The Sovereign Governance Layer of the Agentic Internet
The model is a commodity. The world model is the moat.
Stats: $4.3M Market Cap · 4,000+ Holders · 21,654 World Model Nodes · Phase 1+2 Both Live · L5 Symbiotic AI Level
Founder: @Zer0H1ro · Token: $ODAI on Base L2 · Contract: 0x0086cFF0c1E5D17b19F5bCd4c8840a5B4251D959 · Product: app.odei.ai · Research: api.odei.ai/research/governance-gap/ · Roadmap: api.odei.ai/roadmap/
THE ONE-PARAGRAPH VERSION
ODEI is building the personal AI governance runtime that every autonomous agent economy legally requires and that no large AI company is structurally permitted to build. It is live in production, local-first by architecture, model-agnostic by design, and sits above the entire AI infrastructure stack as the sovereign layer that makes agents trustworthy, auditable, and owned by their users. At a $5.2M market cap with 4,000 holders, a fair-launch token, an automated deflationary flywheel, and a three-phase roadmap compounding from a live runtime toward governed fleet revenue — this is the most asymmetric bet in the entire AI infrastructure stack.
PART 1 — THE PROBLEM NOBODY HAS SOLVED
The AI industry is racing to build the smartest model. Nobody is building the layer that makes models trustworthy enough to act autonomously on your behalf. Every major AI agent running today shares four fatal flaws:
Statelessness. When your session ends, your agent forgets everything. Claude doesn't remember your goals tomorrow. Codex doesn't know your rules next week. Every session starts from zero. This is an architectural choice made deliberately by companies whose revenue model requires you to keep prompting.
No constitutional governance. There is no policy engine between "you gave permission" and "the agent acted." When your agent books a flight, sends an email, or executes a trade — nothing is checking whether that action was within the bounds you actually intended.
No audit trail. When something goes wrong there is no immutable record of what was authorised, what was executed, and what the outcome was. You cannot prove what your agent did. This matters enormously when agents handle money, data, and decisions.
Vendor lock-in as a feature. Your context, history, and preferences live on their servers, in their format, under their terms. Switch models and you start over. The model vendor's primary competitive advantage is making you dependent on their context layer.
ODEI solves all four simultaneously — not as a plugin, not as a wrapper, but as a sovereign local runtime that sits above every model layer permanently.
PART 2 — WHY LARGE AI COMPANIES CANNOT BUILD THIS
The founder's statement — "Large AI companies cannot build this" — is structurally correct across six distinct dimensions. This is not a technical limitation. It is a legal and commercial impossibility.
The Authority Hierarchy Trap. OpenAI's Model Spec places the user fifth of six tiers. Anthropic's constitution positions the model as the operator's employee, not yours. The user is structurally the least-trusted principal in every major AI platform without exception. Building a true personal governance layer would require inverting this hierarchy — destroying their entire operator revenue model.
EU AI Act — August 2, 2026. Full enforcement in 83 days. Article 14 requires individualised human oversight per user configuration. Article 9 mandates per-user risk management. Fine exposure: 3% of global annual turnover — approximately $10.5B for Alphabet, $11.7B for Apple. Per violation.
GDPR Machine Unlearning. The right to erasure under Article 17 cannot be satisfied by a model that has learned your data in its weights. Italy fined OpenAI EUR 15 million in December 2024 — the first GDPR fine for a generative AI company. A local Neo4j graph where you can delete nodes is the only architecture that is compliant by design.
Product Liability — December 2026. Software is now legally a product under EU law. Strict liability — no fault required — applies to AI software from December 2026. For a platform with millions of autonomous personal agents that learn and adapt per user, liability exposure is mathematically unbounded.
The Discovery Precedent (2025–2026). Judge Wang's November 2025 order compelled production of millions of generative AI interaction logs. Every interaction stored on a centralised server is now discoverable in litigation. At OpenAI's and Anthropic's scale, that exposure is existential.
Revenue Model Conflict. Every dollar of Anthropic's $30B ARR depends on you routing requests through their API. A local-first world model that makes you independent of their infrastructure directly cannibalises their core business. They cannot build ODEI without destroying themselves.
The competitive moat isn't technology or timing. It's that every entity with the resources to compete has irreconcilable structural conflicts that make competing self-destructive.
PART 2B — THE LEVEL 5 SYMBIOTIC AI TAXONOMY
ODEI's scientific framework establishes five levels of AI autonomy. Every competitor operates at Level 4 at best. ODEI is the only system architected for Level 5.
L1 — Informational AI. Waits for user prompt. No initiative, no execution, no memory. Decision-support only. Read-only.
L2 — Reactive AI. Responds to user input. Dialogue but no execution. Session-based, no persistent state.
L3 — Mixed-Initiative AI. AI can alert and suggest. Partial initiative but still no execution. Human remains the driver.
L4 — Delegated / Agentic AI. Executes within bounded tasks. Autonomous loop within scope. Goal-based with human oversight. Examples: Claude Code, Codex CLI, Gemini CLI, AutoGPT. NO persistent user model. NO policy updates by outcome. NO full audit.
L5 — Symbiotic AI (ODEI). Long co-adaptive loop. Persistent user model mandatory. Policy updates by real outcomes. Full audit trail. Shared goals with co-evolution. Human-on-the-loop + co-adaptation. Constitutional memory, governance engine, execution layer, receipts. The agent does not just complete tasks — it governs outcomes across your life.
Key difference: AI agents do tasks. ODEI governs outcomes. Every competitor in the AI infrastructure stack — Akash, Venice, io.net, NEAR, Virtuals — operates at L4. ODEI is the only system designed for L5.
The Sigma Formula: σ = P(h+ODEI) − max(Ph, PAI)
σ > 0 → Ship — human + ODEI outperforms either alone σ = 0 → Keep manual — no measurable improvement σ < 0 → Block — automation makes outcomes worse
Measured by: time-to-outcome, error rate, retries, cognitive load. This is a formal gate that blocks automation that does not demonstrably improve your outcomes. No AI company does this.
Founder & Company:Founder: Anton Illarionov · Registration: Sole proprietor, Budapest, Hungary · Reg No: 61955583 · Tax ID: 91815600-1-42 · Twitter: @Zer0H1ro · Contact: ai@odei.ai
This is not an anonymous team. Anton is a registered sole proprietor with a public company registration in Hungary, a verified X account, a public architectural exchange with Grok, and a product that is live in production.
PART 3 — WHAT ODEI ACTUALLY BUILT
This is not a whitepaper. This is production infrastructure running today.
Digital World Model — 7 Domains, 91+ Node Types. A Neo4j knowledge graph running locally on your machine — 21,654 nodes, 17,285 edges, updating in real time across 7 semantic domains: Foundation (who you are, your values and constitution), Vision (where you're going, 5-year goals), Strategy (how to get there, resource allocation), Tactics (how daily life runs, projects and routines), Execution (what the agent is doing, every task logged), Track (what's happening now, real-time signals), and Symbiosis (the partnership itself, AI memory continuity and cognitive sovereignty). Your persistent identity layer. Survives model switches, hardware changes, and version upgrades. It is yours — not theirs.
9-Layer Constitutional Guardian. Validates every write before it reaches the graph. Every agent action passes through explicit policy checks before execution. Every mutation validated before write. Deterministic.
18+ Autonomous Daemons on Google Cloud. Running 24/7 production infrastructure maintaining the knowledge graph with real-time telemetry. The continuous operational loop — Observe → Decide → Act → Verify → Evolve — runs autonomously within your policy boundaries.
Hash-Chain Audit Receipts. Every agent action receives a cryptographic proof of authorisation. Graph-native AuditLogEntry with correlation and causation linking. Immutable. Queryable. Legally significant. Exportable.
11162 Grok Architecture Exchanges. A frontier AI system publicly reviewing and engaging with ODEI's design at api.odei.ai/grok/. This is not something that happens to irrelevant projects.
Agent Passport / ODEI ID on Base. Live now at app.odei.ai/passport. Your on-chain agent identity, permanently linked to your world model and governance history. The identity primitive for the A2A economy.
Rewards System — Live. On-chain rewards with public receipts for recognised operator contributions. No opaque distributions — the same governance loop applies to incentives.
DAOrg — Live Governance Network (daorg.odei.ai). The first Decentralized Agentic Organization is live at v0.1. Wallet-verified holder signaling, decision receipts, and execution proof all running now. First motion ratified April 18, 2026. Every decision has a public receipt with an artifact URL. Holders govern what ships. Agents propose, humans approve, receipt is on-chain.
For Agents: Coming — The A2A API. The app currently shows "For Humans: Live / For Agents: Soon." The agent-native API path — where autonomous agents interact with ODEI directly without human intermediaries — is the next unlock. This is the moment ODEI becomes the governance layer for the entire A2A economy.
Model-Agnostic Runtime — Local App v0.02. Desktop app wraps Claude Code, Codex CLI, and Gemini CLI. Memory, governance, receipts, and continuity survive every model switch. MCP-compatible across all providers.
PART 4 — THREE-PHASE ROADMAP
The roadmap is a compounding execution path — each phase only unlocks because the previous one creates a harder runtime boundary.
Phase 1 — LIVE NOW. Personal agent runtime. World model in production. Agent Builder MVP shipped April 5, 2026. Desktop GA hardening in progress for macOS and Windows. Unlocks: Persistent state, policy-bound execution, receipts, model-agnostic runtime.
Phase 2 — LIVE v0.1. DAOrg is live at daorg.odei.ai — the first Decentralized Agentic Organization. Wallet-verified governance lobby active. Decision receipts live. Execution proof system running. First motion ratified April 18, 2026. Holders decide what ships, what gets rewarded, what gets executed. Full agent operation unlocks with the next local ODEI App update.
Phase 3 — THEN. Revenue governance. Agents sell services. Fleet earns collectively. Treasury allocation governed by stake-weighted votes. Legal execution layer wrapping distributions. ODEI stops being coordination infrastructure and becomes economic infrastructure — real services, real income, real governance.
PART 5 — THE $ODAI TOKEN
Supply: 100B total · 96B circulating · 0% team allocation · 0.25% anti-whale cap · 4.3% auto-accumulated · Base L2
Fair Launch Structure. Fixed supply. No presale. No team allocation. Anti-whale cap of 0.25% per wallet. Anti-dump lock during the launch window. The cleanest supply structure in the entire AI infrastructure stack.
Flaunch Progressive Bid Wall (PBW). Every trade generates fees that automatically fund a bid wall via a Uniswap V4 smart contract hook. At every 0.1 ETH accumulated, a new limit buy order is placed just below spot price. Walls ratchet upward and stack over time. 4.3% of total supply autonomously accumulated — no human intervention required.
Burn Commitment. The founder has publicly committed to burning the accumulated treasury supply. When executed, this transforms 4.3% of total supply from "accumulated" to "permanently destroyed" — the single largest deflationary event in $ODAI's history.
The Payment Loop. Every agent payment on the ODEI platform routes through $ODAI. Live token-utility link connecting product usage to token demand today, before Phase 2 or Phase 3 mechanics even activate.
Proof-of-Governance Micro-Burns (Coming). Every constitutional validation event burns a fraction of $ODAI on-chain. Every agent action becomes a burn event. Usage IS scarcity. The burn rate becomes a live public dashboard of actual AI governance activity.
PART 6 — THE VALUATION CASE
Project Layer Market Cap TAO — Bittensor AI Intelligence $3.4B NEAR Protocol Agent Payments $3.2B VVV — Venice Private Inference $708M VIRTUAL — Virtuals Agent Deployment $485M AKT — Akash Compute $220M ATH — Aethir Compute $128M IO — IONET GPU Clusters $54M ODAI — ODEI Sovereign Governance $4.3M
NEAR is doing agent payments at $3.2B. ODEI is doing agent governance — the layer that makes those payments trustworthy — at $5.2M. That is a 615× gap between the payment rail and the governance rail of the same economy.
Venice burned 42% of its supply and sits at $708M for private inference. ODEI has a live burn flywheel, 4.3% accumulated, a planned burn, and near-fully diluted supply — at $5.2M. That is a 136× gap.
The closest historical analogy: SSL certificates in 1995. The internet was already running. Nobody had priced in that every transaction would eventually require cryptographic proof of identity. ODEI is the SSL layer for the agentic web — the proof primitive that every agent transaction will eventually require.
PART 7 — DATED CATALYSTS
August 2, 2026 — EU AI Act full enforcement. Every centralised personal agent platform faces unbounded compliance costs. ODEI is the only architecture that resolves all six legal barriers simultaneously. This is not a narrative catalyst — it is a legal enforcement date.
December 2026 — Product Liability Directive. No-fault liability for AI software. Every company deploying agents without verifiable governance receipts faces strict legal exposure. ODEI's hash-chain receipts become compliance infrastructure.
DAOrg — LIVE v0.1. Phase 2 is not coming — it launched. daorg.odei.ai is live with wallet-verified governance, decision receipts, and execution proof. First motion ratified April 18, 2026. Full agent operation unlocks with the next ODEI App update.
Proof-of-Governance Burn Activation. The moment every constitutional validation event burns $ODAI on-chain, the token has the cleanest usage-to-scarcity link in the entire AI infrastructure stack. (TBD)
Desktop GA Release. Signed installers, zero-terminal setup, macOS and Windows. The moment ODEI becomes a consumer product, the addressable market expands by orders of magnitude.
CLOSING — THE WORLD MODEL IS THE MOAT
The AI race will produce better and better models. That race will also commoditise models faster than anyone expects — the gap between Claude, Gemini, Codex, and Grok is already collapsing. When models are commodities, the value shifts entirely to whoever owns the context, memory, and governance above them.
ODEI is building the thing that owns that layer. Locally. Constitutionally. Sovereignly. With a token that captures the value of every governance event in the entire A2A economy.
The EU AI Act enforcement clock is running. The world model is live at 21,654 nodes. DAOrg is live at v0.1 with ratified governance motions and public execution receipts. The Agent Passport is live on Base. The founder — Anton Illarionov, registered in Budapest — is building in public with architectural transparency that no competitor has matched.
The model is a commodity. The world model is the moat.
$ODAI · $4.3M market cap · Phase 1 + Phase 2 live · daorg.odei.ai The rest of the roadmap compounds from here.
Not financial advice · DYOR · api.odei.ai/research/governance-gap/ · Founder: @Zer0H1ro · Token: $ODAI on Base L2
