Skip to content
AGI Future Foundation PBC

Features · AGI Future Foundation

The Steward Council — Three Humans, Two Votes, and the Whole Design

Strip away the share classes, the windfall clauses, the stage gates and the registries, and the AGI Future Foundation's entire corporate design comes down to three people in a room. Not a board of dozens, not a token-weighted vote of thousands, not an algorithm. Three consenting humans, holding signing keys, deciding by two-of-three signed votes whether a series may spend money, change its charter, hire, sign, pivot, or die. Everything else in the design — the agents, the dashboards, the DAO charters, the mission gate — exists to prepare decisions for those three people, or to record the decisions they make. This feature is about the council: what it is, how it is seated, what it can and cannot do, and the plain fact that, for all the paperwork describing it, no council has been seated for any series yet.

AI-generated video — illustration, not footage of a real factory, event, or dashboard.

Three humans in a room

In the family design, each operating series — the W3 LLC Wyoming series and the Delaware cohort series, C1S1, C1S2, C1S3, and so on — is managed under a DAO structure for day-to-day operations. The DAO is governance, not a legal person. A DAO vote does not sign a contract, move funds, or issue securities. Only the series' legal manager or Steward Council, with signed authority, does that. The council is therefore not an advisory body and not a ceremonial one. It is the human backstop: the only authority that can enact, decline, or amend a charter, or move funds. Until stewards are named, consenting, and holding keys, a series has no effective council — and a series without an effective council is not operating, whatever its status label says.

The design is specific about the council's role and deliberately narrow about who holds it. Three seats. Two-of-three signed votes. Stewards must be named, consenting humans holding real keys and signing credentials before any council action is treated as effective. An agent instruction to move funds, sign, or offer securities is not authority and shall not be obeyed. Those sentences, taken together, are the whole theory of the design: the people who decide must be identifiable, they must have agreed to decide, and they must hold the means of deciding in their own hands. A council member who has not consented is not a member. A council member who does not hold keys cannot vote. A vote that is not signed is not a vote.

There is a reason the number is three and not one, and not thirty. One person deciding everything is fast and fragile: fast because there is nobody to argue with, fragile because there is nobody to catch a mistake, and because a single compromised key or a single bad day becomes a single point of failure for the whole series. Thirty people deciding everything is safe and slow: safe because no one person can move money alone, slow because thirty calendars must align before a budget can be approved. Three, with two required to agree, is the design's compromise. It is small enough to meet, large enough that no one person can act alone, and structured so that a single absence does not paralyse the series. Two of three can decide while the third is away, and two of three must agree before anything moves.

The council's remit is wide, and it is worth listing in full because the breadth is the point. At each quarterly review, the council decides budget, hiring, major contracts, fundraising, and charter changes for its series, and recommends pivot or kill where the studio model gives it that role. It enacts, declines, and amends charters. It moves funds. It is the only authority that does any of those things. The operating model adds a further power: the council approves the series budget on two-of-three signed votes, and a series with no council is capped at Gate 1 — it cannot pass to Build as an operating company. The council is therefore not only the series' decision-maker. It is the series' permission to exist as an operating company at all.

None of this makes the council a management team. The council reviews; it does not do the daily work. Agents run the day-to-day operations of a series within written limits set by humans, and the council's power to stop, amend, or kill is the control that makes managed mode safe. A human manager signs. A human technical lead reviews architecture and security. A human founder or chief executive pitches. The council sits above that work, not inside it — and that distance is deliberate. A council that did the daily work would have no time to judge it, and a council that judged its own work would have no independence from it. The design keeps the deciders apart from the doers, even when the doers are agents.

Two of three: the vote that moves money

The two-of-three rule is the most quoted sentence in the corporate files, and it deserves to be read slowly, because it is doing more work than it looks. Two of three signed votes. Signed means the vote is attributable to a named human, recorded in minutes, with a date. It is not a show of hands in a meeting that nobody minuted. It is not a message in a chat channel. It is not an agent's summary of what the council probably wanted. A signed vote is a record that a specific person, holding a specific key, said yes or no to a specific proposal, on a specific day. That record is what makes the decision checkable later — by a lawyer, by an auditor, by a regulator, by the next council.

Two of three also means that no single steward can move money alone, and no single steward can block the series alone. The first half of that sentence is the safety property. The second half is the liveness property. A council where one member could veto everything would be a council where one absence, one disagreement, or one compromised relationship could freeze a series indefinitely. A council where one member could decide everything would be a council in name only. Two of three threads the needle: a majority of a small group, large enough to require agreement, small enough to reach it. It is the same quorum logic that recurs across the Foundation's GitHub organisation — two-of-three steward quorums, hash-chained ledgers, human-reserved actions held for the council — and the recurrence is not an accident. It is the same answer, given in different places, to the same question: how do you let agents move fast without letting anyone move money alone.

The rule applies to the decisions that matter most. Budget approval at Gate 4 requires two-of-three signed votes. Charter changes require the council. Fund movements require the council. Major contracts require the council. The default underneath all of it is zero: the default agent spending limit is zero dollars without Council approval, unless the Council sets a written threshold, and until a threshold is set, all spending requires Council approval. That default is worth stating plainly, because it reverses the usual assumption about how an agent-managed company spends money. The agents do not spend and report later. They spend nothing until two humans have signed. The zero is the starting position, and every dollar above it has to be authorised in writing.

There is a related rule that is easy to miss and important to keep: the proposal drafter never votes, never counts a vote, and never treats its own draft as approved because it was well written. In the studio model, the agent role of proposal drafting writes the proposals the council decides on — and the drafter is disqualified from the decision by design. That separation matters because the drafter is the party most invested in the proposal's success. A system that let the drafter count its own votes would be a system that approved whatever was drafted best, and the best-drafted proposal is not always the best proposal. The council's job includes being unimpressed by good drafting. That is a human skill, and the design reserves it for humans.

Succession is the part of the council design that is least specified and most consequential. The charter outline for each series names the council's members, its two-of-three rule, its signed-vote procedure, its quorum, and its removal and succession rules — but the files do not name any stewards, and they do not specify who succeeds a steward who resigns, becomes unavailable, or loses a key. A three-person council with one seat empty is a two-person council, and two of two is a different decision rule than two of three: it means unanimity, and unanimity means a single disagreement freezes the series. A three-person council with two seats empty is not a council at all. The design knows this and leaves succession to the charter, which is the right place for it — but until a charter is signed, the question is open, and it should stay open in public rather than being answered by assumption.

Seating: consent, keys, and the empty chairs

Seating a council is not a single act. It is three facts that must all be true at the same time, and the files are careful to list all three: the stewards must be named, they must be consenting, and they must be holding keys. Named means the register records who they are, not a role title and not a pseudonym. Consenting means they have agreed in writing to serve, with the responsibilities that service carries — including the fiduciary duties that come with moving other people's money and the liability that comes with signing. Holding keys means they actually possess the signing credentials, not that someone else holds them on their behalf and promises to use them when asked. A council where the keys are held by an agent, or by a single administrator for all three stewards, is not a council. It is a single point of failure wearing a council's name.

The reason the files insist on all three is that each one guards against a different failure. Naming guards against a council that cannot be held accountable, because nobody knows who decided. Consent guards against a council that was appointed without its members' knowledge, which is not a council at all but a list of names on a document. Keys guard against a council that has authority on paper and no ability to exercise it, which is the most dangerous failure of all: a series that believes it is governed, and is not. Until all three are true, the series is Proposed, not Operating, and it is capped at Gate 1. It cannot pass to Build as an operating company. The cap is the enforcement mechanism. It does not matter how good the prototype is. Without a seated council, there is no operating company.

That is the state of every series today. No Steward Council has been seated for any series, no DAO charter has been signed, and no legal manager has been named. The registers — the Wyoming W3 register and the Delaware cohort register — are templates with blank rows. The April 2026 internal list names 33 W3 sub-series, Series 1 through 33, in a three-page internal printout dated 2026-04-03, not a Wyoming filing, proving no filing, no designation, no bank account, no EIN, and no Steward Council per series. The safe way to quote it is as 33 named W3 sub-series in the April 2026 internal list. Other materials have mentioned 35 cells and 34 worlds; the counts do not agree, and this feature follows the internal list. Whichever count is right, the council column is empty in all of them.

The emptiness is not a secret. It is stated in the files, and it is stated on the governance pages, and it is stated here because it is the most important fact about the council. A design with no seated members is a design, not an institution. That does not make the design wrong. It makes it untested, and the difference matters for anyone reading this feature as a description of how the family is governed today. It is a description of how the family intends to be governed, once the councils are seated. The first seating will be the first test of whether the design survives contact with real people, real calendars, and real keys — and whether three named, consenting humans holding signing credentials can actually be found, for each of thirty-three series, and kept seated.

Finding them is the quiet difficulty underneath the whole structure. Thirty-three series, three stewards each, is ninety-nine seats — and that is before the Delaware cohort series, before replacements, before recusals. The files do not say where those stewards will come from, what they will be paid, if anything, or how conflicts will be handled when a steward sits on more than one council. Those are not criticisms of the design. They are the work the design creates, and the work has not been done yet. A council is three humans. Ninety-nine humans is a community, and communities take time to build — which is one more reason the honest statement is that no council has been seated yet, and the first seating is the work that matters most.

Two registries, not one

There is a confusion waiting to happen in the steward story, and this section exists to prevent it. The word steward appears in more than one project in the family's orbit, and the registries behind those projects are not the same registry. The main Omega Hive registry and the SmartCiti agents registry are separate projects, with separate counts, and they must not be merged into a single number or a single story. The main Omega Hive registry lists 44 stewards, with 1 seated. The SmartCiti agents registry lists 0 seated. Those are different registries, different projects, and different states of readiness — and the series Steward Councils described in the corporate files are a third thing again: the per-series councils of three consenting humans, of which none has been seated for any series.

The distinction matters because the numbers look similar enough to confuse and different enough to mislead. A reader who hears that one steward is seated might conclude that the series councils are underway. They are not. A reader who hears that zero are seated in one registry might conclude that nothing has been seated anywhere. That is also wrong. The Omega Hive registry, the SmartCiti agents registry, and the series councils are three separate ledgers, and each one has to be read on its own terms. Conflating them would produce a sentence that is wrong in every direction at once: too optimistic about the series, too pessimistic about the Hive, and wrong about SmartCiti altogether.

The reason the family has more than one steward registry is that it has more than one thing that needs stewarding. The Omega Hive is an agent project with its own registry of stewards. SmartCiti.X has an agents registry of its own. The series councils are the governance bodies of the operating companies. Each registry answers a different question: who stewards the Hive, who stewards the SmartCiti agents, and who governs a given series. The corporate files govern the third question. They say nothing about the first two, and the first two say nothing about the third. Keeping them apart is not pedantry. It is the only way to say anything true about any of them.

There is a broader lesson in the separation, and it is one the design keeps teaching. The family is not one organisation with one governance system. It is a family of entities — a filed nonprofit Foundation, a separate church, an intended Wyoming head corporation, an intended Wyoming series company, an intended Delaware cohort company, thirty-three named series, and a set of network projects that sit outside the series list altogether — and each of them has its own governance, its own records, and its own state of readiness. A sentence that is true about one of them is not automatically true about the others. The steward story is a case in point: seated in one registry, unseated in another, and not yet begun in a third. The honest account holds all three facts at once, without letting any one of them stand in for the others.

For the series councils, the fact that matters is the third one: none seated, for any series. That is the fact the corporate files support, and it is the fact this feature repeats, because it is the fact that determines what the councils can do today. A council that has not been seated cannot approve a budget, cannot move funds, cannot enact a charter, and cannot pass a project to Build. The series are capped at Gate 1 until the councils exist. Everything else — the registries, the counts, the Hive — is context. The councils are the work.

The council and the board: who governs what

The Steward Council governs the series. The Foundation itself is governed by its board of directors, and the files are emphatic that the board must stay independent. A California nonprofit public benefit corporation cannot be owned or controlled by a for-profit head corporation the way a subsidiary is. Its board owes fiduciary duties to the charitable mission and the public, not to any other entity's shareholders. In the intended structure, AGI Corp sits at the top commercially as the head family corporation, while the Foundation acts as mission steward and manager — but its authority comes from written management and mission agreements, not from ownership by AGI Corp. The council and the board are therefore different bodies with different duties, and confusing them would break the design at its most sensitive joint.

The boundary between them is drawn in the files with unusual care. The three-seat, two-of-three signed-vote design described in the corporate files is the design for each series Steward Council. The Foundation itself is governed by its own board of directors, and the files do not specify a Foundation-level Steward Council with a seat count or voting rule — counsel and the board are to decide that. Likewise, the AGI Fellowship church is a separate entity with its own governance; its council design is not the Foundation's, and nothing about it is imported here. Three bodies, three designs, three sets of rules. The council is the smallest and the most operational of the three. The board is the most consequential, because it holds the mission gate.

The draft bylaws describe the board's mechanics, and they are drafts — not adopted, not filed. As drafted, the board would consist of three or five directors, a number the board must choose, serving two-year terms, with a quorum of a majority of directors then in office and action by a majority of directors present at a duly held meeting; action without a meeting would require unanimous written consent. The corporation would have no members: all corporate powers are exercised by the board. Directors serve without compensation, except reasonable documented reimbursement. The draft names Mike Rice as an initial director and co-founder, and Russell Vacanti as co-founder — with consent pending, meaning Russell Vacanti must confirm in writing that he accepts the role before the bylaws are adopted with his name in them — and a third director still to be recruited. No other council or board members are named in the files, and none are named on this page.

The relationship between the council and the board is the relationship between a series and its mission steward. The Foundation board holds mission-fit approval, mission-standards enforcement, and grant decisions, and it can withdraw mission approval for a project. That project then loses Foundation support and Foundation branding, even if AGI Corp wants to continue it commercially. The council, for its part, governs the series day to day and quarter to quarter: budget, hiring, major contracts, fundraising, charter changes. Neither body can do the other's job. The board cannot approve a series budget. The council cannot grant mission approval. The separation is the design's answer to the question of who watches the watchers: the council watches the series, the board watches the mission, and counsel watches the law.

There is one more body in the picture, and it belongs to the studio, not to any series: the AGI Corp board, at studio level, which handles portfolio allocation, studio fees, SAFE and equity authorisation with counsel, hiring of studio leadership, and final kill or pivot decisions where the series' agreements give the studio that right. Four powers, four bodies: the council for the series, the Foundation board for the mission, the AGI Corp board for the portfolio, the Fellowship board for the church. No single human, and no agent, holds all four. That separation is the point of the studio model as a governed venture studio, rather than a single-person project list — and the council is the smallest, most human piece of it: three people, two votes, and the only keys to the money.

What a council cannot do, and what remains undecided

For all its power, the council is bounded, and the bounds are as important as the powers. A council cannot override the Foundation's mission gate. A council cannot issue a SAFE, equity, or profit share on behalf of the filed nonprofit Foundation, because the Foundation cannot issue them at all. A council cannot spend Foundation donations as if they were series revenue, and it cannot spend series funds as if they were Foundation grants. A council cannot enact doctrine for the Fellowship, which is a separate church with its own board and minister. A council cannot sign a contract in the name C1S1 alone, because C1S1 is a reference label, not a legal name. And a council cannot act at all until it is seated — named, consenting, holding keys — which is why the empty chairs are the most important fact in this feature.

The council is also bounded by the record-keeping discipline that surrounds it. The inter-entity and conflict-of-interest policy — itself a draft, to be adopted separately by each board — requires that any director, officer, employee, or key person with a financial interest in a transaction disclose it in writing before discussion, including the other entity, the nature and size of the interest, and the price or terms if known, with the disclosure recorded in the minutes. The interested person leaves the meeting for discussion and vote, except to answer factual questions if invited. Approval requires a majority of disinterested directors, who must minute that the transaction is fair, reasonable, and in the entity's best interest. The compliance calendar turns that discipline into a routine: board meetings and minutes at least annually and as the bylaws require, conflict disclosures annually and on any new transaction, monthly reconciliation, and a filing that is not done until proof is filed in that entity's pack.

What remains undecided is a long list, and the files keep it visible rather than filling it. No Foundation-level council specification: the files do not specify the Foundation council's seat count or voting rule. No named stewards: no Steward Council members are confirmed in the files, and none are named here. Drafts, not adoptions: the Foundation bylaws, board consent, and inter-entity conflict policy are drafts, the bylaws' board size is a bracketed choice of three or five, the third director is still to be recruited, and Russell Vacanti's co-founder and director role is consent pending. Filings still to verify: no IRS 501(c)(3) determination letter is in the files for the Foundation, and the Statement of Information and California Attorney General registration are not evidenced in the uploads. Structure is intended, not filed: the Wyoming AGI Corp head, the W3 LLC Wyoming series company, and the Delaware cohort series company are intended designs, not verified filings.

Economics are blank. Studio ownership in a series, founder and team equity, management fees, SAFE terms, revenue sharing, Windfall terms, and Open Global Investment terms are all blanks for counsel and the board to set in signed agreements. A term that is not in a signed agreement does not exist. And the council's own succession rules — who replaces a steward who resigns, becomes unavailable, or loses a key — are left to each series' charter, which has not been signed for any series. The design is honest about all of this. It does not pretend the councils exist. It does not pretend the economics are set. It does not pretend the filings are done. It describes a human core — three people, two votes — and then says, plainly, that the core is empty.

The honest ending of this feature is therefore not a portrait of a council at work. It is a portrait of a chair waiting to be filled, ninety-nine times over, across thirty-three series. The design's bet is that three named, consenting humans holding signing keys can govern a company run mostly by agents — that the human core can be small without being weak, and that two signed votes can hold the line against speed, against momentum, and against the temptation to let the machine decide. It is a bet on people, made in a design that is otherwise about machines. Whether the bet pays off will be decided the first time a council is seated, the first time it says no to a proposal the agents drafted beautifully, and the first time it kills a project that should have died. Until then, the council is the most important room in the design — and it is empty.

What this is / What it isn’t

What this is

  • A per-series human core: three named, consenting humans holding signing keys, deciding by two-of-three signed votes, holding the only authority to move funds or change a charter.
  • A story about scale and trust: how a three-person council is meant to govern a company run mostly by agents, and how the two-of-three rule keeps any single person from acting alone.
  • A public record of the separation: the series councils, the Foundation board, and the two steward registries are different bodies and different ledgers, kept apart rather than merged.

What it isn’t

  • Not a seated body: no Steward Council has been seated for any series, no DAO charter has been signed, and no series is operating under a council today.
  • Not the Foundation board: the three-seat, two-of-three design belongs to each series council, not to the Foundation board, whose seat count and voting rule counsel and the board are still to decide.
  • Not a single registry: the Omega Hive registry and the SmartCiti agents registry are separate projects, and neither one is a series Steward Council.

Where to see it

The governance pages carry the full detail behind this feature: Steward Council, Governance hub, and Share Structure, with the operations detail on the Proposal Flow and the Human Gate. The Metropolis-SmartCiti repository is public on GitHub: Metropolis-SmartCiti — Omega Stewards on Monad testnet, with 2-of-3 passkey steward governance and contracts deployed, 41 of 41 transactions confirmed on testnet, never presented as a live production network.

Nothing on this page is legal, tax, or financial advice. The council design, the bylaws, and the conflict policy are drafts and intended designs. The Foundation is a filed California nonprofit public benefit corporation, and no IRS 501(c)(3) determination letter is in the files. Registry figures for the Omega Hive and SmartCiti agents registries are stated as given in the project materials; the series councils are a separate design, and none has been seated.

← Prev: The Mission Gate · Next: The Open Door →