Agent for everyone
Personal intelligence with persistent memory, permissions, and access to open or proprietary models.
Strategic assessment · 5 September 2026
The agent and integration layers can work. The ownership experiment deserves a pilot. The token protocol is not ready to carry the social contract.
Executive verdict
ii.inc has identified the right layer of the AI economy and placed too much machinery around it.
The right layer is deployment: the durable relationship among an agent, its memory, its permissions, the people it serves, and the institution that remains accountable when models change. ii.inc calls the institution a Champion—a locally owned private company that would deliver personal agents, embed integration engineers in organizations, and eventually operate humanoid robots as a service.12
That core can work. A public II-Agent repository already exists, and the newer Genii project uses OpenClaw as its execution layer. The integration-engineer model is a recognizable services business. Local ownership can align some of the returns from AI deployment with the people and institutions that create the demand.56
The whole system is not yet credible. The public evidence shows working agent software and an early local-development product core. It does not yet show an operating Champion, a production protocol, a universal-service institution, or a humanoid fleet. The whitepaper also makes a category error at its center: a cryptographic receipt can prove that defined work was performed and scored above a threshold. It cannot prove that the work produced a public benefit, that the benefit was distributed fairly, or that the authority defining “benefit” was legitimate.39
Our answer is yes in pieces, no as one leap. Build the agent, integration, portability, accountability, and local-ownership layers first. Keep the coin, population identity, exclusive franchise, and robotics claims outside the critical path until evidence earns them.
The proposal
The homepage compresses the argument into three moves. Intelligence is becoming cheaper. The value therefore moves from access to deployment and integration. Whoever owns the last mile—the agent, organizational context, customer relationship, and machines acting in the physical world—will control a larger share of what AI creates.1
The Champion is ii.inc's answer. It would be a locally owned intelligence utility formed as a private company with a public-benefit purpose. It would keep the operating record and customer relationship while remaining able to replace II, a model provider, or a cloud provider without losing keys, contracts, or continuity.2
Personal intelligence with persistent memory, permissions, and access to open or proprietary models.
Local technical teams that embed agents and automation inside businesses and public institutions.
Local financing, deployment, maintenance, and accountability for physical AI made by other manufacturers.
The proposal then expands. A Bitcoin-derived ledger would coordinate Champions. “Proof-of-Benefit” receipts would permit the minting of Foundation Coins when approved compute, training, curation, or agent work clears an evaluation threshold. A WebAuthn credential would anchor each person's II-Account. Culture Credits, validator franchises, Guardian agents, an Oracle Council, universal inference quotas, and a staged ownership model would turn the operating company into a new economic and civic institution.34
These are not one bet. They are at least six bets with different evidence, capital needs, legal structures, and failure modes. They should be judged separately.
Component verdicts
II-Agent is public under Apache 2.0 and supports self-hosting, multiple model providers, research, document work, skills, and integrations. That establishes a real software base, not production performance or user sovereignty by itself.5
Genii packages messaging, persistent memory, schedules, app connections, and voice around OpenClaw. Its own documentation calls v0.1 a local-development milestone, depends on E2B and external model services, and excludes production multi-tenancy, high availability, managed credentials, and stable upgrades. The direction is credible; the sovereignty claim remains ahead of the product.67
Organizations buy outcomes, migration, controls, training, and support—not raw model access. A local team that understands both the institution and the agent stack can create durable value. The Champion thesis's specific revenue, valuation, staffing, and switching-cost figures are explicitly modeled, however. They are scenarios, not results.4
A locally rooted public-benefit company could aggregate procurement, preserve continuity, and distribute ownership more broadly than a foreign software vendor. It still needs ordinary governance: directors, fiduciary duties, capital discipline, privacy, sector licensing, procurement legitimacy, competition, exit rights, and a way to replace failed management.
The downstream service opportunity is believable. A robot must be financed, integrated, insured, secured, maintained, and stopped when it behaves incorrectly. ii.inc has not yet published equivalent evidence of an operating humanoid service. The economics will depend on specific machines and jobs, not on a general belief that physical intelligence is coming.
The protocol binds issuance to an approved receipt and score. It does not establish the value, legitimacy, additionality, or distribution of the claimed benefit. Key parameters remain open, and a small validator set, Oracle Council, emergency pause agents, wrapped tokens, cross-chain bridges, and population identity create more governance and attack surface than the useful operating model requires.3
The hard problems
The whitepaper's central receipt contains a work commitment, a benefit class, an evaluation score, a producer identifier, and a timestamp. Consensus checks that the fields exist, the score clears the active threshold, and a validator quorum signs. That can support auditability. It does not close the moral or political question.3
A tutoring system can answer accurately while excluding people who lack devices. A medical model can clear a benchmark while being deployed without lawful authority. A data-curation job can improve an index while violating consent. A public-service agent can satisfy its service-level agreement while making appeal impossible. The ledger would faithfully record the result of the chosen definition. It would not prove that the definition deserved to govern.
ii.inc's newer Champion thesis partly recognizes this limit. It says protocol quality assurance is not full alignment and that the advisory layer must not author the criterion. That is an important correction. It also means “Proof-of-Benefit” should be read as a proposed reward rule, not a solved proof of social value.4
WebAuthn is excellent for strong, scoped public-key authentication. It proves control of a previously registered credential. The W3C specification does not make it a proof that one natural person created only one account.10 The whitepaper says every citizen may create exactly one II-Account without compulsory know-your-customer checks, then uses optional proof of personhood for higher access tiers. The baseline uniqueness and recovery model therefore remains unresolved.
A local monopoly is local. It is not automatically democratic. A Champion that holds identity relationships, agent memory, organizational context, compute contracts, policy mediation, and physical machines could become unusually difficult to contest. Broad equity helps distribute returns, but shares do not guarantee privacy, due process, worker authority, technical portability, or an effective right to leave.
The ownership formula also needs legal and mechanical detail. The homepage says 10% of equity would be allocated to people under 21 at formation and 0.5% issued to each future birth cohort in perpetuity. The later thesis adds local rounds, future public listings, contribution rights, governance participation, and an II warrant. Each device may be defensible. Their interaction—dilution, custody for minors, voting, eligibility, cross-border movement, tax, securities compliance, and control—has not yet been demonstrated.14
Current regulation supports documentation, risk management, oversight, and sector obligations. It does not establish that jurisdictions will award exclusive territorial AI-utility franchises. The European Union's AI Act imposes duties on providers and deployers; the United States' NIST framework organizes continuous governance, mapping, measurement, and management. Neither implies the Champion structure.1112
The coin layer also enters a legal environment that is still moving. The U.S. Securities and Exchange Commission issued crypto-asset guidance in March 2026 and proposed a tailored offering regime in August 2026. A global network combining ordinary equity, token issuance, mining rewards, wrapped assets, and local credits cannot treat compliance as an implementation detail.1314
Fit with RVA Cyber
Agreement: model access is not the final question. Human authority, capability, standing, and claims on productivity gains must survive deployment.
Tension: ii.inc's economic case leans on aggressive displacement forecasts. Our framework rejects inevitability and asks which people, tasks, time horizons, gains, and decision rights actually changed. Shared ownership is useful, but it cannot substitute for a Human Relevance Ledger.15
Agreement: broadly available computational capability and shared capital ownership can help convert productive capacity into freedom.
Tension: an inference quota is capacity, not abundance. The service must also be reliable, affordable in practice, rights-protected, sustainable, resilient, contestable, and available through a human or offline path when the agent fails.16
Agreement: formal measures, audit records, and revision mechanisms are necessary.
Tension: this is the direct objection to Proof-of-Benefit. Better performance on an evaluation outside its empirical and normative warrant is not proof that the purpose has been realized. Benefit definitions need evidence, legitimate authority, contestability, and a route to revise or retire the metric.9
Agreement: keep raw context and keys in a local sovereign plane; share minimized signals with a larger intelligence network; preserve receipts, continuity, and a named institution that can answer when automation fails.
Tension: our cyber-steward design makes authority explicit. It separates observation, inference, approval, execution, independent verification, recovery, and oversight. The Champion architecture needs the same control plane before it mediates public policy, critical services, or physical action.17
There is also a literal technical connection. Genii assigns agent execution to OpenClaw while its own server retains product state, identity, schedules, integration policy, callback verification, and delivery decisions. That separation is sound in principle: the agent runtime should not silently become the business authority. It is also close to the architecture RVA Cyber has been developing in practice.8
Recommendation
RVA Cyber should treat ii.inc as a useful adjacent project and a potential implementation partner, not as a complete institutional blueprint. The fastest honest test is a bounded regional pilot that removes every claim unnecessary to producing value.
Stop or redesign if users cannot export and leave; if the service requires provider lock-in to remain useful; if consequential actions occur without clear approval and recovery; if the integration cost cannot be recovered without speculative capital; if security incidents outpace detection and repair; if productivity gains do not reach users or workers; or if local ownership exists on paper while practical authority remains elsewhere.
ii.inc is right that the last mile matters. Models will change. The institution holding memory, permissions, relationships, physical access, and responsibility may endure.
That is precisely why the institution must earn trust before it acquires monetary policy, civic identity, territorial privilege, and machines that can act in the world. The agent can be built. The integration company can be tested. Ownership can broaden. Every additional layer should arrive only after the layer below it works and remains governable.
Do not begin by minting proof of benefit. Begin by proving who benefits.
Methods and limits
RVA Cyber reviewed ii.inc's homepage, 4 September Champion essay, current whitepaper, current strategic thesis, Champion and product pages, II-Agent repository, and Genii repository and documentation. We compared those materials with current W3C, NIST, European Union, and U.S. Securities and Exchange Commission sources and with four RVA Cyber research programs.
We treated ii.inc's modeled demand, revenue, valuation, employment, licensing, and adoption figures as forecasts unless independently demonstrated. “No public evidence” means none was identified in the reviewed official materials and public repositories through 5 September 2026; it is not a claim that private pilots or unpublished work do not exist.
This is a strategic and technical assessment, not an audit of ii.inc, investment advice, legal advice, or a recommendation to buy equity or tokens. RVA Cyber has no disclosed partnership with ii.inc. Product and regulatory facts can change after publication.
Source ledger