Publications
Research output from the lab — white papers on governance architecture, technical articles, tutorials, and courses.
Public, Chained, Provable — DGV Goes Live
A real Hetzner host, real DNS and OIDC, the T0/T1 revocation and delegation-cascade proofs re-run externally, plus hash-chained offline-verifiable receipts and SLSA provenance
The previous update's honest non-claim — 'nothing is publicly deployed' — no longer holds. dgv-gate now runs on a real host behind real TLS, DNS, and a real Google OIDC issuer. The...
PrimeSwarm vs Google Gemini Enterprise Agent Platform
Google's own docs admit its flagship governance layer uses an LLM as judge, and can make mistakes — plus the documented integration point where a deterministic layer belongs instead
Google's shipped, flagship governance feature — Semantic Governance Policies — reads a proposed action and has an LLM judge it in natural language, with Google's own docs stating p...
PrimeSwarm vs Blue
A different layer, not a competitor — and a small, well-built project whose source turned up a real gap in our own evidence verification
Blue governs which coding-agent CLI a developer is allowed to launch and how it's configured. PrimeSwarm governs what a running agent is allowed to do, and re-checks at the moment ...
The Gate Is a Running Service Now — DGV v0.4.0
Verified identity, signed policies, fail-closed partitions, and signed revocation gossip — measured, not asserted
dgv-gate is now a live enforcement service: T₀/T₁ authority re-checks, JWT/OIDC identity, Ed25519-signed versioned policies, verified human approvals, fail-closed partition handlin...
Sealed, Delegated, Deployable — DGV After v0.4.0
Encrypted agent-to-agent transport, tokens that can only narrow, and a deployment posture check that passed on real Postgres + RS256
dgv-sealed-v1 encrypts A2A payloads end-to-end through a zero-knowledge relay (X25519 ECDH + AEAD, payload hash binding the authorized ciphertext). dgv-delegate-v1 lets a token's g...
What Onboarding Requires — The Customer Checklist
What your organisation must have, decide, and provide before a deployment goes live — plus where we sit relative to the layers you already run
The complete onboarding requirements: the eight questions or a node map, an IdP we sit in front of, the write-and-spend surface, where the gate runs, named humans for YELLOW bindin...
The AI Did It. Who Allowed It?
A plain-language answer for the people who run the business — what ungoverned AI costs, what a gate is, and what we built
When the AI writes the record, sends the email, or spends the money — who allowed it? At most companies the honest answer is nobody knows. This page explains, without jargon, what ...
How DGV-TC-069 Works — and Why It Matters for Regulated AI
The test that catches what audits miss: negative correctness in governance gates
DGV-TC-069 tests negative correctness — the gate must DENY a proposal that violates constraints, even when it looks superficially valid. Most governance systems fail this test beca...
ONLY Lang for Compliance Officers
A governance DSL, not a programming language — inspectable, testable, replayable
ONLY Lang is a domain-specific language for declaring governance constraints. Compliance officers can read it, approve it, and replay it. It is not a PDF policy — it is a script th...
Why Deterministic Receipts Matter to Auditors
The difference between a log and a receipt — and why auditors need the latter
Logs say 'this happened.' Receipts say 'this happened, here is the proof, and here is how you can verify it.' PrimeSwarm produces SHA-256 receipts for every governance decision. Au...
PrimeSwarm vs Microsoft Agent Governance Toolkit
An honest comparison with the toolkit that covers both layers — where AGT is broader, where PrimeSwarm is deeper, where they could be complementary
Microsoft's AGT covers sandboxing, multi-agent identity, SRE, 14 frameworks, and 6 compliance frameworks. PrimeSwarm fills gaps AGT explicitly documents: knowledge provenance, memo...
PSR-001: A Receipt Format for AI Governance Decisions
A framework-agnostic spec for cryptographically verifiable receipts — any governance system can produce them, any auditor can verify them
We published a common receipt format for AI governance decisions. Any system — PrimeSwarm, Microsoft AGT, NVIDIA NeMoClaw, or custom — can produce PSR-001 receipts. An auditor can ...
What Is On Offer
Live, Assessment-only, Lab, Frozen. The quote cannot sell the laboratory.
The PrimeSwarm offer sheet: Guard is live and the portal may accept; Estate and Federation are assessment-only; marketplace, PIR, and Forum are lab; kill switch, per-agent creds, a...
We Sit in Front of the Model You Already Have
Rent it, buy the seats, or run it on localhost. Tokens stay on their bill. We gate the write.
ChatGPT, Claude, Copilot, Azure OpenAI, Bedrock, and the weights on a GPU in the cupboard are cognitive nodes — not our product. PrimeSwarm sits in front of send, file, and record ...
The HITL Contract
Send and file are always YELLOW. Reads may GREEN. Two names on the receipt.
The human-in-the-loop contract the Cognitive API enforces: send and file halt at YELLOW with a proof id until a human binds; reads may GREEN; the receipt names the agent that propo...
The Industry Wants a Write-Path Gate. Most Vendors Sell a PDF.
How PrimeSwarm compares to the organisations and systems the market already shortlists
Buyers, analysts, and regulators are asking for a write-path gate: semantics, live operational state, and provenance. Agent frameworks ship three of twelve of those capabilities. G...
Two Keys, Not a Cliché
SOC 2 and ISO/IEC 42001 as separate questions, mapped to the gate — plus the dictionary that replaces marketing adjectives with evidence
SOC 2 asks whether the controls around the service worked. ISO 42001 asks whether you govern AI across its life. Corporates need both letters. This paper maps AICPA TSC and Annex A...
PIR Lattice Analysis: Detecting Hidden Physiological Stress in Real Biometric Data
How Prime Integer Relations reveal what raw glucose readings cannot
Using real Dexcom CGM, Empatica E4 (HR, EDA, temperature), and food log data from a 14-day continuous monitoring period, we demonstrate five distinct clinical advantages of PIR lat...
Prime Integer Relations: The Mathematical Foundation
How Thue-Morse equilibrium detects instability across markets, neural networks, and knowledge graphs
PIR treats complex systems as arithmetic structures. The Thue-Morse sequence assigns deterministic signs to measure structural balance. Balance Gap and Pattern Area detect regime s...
onlyDB: Graph-Vector Storage for Cognitive AI
Structure, provenance, and constraints — not just embeddings
onlyDB is our cognitive database layer built on HelixDB-style graph-vector storage. Every fact is a node, every relationship a typed edge, every query respects graph invariants. St...
Only OS: Deterministic Runtime Governance via Arithmetic Equilibrium
Validity is not boolean — it is balance
Only OS treats validity as balance, not boolean logic. If a value corrupts or a proposed action breaks equilibrium, the system detects the imbalance and either solves the missing v...
OnlyState: Zero-Knowledge Prime Integer Relations
Social consensus, distilled to a single integer
OnlyState compresses social graphs and group memberships into a single constant-size on-chain state using ZK-PIR and the Prouhet-Tarry-Escott problem. 100,000 members represented a...
TPNN + Quilt: Topological Constraints on Computation
Spatial manifolds and cellular governance — making attacks physically impossible
TPNN enforces topological manifolds on neural network output, making prompt injection physically incapable of forming valid vectors. Quilt governs cellular automata evolution with ...
Governance Without Monitoring: A Free Report on Mathematical Agent Safety
Why constraint beats observation for autonomous systems
Most AI safety frameworks monitor agent behavior and intervene after the fact. We argue this is fundamentally insufficient for systems that execute at machine speed. This free repo...
PIR Evolver Deep Dive: From Thue-Morse to Trading Signals
A technical walkthrough of the engine that powers SignalO
How does a sequence discovered in 1850 detect market regime shifts in 2026? We walk through the PIR Evolver Engine code, step by step — from raw price data to regime classification...
Quilt Explained: Cellular Automata as a Substrate for Safe AGI
Why the simplest model of computation might be the safest
Cellular automata are grids of cells that follow simple local rules and produce complex global behavior. Quilt takes this substrate and asks: can we govern the evolution? The Quilt...
Learning Trails: Why Knowledge Equilibrium Beats Completion Percentages
Using Prime Integer Relations to measure structural soundness of understanding
Traditional L&D platforms track completion — but completion is not comprehension. We introduce Learning Trails, a knowledge graph system that uses Prime Integer Relations (PIR) to ...
From Moscow to Melbourne: A 41-Year Journey to Only Institute
How childhood graphics for my father's papers became a research laboratory
I was drawing graphics for my father's scientific papers before I could read the equations. Forty-one years later, those early mornings in Moscow and the journey to Australia have ...
PrimeSwarm's Governance Stack: Why Mathematical Constraints Beat Monitoring
A 10-layer architecture for autonomous agent safety
No competing framework offers more than 3 of our 10 governance capabilities. We examine why mathematical constraint — not post-hoc monitoring — is the correct paradigm for autonomo...
The Quilt-AGI Bridge: Cellular Automata Meet Mathematical Governance
Five new cell kinds for safe autonomous evolution
We extend Quilt's cellular automata substrate with five new cell kinds — equilibrium, concept, gate, heal, and topology — that provide mathematical guarantees for autonomous cell e...
Getting Started with PrimeSwarm: Your First Governed Agent
A step-by-step guide to deploying your first PrimeSwarm agent with the cognitive API, including JWT authentication, session management, and streaming chat completions....
TPNN: Why Topological Constraints Block Prompt Injection
A technical deep-dive into how the Topological Prime Neural Network (TPNN) uses spatial constraints to make prompt injection attacks mathematically incapable of forming valid outpu...
Prime Integer Relations for Market Regime Detection
How we use Prime Integer Relations (PIR) to detect market regimes and build a momentum rebalancer for the top 16 crypto assets with risk-adjusted position sizing....