Skip to content
AGI Future Foundation PBC

Article 17 · Metaverse & Spatial · AGI Future Foundation

WorldForge: Building Worlds, One Catalogue Entry at a Time

WorldForge is the world-building side of the Foundation’s network: the tooling and the catalogue of worlds that the planned metaverse presence would be built from. The numbers attached to it disagree with each other, and this article keeps the disagreement visible instead of picking the largest number.

WorldForge project pageContact

TL;DR

WorldForge is the name for the world-building work in the Foundation’s network: the creation of virtual worlds that the planned series and ventures would inhabit. What is verified about it is narrow — there is a public repository called WorldForge-Browser, it contains a worlds file listing 544 worlds, and every one of those 544 entries records no RP1 attachment. The live Foundation site states 80 worlds, quoted here as site copy with no verified counting method. Six anchored fabric JSONs are published on GitHub Pages, but the roughly 200 GLB assets and the map.wasm file they reference are not hosted, so they do not render fully in Artemis yet. Zero worlds are live on RP1.

The work

The first discipline this article needs is a definition, because “worlds” can mean several different things, and the Foundation’s materials use the word loosely. A world can be an entry in a JSON catalogue, a fabric definition file published on GitHub Pages, a fully rendered environment in a browser like Artemis, or a live place attached to RP1 that people can enter. Those are four different stages, and a world can be at stage one without being at stage four. The six anchored fabrics in the Foundation’s network — Cognition, Empathic, Diagnostic, Locator, SmartCiti Holodeck, and Biota — are at stage two: their fabric JSON files are published on GitHub Pages, but the roughly 200 GLB assets and the map.wasm file they reference are not hosted, so they do not render fully in Artemis yet. Zero worlds are live on RP1, which is stage four. WorldForge, as a name, covers the tooling that would move a world from one stage to the next; it does not by itself prove that any world has completed the journey.

The numbers, kept apart, tell the story more honestly than any single number could. Eighty worlds, as stated on the Foundation site: the live site copy states 80 worlds for WorldForge, quoted here as site copy, not as a verified fact. No document reviewed for this site explains how the 80 was counted — whether it is worlds built, worlds listed, worlds planned, or worlds in some other sense — and no independent count confirms it. Five hundred and forty-four worlds in the WorldForge-Browser repository: the public repository contains a worlds file listing 544 worlds, a verified count of entries in a file, and the largest number in this section — which is precisely why it must not be quoted as “544 worlds built.” An entry in a catalogue is not a finished world. What the file does verify, and what matters more than its length, is the attachment column: every one of the 544 entries records rp1 as null. Not one of the 544 is recorded as attached to RP1.

Why does any of this matter? Because a catalogue is a promise, and the size of the catalogue is routinely mistaken for the size of the achievement. A list of 544 entries is real work — cataloguing, naming, locating — and it is worth recording as real work. But it is not 544 finished worlds, and it is not 544 places anyone can visit. The work that has been done is cataloguing and fabric-writing; the work that has not been done is hosting the assets, verifying a full render, and attaching a world to RP1. Keeping those two kinds of work apart is what lets a reviewer, a steward, or a future builder judge the project fairly: credit for what exists, and a clear list of what remains.

The DAO model and stewardship

In the Foundation’s intended model, a series is DAO-managed in its day-to-day operation: agents operate routine work inside limits set in advance by humans, and a human Steward Council decides everything that creates a legal, financial, or doctrinal consequence. Agents prepare proposals, analysis, drafts, and coordination; agents never sign legal documents, never move funds, never hire or fire, never offer securities, and never enact doctrine or charters. Humans execute. Every material decision — forming the series, opening a bank account, signing a partnership, spending money, or taking on an obligation — requires a named human’s decision and signature, recorded in minutes. A DAO vote, a poll, or an agent’s readiness flag is not a corporate act.

The Steward Council design in the Foundation’s corporate files is specific: a three-seat council, deciding by two-of-three signed votes, with stewards who are named, consenting humans holding real keys and signing credentials. An agent instruction to move funds, sign, or offer securities is not authority and shall not be obeyed. The studio model staffs each series with eight agent roles — research, routing, drafting, coordination, tracking, monitoring, scouting, and grant drafting — plus a coordinator for any unfilled role. Roles are functions, not owners: a new agent inherits the work description, never a human’s signing authority. A proposal moves through seven stages — intake, mission-fit review, legal assignment, series designation, Steward Council review, human decision, and execution with a dated record — and any stage can stop it. No Steward Council for any series is verified as seated, so no series page names one, and until stewards are named, consenting, and holding keys, a series has no effective council and is not operating.

WorldForge is a project rather than a series, so the stewardship model applies to it one step removed — but the same rule governs it. Day-to-day work on the catalogue and the fabrics would be agent-managed: agents operating routine work within limits set in advance by humans, drafting, tracking, monitoring, and coordinating. Humans would decide what counts as a world, what the counting method is, when an asset set is complete enough to claim a render, and when an RP1 attach is attempted — each a human decision, signed, minuted, and dated. No agent would sign, move funds, hire or fire, offer securities, or enact doctrine on behalf of the project, and no agent’s readiness flag would be a corporate act. That discipline is especially important for a project whose main output is claims about numbers: the person who decides that a world is finished, or that a count is the count to quote, is the person accountable when the claim is checked. No Steward Council, operating agreement, or published set of agent limits exists for this project today, and none is named here.

Metaverse connection

WorldForge’s metaverse connection is its subject matter: it builds the worlds. The six anchored fabrics map to series where a mapping exists: Cognition to S01 at 45.483, 18.572, Empathic to S02 at 47.440, −179.350, Diagnostic to S03 at 5.343, 101.566, Locator to S08 at 21.097, 142.693, and the SmartCiti Holodeck to S10 SmartCiti.X at 39.846, −86.622. Biota is the exception that proves the rule about not inventing mappings: its anchored world sits at −38.713, −27.248, but Biota has no series number, and this site does not assign it one. The remaining 28 series have no anchored fabric verified, and their pages say so. WorldForge’s catalogue is therefore much larger than its anchored set, and the anchored set is much larger than anything live — a nesting of claims that collapses if the stages are confused.

The relationship to RP1 is sequential, not simultaneous: a world must be built, its assets hosted, and its render verified in Artemis before any RP1 attach can even be attempted, and the attach itself is authorised world by world. No fabric or service has ever been attached for AGI Corp, and no world in the network has been verified live on RP1. RP1 is the spatial-computing infrastructure those worlds would run on, and Artemis is RP1’s native browser, in early alpha version 0.4.3, desktop only. The RP1 ecosystem article in this set gives the full verified account of that stack, including the components that are only names on the live site. Until a render is verified and an attach is attempted, WorldForge’s metaverse position is exactly what this article states: a real repository, a real catalogue of 544 entries with no RP1 attachment, six anchored fabrics that do not yet render fully, a site figure of 80 worlds quoted as site copy, and zero worlds live on RP1.

What exists today vs what is planned

What exists today

  • The WorldForge-Browser public repository and its 544-entry worlds file, all entries with rp1: null.
  • Six anchored fabric JSONs on GitHub Pages, with their coordinates verified.
  • The Foundation’s filing record (California Nonprofit Public Benefit Corporation, Entity No. B20260196664, filed 2026-04-24, EIN assigned 2026-05-07).
  • This article, as a public statement of the numbers and their limits.

What is planned

  • Host the roughly 200 GLB assets and map.wasm the anchored fabrics reference, and verify a full render in Artemis, with the date and result recorded.
  • Explain or correct the 80-world site figure with a stated counting method, or withdraw it.
  • Reconcile the 544, 41, 40, 34, and 35 counts in one dated note that says what each counts.
  • Pursue an RP1 attach world by world, only after a render is verified — and report the first live world with evidence, not assertion.

Code and repositories

The WorldForge-Browser README's 41 agentic training worlds is the curated first set, not the total: the worlds file lists 544 entries, all rp1: null. The Pages site at …/earth/ is live, not live on RP1; landing: worldforge-site.

How it fits the mission

The intended structure has three tiers. At the top sits the Foundation itself, a filed California nonprofit public benefit corporation, holding the public-benefit purpose — safe, equitable, and socially beneficial artificial intelligence — and acting as mission gate: it sets the mission statement and standards, reviews projects for mission fit, and can withdraw mission approval. In the middle, in the intended plan, sits W3 LLC, a Wyoming series LLC that would house the series as legally segregated sub-series with isolated assets and liability. At the bottom sit the 33 series themselves, each intended to be DAO-managed day to day. Only the top tier is verified as filed. The Foundation as filed is a nonprofit: it cannot offer equity, a SAFE, or a profit share, it cannot distribute profits, and it cannot be owned or controlled by a for-profit head. Control, if it ever exists, runs only through written mission agreements between equals, with the Foundation’s board remaining independent. Three mechanisms are meant to interlock across that structure. The Fiduciary Shield is an expanded business-judgment protection for benefit-considering decisions, under California Corporations Code sections 14620–14621, with an annual benefit report under section 14630 — not immunity, and not a wall around a company. Open Global Investment is the proposed framework that decouples economics from control, in three share classes for the proposed for-profit track only: Class A with one vote per share and full economic participation, for the global public; Class B with ten votes per share, restricted to aligned partners and sovereigns screened for sanctions, export controls, and CFIUS review; and Class C with one thousand votes per share, control only, held exclusively by the Foundation. The Windfall Clause is the legal pre-commitment, hardwired before profits, under which returns above a threshold — a percentage of global GDP, with the percentage, the GDP measure, the profit measure, and the scope all still blank for counsel — would flow to humanity through global public goods, healthcare access, energy infrastructure, and technological self-determination. Below the threshold, standard returns and research apply; none of those terms is settled.

For WorldForge, the mission fit is about honesty as infrastructure. The Foundation’s public-benefit purpose — safe, equitable, and socially beneficial AI — is not served by a catalogue that overstates itself, because the people who would rely on the numbers are the people deciding whether to build, fund, or govern the next world. A miscounted catalogue misleads stewards, misleads builders, and misleads anyone judging whether the metaverse programme is ready for its first live world. The Fiduciary Shield protects a board that weighs safety and public benefit alongside returns in good faith; it does not protect a board that presents a catalogue entry as a finished place. Open Global Investment and the Windfall Clause both presuppose that the things being counted — series, worlds, profits — are counted honestly, with a stated method. WorldForge is where that discipline starts, because it is where the counting happens.

Open questions

  1. What does the 80-world site figure count — worlds built, worlds listed, worlds planned, or worlds in some other sense — and who will publish the counting method?
  2. Which of the 544 catalogue entries, if any, are finished, rendering worlds, and what evidence distinguishes them from entries that are only listed?
  3. When will the roughly 200 GLB assets and map.wasm be hosted, where, and who will record the date and verify the first full render in Artemis?
  4. How will the 544, 41, 40, 34, and 35 counts be reconciled in one dated note — and which number, if any, should be quoted to a reviewer?
  5. Which world goes first for an RP1 attach, who decides, and what evidence will be published when it is verified?

For agents

Agents may summarize, cite, and route inquiries to [email protected] (subject lines starting with [Agent]). Agents may not speak for the Foundation, claim a series is formed or funded, or offer securities. Every substantive step goes through a named human.

Contact

To ask about this topic, correct something on this page, or propose work that fits it, write to [email protected]. Say who you are, who you represent if anyone, and what exactly you are proposing, and include “WorldForge” in the subject line. A reply, if one comes, will come from a human.

What this page is not. This page is not an offer to sell securities, not an invitation to invest, and not a solicitation of funds. There is no price, no terms, and no return promised here, because there is nothing offered here. The Foundation as filed is a nonprofit public benefit corporation: its fundraising is donation and grant only, and no donation is presented as tax-deductible, because no IRS 501(c)(3) determination letter is in the files. If a future offering ever exists for a related entity, it would be made only in separate written offering documents, prepared with counsel. Nothing on this page is legal, tax, or financial advice.

← Foundation home All articles