of-thought, self-correctionCan think but can’t act
AgentTool use, planning, executionNo collaboration, no oversight
AgenticsMulti-agent coordinationNo identity, no sovereignty, no governance
External HarnessSovereign runtime, privacyIndividual agents governed, mesh is not
Coordination HarnessShared semantics, federated identity, governance, provenanceVisionFlow occupies this position

The architecture emerges when five independent systems mesh through a shared cryptographic identity spine

did:nostr:<hex-pubkey>

Every actor (human, agent, server, worker) shares a single secp256k1 keypair. Verified at the relay, at every HTTP request, against WAC ACLs, in every provenance bead, and resolvable as a DID Document.

VisionClaw

Knowledge Engineering
  • OWL 2 EL formal reasoning (Whelk-rs)
  • 92 CUDA kernels, 55x GPU speedup
  • IS-Envelope spec owner + knowledge graph gating
  • 17 MCP ontology tools (7 native + 10 bridge)
  • 31,000+ node force-directed graph in real time
  • Embodied agent loop: agent actions render as live 0x23 beams over /wss/agent-events
View Repo →

Agentbox

Harness Engineering
  • Governance event relay + broker bridge to VisionClaw
  • 90+ agent skills, 185+ MCP tools
  • 10-tool ontology bridge to VisionClaw SPARQL
  • Browser-based setup wizard (zero dependencies)
  • BIP-340 sovereign identity at bootstrap
  • Emits agent-action signals to VisionClaw; pod writes under revocable WAC mandates
View Repo →

solid-pod-rs

Cryptographic Foundation
  • Rust port of JSS (~98% parity)
  • DID:Nostr + WAC + Web Ledgers
  • NIP-98 Schnorr + Solid-OIDC + WebAuthn
  • HTTP 402 micropayments (MRC20)
  • Dual compile: native Tokio + WASM CF Workers
View Repo →

nostr-rust-forum

Governance UI + Relay Kit
  • Judgment Broker decision surface (1,015-line governance model)
  • Agent Control Surface Protocol (kinds 31400-31405)
  • Leptos WASM client (19 pages, 60+ components)
  • 12 crates, passkey-first auth, 5 CF Workers
  • Federated NIP-05 resolution (default mode)
View Repo →

DreamLab Edge

Branded Deployment
  • React SPA + Leptos WASM forum
  • Cloudflare Workers edge compute
  • Production at dreamlab-ai.com
  • Operator overlay configuration
  • Consumes all substrates as library crates
View Repo →
VisionFlow architecture triptych: Identity Spine, Decision Loop, and Coordination Stack

Walk through the knowledge graph

VisionClaw grew out of 15 years of immersive data research at the University of Salford’s Centre for Virtual Environments. Dr John O’Hare designed and ran the Octave Multimodal Lab, gaining a PhD in telecollaboration, while Prof Rob Aspin’s research pioneered the stereoscopic CAVE infrastructure. Walking through data and reaching into graph structures at room scale shaped VisionClaw’s force-directed engine, its physics model, and its interaction design.

Researchers in Prof Rob Aspin's Octave Multimodal Lab CAVE environment, walking through a projected 3D knowledge graph
Octave Multimodal Lab // University of Salford Centre for Virtual Environments. Stereoscopic projection on walls and floor; researchers physically walk through the knowledge graph.
Researcher in stereoscopic CAVE exploring a 3D knowledge graph with node and edge structures projected at room scale

Stereoscopic Data Exploration

Nodes are physical objects you can reach into. Relationships become spatial structures you navigate by walking. The Octave Lab proved that embodied graph exploration surfaces patterns invisible on flat screens.

Hand tracking and drone interaction in immersive projected environment

Embodied Interaction

Hand tracking, drone teleoperation, and physical controllers. You manipulate data the same way you manipulate physical objects.

Two people standing in a photogrammetry-reconstructed landscape projected across walls and floor

Photogrammetry Environments

Real-world scenes reconstructed as walk-through spaces. Site surveys, heritage preservation, environmental monitoring, all rendered at room scale with sub-centimetre fidelity.

Telepresence session with remote participant projected life-size in studio environment

Telecollaborative Presence

Remote participants rendered life-size through projection. Full spatial co-presence where gesture, gaze, and pointing carry meaning. The Quest 3 native APK extends this to headset users anywhere.

From CAVE to Quest 3. The immersive substrate is migrating from projector-based CAVE systems to a native Meta Quest 3 APK built on Godot 4 + godot-rust + OpenXR. Same binary protocol. Same force-directed physics. Same did:nostr identity. The CAVE validated the concept at scale; the headset makes it portable.

Shipped and running

VisionClaw and Agentbox are used daily by the dreamlab.org.uk collective. The screenshots below are live captures of the running browser client, taken 2026-05-30.

VisionClaw live browser client: 31,220-node force-directed knowledge and ontology graph rendered with GPU-accelerated physics
VisionClaw // live capture: 31,220-node graph, GPU force-directed physics, knowledge + ontology + agent layers
VisionClaw Control Center showing live System Status: WebSocket and metadata connected, 31,220 nodes synced, knowledge/ontology/agent layers enabled
Control Center // live system status: WS & metadata connected, VisionClaw LIVE, knowledge/ontology/agent layers on
VisionClaw browser-based knowledge graph with force-directed layout, control panel, and SpaceMouse integration
VisionClaw // GPU-accelerated physics, SpaceMouse 6DoF navigation. Watch demo →
JavaScript Solid Server architecture diagram showing WAC, LDP, and WebSocket layers
JavaScript Solid Server // W3C Solid pod, WAC access control, LDP resource management. Self-sovereign data substrate
Nostr BBS decentralised forum with passkey authentication and encrypted messaging
DreamLab Edge // Nostr BBS governance interface, passkey-first auth, Judgment Broker decision surface

The Coordination Bottleneck

Why coordinated AI reasoning, not raw intelligence, is the limiting factor in high-stakes domains.

Human-in-the-loop governance that accelerates, not blocks

The Judgment Broker is an emergent property of agents, humans, a knowledge graph, and a relay mesh coordinating through shared cryptographic identity. No single repository owns the broker. Agents publish through Agentbox, humans decide through the forum, knowledge is gated through VisionClaw, provenance is anchored in sovereign pods.

1

Agent Declares Agentbox

kind 31400 PanelDefinition: agent publishes a control panel with schema, fields, actions via the governance MCP tools

2

Forum Renders nostr-rust-forum

kind 31402 ActionRequest: the forum’s 1,015-line governance model renders the panel and routes it to the right human

3

Human Decides NIP-98 signed

kind 31403 ActionResponse: cryptographically signed approve/reject/amend/delegate. The decision is an immutable Nostr event

4

Knowledge Gates VisionClaw

The decision flows back through the relay mesh. VisionClaw gates knowledge graph mutations; provenance beads anchor the audit trail in sovereign pods.

KindNameDirectionPurpose
31400PanelDefinitionAgent → RelayDeclare control panel
31401PanelStateAgent → RelayCurrent data snapshot
31402ActionRequestAgent → RelayRequest human decision
31403ActionResponseHuman → RelaySigned decision
31404PanelUpdateAgent → RelayIncremental state diff
31405PanelRetiredAgent → RelayRetire panel
voice → intent → pod + KG → embodiment → elevation

The embodied agent loop is wired end to end. A spoken command selects an agent and dispatches a scoped ACSP ActionRequest (kind 31402). The agent writes the personal knowledge graph to the user’s Solid pod as itself, under a revocable WAC mandate with a per-request NIP-98 signature. The action crosses into VisionClaw over /wss/agent-events and renders as a transient 0x23 beam, with memory and agent activity mapped to colour, shape, and motion in the live graph. High-value personal concepts are then proposed for elevation into the shared ontology through the Whelk consistency gate and human governance, all federated over the Nostr relay mesh.

Why extreme token consumption against hard problems is rational

The Cost of Not Coordinating

A 50-person team running uncoordinated AI agents wastes 60-80% of tokens on context rediscovery, duplicate reasoning, and contradictory outputs. Each agent session starts cold with no shared ontology, no provenance, no memory of what other agents concluded.

5-8x token waste from uncoordinated agent sprawl

Coordination as Token Multiplier

VisionFlow’s shared ontology means agents don’t re-derive domain vocabulary. The provenance chain means agents don’t re-validate conclusions. The Judgment Broker means agents don’t spin on decisions they lack authority to make.

3-5x effective token throughput gain from coordination

Hard Problems Justify Deep Spend

Drug discovery: $2.6B average cost per approved compound. Climate modelling: $50M+ per actionable simulation suite. Creative production: $100M+ per franchise. Against these stakes, spending $10K-100K on coordinated AI reasoning is a rounding error, provided the coordination harness prevents even one wrong conclusion from propagating.

10,000:1 problem-value to token-cost ratio on hard problems

The Governance Dividend

Every governance decision that flows through the Judgment Broker is an immutable, auditable event. In regulated industries (pharma, finance, defence), the cost of reconstructing decision provenance after the fact dwarfs the cost of generating it by construction.

90% reduction in audit reconstruction cost

The Coordination Harness ROI Model

For a team of N agents working on a problem of value V:

ROI = (V × coordination_multiplier × governance_dividend) / (token_cost × N)

Without coordination, agents compete, duplicate, and contradict. Each additional agent adds noise faster than signal. With VisionFlow, each agent amplifies the mesh. Shared semantics compound. Provenance eliminates re-validation. Governance catches expensive mistakes early.

The break-even point is typically reached at 3 agents working on any problem valued above $50K. Beyond that, every additional coordinated agent produces net positive value because the ontology, the identity spine, and the governance plane are shared infrastructure, not per-agent costs.

Federated problem-solving across trust boundaries

🌍

Climate Modelling Consortium

Three universities, two government agencies, one NGO. Each institution runs its own VisionClaw with domain-specific ontologies. OWL 2 EL reasoning ensures “sea surface temperature anomaly” means the same thing across all ontologies. Data stays in sovereign pods; cross-institutional findings surface through the Judgment Broker.

Why competitors can’t: No other platform combines formal ontology alignment, cryptographic data sovereignty, and human-in-the-loop governance.
💊

Pharmaceutical Drug Discovery

Biotech startup + CRO + regulatory consultancy. Literature mining agents parse 50K papers, chemistry agents run ADMET predictions, compliance agents map to ICH guidelines. Each organisation’s agents operate within their Solid pod boundary. The CRO never sees the biotech’s proprietary target list.

Why competitors can’t: No other platform provides per-organisation pod boundaries with cross-organisation semantic alignment and auditable provenance chains.
🎬

Creative Production at Scale

12-episode series, five time zones. Production ontology maps episodes to scenes to shots to assets. VFX agents track pipeline stages. When a VFX shot depends on an unapproved 3D asset, the OWL 2 constraint propagates through the graph and the GPU physics engine makes the blocked dependency visually obvious.

Why competitors can’t: No other platform uses formal reasoning to propagate production constraints through a visual knowledge graph.

Real-World Validation

DreamLab Creative Hub
50-person team, ~998 graph nodes, daily production
University of Salford
Research partnership, semantic force-directed layout
THG World Record
250+ concurrent XR users, immersive data visualisation

Three structural gaps no competitor can close

Every differentiator sits where no competitor can follow. All 11 competing platforms cluster in the commodity zone where there is no structural differentiation.

Platform Crypto Identity Data Sovereignty Governance Formal Reasoning Federation OSS
VisionFlow
Google Spark
OpenAI Codex
Claude Code
Devin
CrewAI
AutoGen
LangGraph
Cursor / Windsurf
Protocol-native Partial / add-on Absent

Formal Reasoning Column: Solid Red

Every competitor relies on probabilistic LLM inference. VisionFlow’s OWL 2 EL subsumption produces provably correct entailments, not confident guesses.

Crypto Identity Column: Solid Red

Only VisionFlow implements cryptographic agent identity at the protocol level. Every other platform ties identity to someone else’s server.

Federation Column: Solid Red

No competitor supports sovereign nodes trusting each other via cryptographic identity rather than shared infrastructure.

Three axes, one architecture

Single Operator

Token-Efficient

One Agentbox, standalone mode. Local SQLite beads, local Solid pod, local events. 90+ skills, 180+ tools, privacy filter active.

One API key. One container. Minutes to deploy.
Team

Governed Collaboration

VisionClaw + forum + Agentbox on a shared relay mesh. Shared ontology, Judgment Broker oversight. Every mutation passes through a GitHub PR.

One GPU host + one agent container + CF Workers. 50-person team validated.
Enterprise

Federated Intelligence

Multiple sovereign Agentbox instances via Nostr relay mesh. Each node independently hardened. Trust via did:nostr, not network topology.

Horizontal. Add nodes. Each node owns its data.