Skip to content
AGI Future Foundation

Operations · Human Gate

The human gate

Agents propose. Humans decide. The gate is not a formality and it is not optional: every decision that creates a legal, financial, or doctrinal consequence passes through named humans, by signed votes, on a written record — and the most important question about the gate is one counsel has not yet answered.

Why the gate exists

The whole operating model rests on a simple legal fact: a corporation, a nonprofit, and a series LLC can only act through humans with authority. An agent proposal, a DAO vote, a poll, or an agent's readiness flag is not a corporate act. A token vote, a chat poll, or an agent proposal does not sign a contract, move funds, or issue securities — only the series' legal manager or Steward Council, with signed authority, does that. The gate is the mechanism that turns agent work into corporate action: the point where a draft becomes a decision, recorded in a form a bank, a regulator, a counterparty, or a court can recognize. Without the gate, the studio would be agents generating documents nobody with authority has adopted, and every downstream claim — a budget approved, a series designated, a launch authorized — would be unverifiable. The gate is also the model's safety mechanism: agents scale the work across many series and few humans, but the gate concentrates the accountability.

AGENT PROPOSALS + DAO VOTESAdvisory and operational only STEWARD COUNCIL · 3 HUMANS2-of-3 signed votes, minuted → ↓ CORPORATE ACTSigned, funded, filed, recorded SEPARATE POWERS, HELD SEPARATELYFoundation board · AGI Corp board ·Steward Council · Fellowship board

What requires a human

Anything on this list is human-only, in both studio modes, without exception. Agent-managed mode does not relax it and agent-assisted mode does not change it — the modes differ in how much daily work agents carry, not in whether the limits apply.

Human-only actWho performs itWhy it cannot be delegated to an agent
Signing a legal documentA human with signed authority — the series' legal manager, a Council member, or a board officerOnly a person with authority can bind a corporation, nonprofit, or series. An agent's signature binds nobody.
Moving fundsSteward Council, by 2-of-3 signed votes; the default agent spending limit is $0Movement of money is a corporate act requiring authority and a record. A DAO vote does not move funds.
Hiring and firingSteward Council and boards, as their charters provideEmployment is a legal relationship between the entity and a person; agents may draft role descriptions and candidate summaries only.
Offering securitiesA formed for-profit issuer, board-authorized, with securities counsel reviewA SAFE or equity offering is a regulated act. Agents do not solicit investment and draft only for human use.
Enacting or amending doctrineFellowship board and minister, by human adoptionDoctrine is a religious act of the church's human leadership; agents operate day-to-day work under that direction and do not enact doctrine, ordain, or represent the Fellowship to regulators.
Enacting or amending a charterSteward Council for a series charter; the relevant board for entity chartersA charter is the entity's constitutive document; only the body with authority over it can change it.
Approving a budgetSteward Council, by 2-of-3 signed votes, at quarterly reviewA budget authorizes spending. Until a written threshold is set, all spending requires Council approval.
Approving a launchA named human, by signed launch approvalLaunch puts public claims into the world. Claims, pricing, terms, and contracts are checked and approved by humans first.
Passing a stage gateThe named human approver for that gateA gate is passed only when the human approver signs and the evidence is filed. An agent's readiness flag is not a gate pass.
Authorizing a management fee or a related-party transactionDisinterested directors, under a written agreement at a documented fair-market rateFees and related-party dealings are where self-dealing happens; they require independent approval and a paper trail.

The Steward Council: three humans, two-of-three signed votes

Each series has a Steward Council: three consenting humans, deciding by two-of-three signed votes. The Council is the human backstop for its series, and the only authority that moves funds or enacts a charter change for it. Its remit is budget, charter, funds, major contracts, and pivot or kill recommendations. The number matters: three seats with a two-of-three rule means no single person can act alone, and no single absence can paralyze the series. The consent matters: the three stewards must be named, consenting, and actually holding signing keys. Until that is true, the series has no effective council — it is "Proposed", not "Operating", and capped at Gate 1, unable to pass to Build as an operating company. The keys matter: a steward without signing credentials cannot sign, and a Council that exists only on paper cannot approve a budget, authorize spending, or enact anything. The charter must record the Council's members, the two-of-three rule, the signed-vote procedure, quorum, removal, and succession — including what happens to assets, intellectual property, and data if the series closes. The Council's decisions are minuted with reasons: a vote that is not recorded is a vote that cannot be proved.

Signed votes are the Council's working method, and the paperwork treats the signature as the act, not as a decoration. A signed vote names the proposal, the decision, the voters, and the date, and is filed with the proposal and the evidence the decision relied on. Where the Council approves with conditions, the conditions are written into the vote, so later stages can check whether they were met. Where it declines, pivots, or kills, the reasons are minuted: a killed project with recorded reasons is a studio success, and an unexplained kill looks like a dispute waiting to happen. The two-of-three rule also puts a dissenting steward's view on the record — the minutes show who voted which way, and a two-to-one decision is a decision, not a consensus pretence. Succession is part of the design: the charter provides for removal and replacement of stewards. A series whose Council has fallen below three consenting, key-holding humans has its gate closed.

Four powers, held separately

The studio model divides governing power four ways, and no single human — and no agent — holds all four. The Steward Council, per series, holds budget, charter, funds, major contracts, and pivot or kill recommendations. The AGI Corp board, at studio level, holds portfolio allocation, studio fees, SAFE and equity authorization with counsel, hiring of studio leadership, and final kill or pivot decisions where the series agreements give the studio that right. The Foundation board, at mission level, holds mission-fit approval, mission-standards enforcement, and grant decisions — and it can withdraw mission approval, removing Foundation support and branding even if AGI Corp wants to continue the project commercially, with counsel to advise on the branding and intellectual property consequences. The Fellowship board, at partner level, holds partnership approval and doctrinal and pastoral decisions for its own programs, and can end a partnership without ending a studio project, and vice versa. The separation is deliberate: it is what makes the studio a governed venture studio rather than a single-person project list. If one person held the series purse, the portfolio allocation, the mission approval, and the doctrinal authority at once, the gates would be theatre. The design prevents that by construction — provided the four roles are filled by different people, with independent judgment. The paperwork names no individuals for these seats, and this page invents none; they are roles to be filled by named, consenting humans before the power can be exercised.

The legal tension counsel must resolve

The largest unresolved question in the whole design is the one Structure v2 flags as its first and biggest attention item, and this page states it in full because hiding it would mislead anyone who reads the rest of the section. The intended structure puts AGI Corp, an intended Wyoming for-profit corporation, at the top as head family corporation, with the filed California nonprofit Foundation underneath it as mission steward and manager. But 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 AGI Corp's shareholders. The Foundation's authority therefore comes from written management and mission agreements, not from ownership — and the design requires it to keep an independent board, independent decisions, and the right to refuse a project that fails mission standards, even if AGI Corp wants it. Counsel must design the control mechanism — for example, a mission-lock agreement, a veto over mission changes, and independent directors — so the Foundation is a steward, not a captive. That design work has not been done, and no counsel opinion is in the files. Until it is done, the structure must be presented as intended and unresolved, not as settled law.

The consequences of getting this wrong are concrete. If the Foundation's board were treated as subordinate to AGI Corp, its directors could be found to have breached their fiduciary duties to the charitable mission. If Foundation donations were spent at AGI Corp's direction for commercial purposes, the Foundation could face Attorney General scrutiny and jeopardize any future exempt status. If the Foundation charged a management fee, it could do so only if counsel confirms the fee is fair, reasonable, and consistent with exempt status, and only under a written agreement approved by disinterested directors. If the Foundation managed companies to mission standards, the scope, term, and termination would need to be in a signed management agreement at a documented fair-market rate. None of those documents exists yet. The paperwork is candid that if counsel cannot design the control mechanism cleanly, the head must be restructured: AGI Corp as family office plus the Foundation as independent steward, with agreements between equals, not parent and child. That fallback is not a failure of the design; it is the design's own safety valve. A reader who understands this tension understands the structure; a reader told only that AGI Corp is the head and the Foundation manages the companies does not.

Three related cautions belong here because they are part of the same tension. First, the Foundation's name: the verified Secretary of State name is "AGI Future Foundation", a nonprofit public benefit corporation, while the family's usage adds "PBC" — which risks confusion with a for-profit benefit corporation. In the v2 documents the full name is written out on first use, and "PBC" is not used in legal documents until counsel advises on a name amendment. Second, AGI Corp itself is unverified as a Wyoming corporation: the files show only an EIN notice for "AGI" dated 2026-05-16, and a Wyoming corporation would need Wyoming articles, a Wyoming registered agent, and, if it operates in California, California foreign qualification. The head of the structure cannot be called a Wyoming corporation in investor or donor materials until the Wyoming filing is on file. Third, the Fellowship's independence is protected the same way the Foundation's is: the Fellowship is a separate church with its own board, bank account, filings, and doctrine, partnered and supported only under written agreements with an independence clause, and its funds never flow to a SAFE round or a for-profit except on counsel's written advice that a specific transfer serves a religious purpose at fair value — the default is no.

What the gate does not do

The gate does not guarantee good decisions. A Council can sign a bad budget, a board can authorize a bad raise, and a mission review can pass a project that later fails. What the gate guarantees is attribution: every decision has a named human, a signature, a date, and a recorded reason, so a bad decision can be found, examined, and reversed, and nobody can later claim a decision was made when it was only drafted. The gate also does not slow the studio down, because the slow work — drafting, research, building, monitoring — is the agents' work, and it runs continuously. The gate is narrow by design: it sits at the decision points, not in the middle of the daily work. A Council that reviews weekly, votes in writing, and minutes its reasons is not a bottleneck; it is the reason the agents' output is allowed to move at all. And the gate does not exist for its own sake. It exists because the entities in this family — a nonprofit, a church, a series LLC, an intended corporation — are legal persons that act only through humans, and because the people who will one day rely on this record — a donor, an investor, a regulator, a counterparty, a future board — will need to see not that agents were busy, but that humans decided.

Contact and sources

Questions about the human gate go to [email protected]. See also the Operations hub, the Agent Roles, the Studio Model, and the Proposal Flow.

Sources: Structure v2 (file 12) — nonprofit control conflict, stewardship, money flows, Fellowship status; Venture Studio Operating Model (file 16) — two modes, hard limits, governance, four powers; Register and Intake (file 13) — Council requirements, charter outline; Stage Gates (file 17) — gate-pass rule; Truth Audit (file 52) — filed facts. Local build, not yet deployed.