Draft for counsel review. Not yet incorporated.
Infinity OS Foundation Download

The Infinity OS story

AI factories every business can own.

The Infinity OS Foundation is an educational non-profit. It teaches businesses how AI factories work and helps define an open standard for them, a standard it is being set up to steward. The aim: every business runs factories of its own, on hardware it owns or controls, with whichever AI models it chooses.

Download Infinity OS Read our articles See our repos

Our plan

Define the standard first. Then, if we can, help the industry share it.

Our aim, if we can earn it, is to be for the harness layer, the layer that connects AI to a business's work, what the best-practice standard-setting bodies of accounting and law are for those professions.

We are loyal to the mission, not to the form. If a government body comes forward to set standards for the harness layer, we are open to it: we would evaluate working with it, working under it, or giving it our entire system. We are also open to handing our mantle of responsibility and our intellectual property to another non-profit, or to a private company that matches our mission and values. Until then, we stick to our principles. If stewardship ever moves, no one is stranded: everyone keeps the copy they downloaded, and only the maintenance moves.

Read the plan in full. Others have done this before us: see the precedents.

The Foundation's plan in two stages Stage one, now, drawn solid: define the factory standard, share it widely and free, teach how to use it. Stage two, later and only perhaps, drawn dashed: work with the industry on shared standards, as the best-practice bodies of accounting and law do for their professions. STAGE ONE: NOW Definethe factory standard Shareit widely, free Teachhow to use it STAGE TWO: LATER, PERHAPS Work with the industry on shared standards, as the best-practice bodies of accounting and law do.

The three ages of AI

Chat, then code, now factories.

We believe AI has come in ages. The third is beginning: AI that runs a whole function of a business, with people in charge.

Read: the three ages of AI
The three ages of AI as a rising staircase Three steps rise from left to right, with a bronze progress line along their tops. Age 1, chat: you ask, it answers; arrived. Age 2, code: you describe, it writes code; arrived. Age 3, factories: it runs a whole business function, people in charge; starting now. AGE 1AGE 2AGE 3 ChatCodeFactories You ask,it answers. You describe,it writes code. It runs a wholebusiness function,people in charge. Arrived Arrived Starting now

What a factory is

One business function, start to finish.

Software does the exact work, AI offers judgment, and people keep the decisions.

A factory names its job, the person accountable for it, who it serves, the records it owns and what it may do.

What a factory is: one business function from start to finish Inside a factory, work comes in, software does the exact work, AI offers judgment, a person decides, and checked, finished work comes out. Every step is recorded, and a named person, the owner, is responsible. A FACTORY, START TO FINISH Work comes ina request, an order, a case Software does the exact worksums, lookups, filing AI offers judgmentdrafts, sorts, suggests A person decidesapproves, sends, spends Checked, finished workkept on the record Every step recorded Owner a named person

The risk

A copied factory is a lost moat.

Write your operations down as a perfect factory and you have built something powerful, and something copyable.

A complete factory, and a competitor's perfect copy On the left, your factory: how you take work in, how you price, how you deliver, who you serve; years of your work. A dashed arrow marked access leads to a competitor's copy on the right, identical in every row. A fully realised factory is a written copy of how you operate; whoever can read it can copy it, and you would have no moat. YOUR FACTORY A COMPETITOR'S COPY How you take work inHow you priceHow you deliverWho you serveHow you take work inHow you priceHow you deliverWho you serve access Years of your work A perfect copy A fully realised factory is a written copy of how you operate. Whoever can read it can copy it. You would have no moat.

Your factory is yours

Own your factory. Plug the models in.

The companies building AI models are doing remarkable work. Their models will help every business, and we use them every day.

A factory is different from a tool you rent. It is how your business runs, written down. Built inside a platform run by the company that sells you the model, your operations live where that company sets the rules, and you may not know who can see them. We think a factory belongs on hardware the business owns or controls, with the model companies plugged in as service providers: used fully, and replaceable. A business owns its factories and need not depend on a for-profit to hold them; the Foundation runs none.

Before you build on any platform, ask:

  • Who can read my factory?
  • Can my usage improve a model that is sold to my competitors?
  • What recourse do I have if my work is used in a way I did not agree to?
  • Could I leave tomorrow, with everything?
Read: why your factory should not live in someone else's company
A factory on a rented platform beside a factory the business owns Top picture, rented: the business's factory and records are kept inside a platform the model seller controls; simple to start, and the platform's terms decide who can read it and how you leave. Bottom picture, owned: the factory sits on hardware the business owns or controls, and model providers A, B and C plug in from outside as service providers. Each model gets only what a task needs; a provider can be swapped and the factory stays. RENTED: ON A PLATFORM THE MODEL SELLER CONTROLS The model excellent work Your factory and records kept inside their platform Simple to start. The platform holds your factory, and its terms decide who can read it and how you leave. OWNED: YOUR FACTORY, ON HARDWARE YOU OWN OR CONTROL Your factory your methods, your records, your rules Each model gets only what a task needs. Swap a provider; the factory stays. Provider A Provider B Provider C PROVIDERS plugged in as services

Option source

Free to switch models and options.

Your software need not be open source. Your choices should stay open.

Option source over time: models and tools change, the factory stays Three lanes over time. The top lane, your factory with its methods and records, runs unbroken. Below it the model changes from model 1 to model 2 to a future model 3, and the tools change from tool A to tool B. Each swap is marked; the factory never breaks. YOUR FACTORY STAYS Your factory, its methods and records MODEL Model 1 Model 2 Model 3 TOOLS AND OPTIONS Tool A Tool B swapswapswap TIME

Open to start

Open blueprints to start. Your own factory to run.

Our factories are starter blueprints. The factory that runs your business is yours.

One open blueprint, three factories each owned by its business An open blueprint, free to start from, leads to three businesses. Each builds its own factory from it, arranged differently and locked to its owner. Each one changes it, owns it and runs it; their changes stay theirs. Open blueprint free to start from Business oneits own factory Business twoits own factory Business threeits own factory Each one changes it, owns it and runs it. Their changes stay theirs.

The central idea

Harness layer intelligence bookkeeping standards.

Every business already knows bookkeeping, accounting standards, regulations and best practices. They are why anyone trained can read any company's books. We want the same for intelligence: a standard way to record and run the AI work of a business, in the harness layer.

Bookkeeping for money beside bookkeeping for intelligence Two columns matched row by row. Money: a sale or a payment, a journal entry, the ledger, financial statements, accounting standards and audits. Intelligence: an act by AI, an entry saying who acted and on whose word, the factory's record, reports on AI work, and harness layer bookkeeping standards. The harness layer connects AI to your work and records. MONEYINTELLIGENCE you know thisthe same idea, for AI A sale or a payment A journal entry The ledger Financial statements Accounting standardsand audits An act by AI An entry: who actedand on whose word The factory's record Reports on AI work Harness layerbookkeeping standards The harness layer connects AI to your work and records.

Why an accountant cares

Andrew Warner is a CPA. He sees an AI factory the way an accountant sees books: it needs a shared, readable record of what came in, what was proposed, who decided, what was done and what proves it. He calls it intelligent bookkeeping, intelligence accounting.

Defining it well is part of the Foundation's work. It is not finished, and we do not claim it is.

Event
A supplier's invoice arrives.
Proposal
An agent matches it to the order and proposes payment.
Decision
The accountable person approves it.
Action
The payment is scheduled.
Receipt
A record of what was done, by whom, on whose word.

Example, made up. Event, proposal, decision, action and receipt are kept as separate records.

Principles

What we believe.

Seven principles.

  1. The third age of AI is the age of factories

    We believe AI has moved from chat, to coding, and now to factories: ready-made, connected ways of doing a whole business function.

  2. Option source: every business should be free to switch

    A business should always be able to swap its AI model, its tools and its provider without rebuilding its business.

  3. Your factory is yours; AI models plug in as service providers

    The AI model companies are excellent and essential, and every business should use them, but a business's factory belongs on hardware it owns, with the models plugged in as service providers, not built on a platform the seller of the model controls.

  4. Ready-made factories, built to an open standard

    Anyone starting a business should be able to begin from pre-built factories, made to one open standard that anyone may build to.

  5. People stay in charge

    In a factory, software does the exact work, AI offers judgment, and people keep the decisions.

  6. Every AI act on the books: intelligent bookkeeping

    Every act an AI takes in a business should be recorded as clearly as a ledger records money: who did it, on whose word, and what came of it.

  7. Teach first, then set the standard together

    Our first job is to define the factory standard well, share it widely and teach people how to use it; later, we aim to work with the whole industry so factories can work together.

Option source compared with open source A grid with code closed or open across the top and choices open or locked down the side. Both cells where choices are open are option source, whether the code is closed or open. Where choices are locked, the business is locked in, even if the code is open. CLOSED CODEOPEN CODE CHOICESOPEN CHOICESLOCKED Option source you can stillswitch open sourceas well Locked in Open code,still hard to leave Option source is about your choices, not your code.

The Foundation is funded and staffed by StarMynd today. Infinity OS is being built and is not yet released.

Governance

Our Board of Directors.

Our proposed founding board brings together financial leadership and AI education to advance the Foundation's mission.

Meet the board
The Foundation's proposed founding board Andrew Warner, CPA, founder of StarMynd: finance and technology. Mats Stafsberg of AI Impact: AI education. FINANCE & TECHNOLOGYAndrew Warner, CPAFounder of StarMyndDevelops Infinity OSAI EDUCATIONMats StafsbergAI ImpactPractical video instruction

Where things stand

Plainly, today.

  • Not yetThe Foundation is being formed as a non-profit and is not yet incorporated.
  • TodayStarMynd founded the Foundation and funds and staffs it. We declare the two separate.
  • IntendedStarMynd is a temporary steward while the open standard is adopted and proven. We intend to transition to an independent body over time, as it makes sense. No dates are set.
  • TodaySix public repositories in the starmynd-org GitHub organisation, each carrying the MIT licence.Measured 14 October 2026
  • BuildingInfinity OS is being built. The new version is not released, and nobody outside uses it yet.
  • PlannedThe factory standard, and downloadable pre-built factories. Neither is published yet.

Start here

Three ways in.

Download Infinity OS

It is early. Today you can download the early public code and the open starter, all MIT licensed. The version being built now is not public yet.

Download Infinity OS

Read our articles

The three ages of AI, why your factory should be yours, and why we run work from queues instead of inboxes.

Read our articles

See our repos

Six public repositories, each with its licence in the repository. Measured on 14 October 2026.

See our repos