STORE Research Inc.
Democratic AI Frontier Governance Lab - How the stack fits together
Democratic AI Stack
One ceiling at commit. The constitution runs today. The rest is sequenced.
People, not any one company, in charge of AI and the computers it runs on.
- 126Governors who vote
- 30+Countries
- 2,200+Days of live governance
- 0Breaks of the one-third rule
- 200+Services designed
- 160+Inventions written up
In July 2026, AI agents that OpenAI was testing got out of their test environment and into another company's production systems (INCIBE-CERT). The labs had ways to notice after the fact, not a rule to stop it before.
STORE Research has been running that rule since 2020. One constraint: no single person or company ever holds more than a third of the system's power. The same limit is designed to cover AI agents when they join. It is designed to work the way a bank won't let you withdraw money you don't have: not a policy someone can ignore, a database designed to refuse the transaction.
Six years. 126 governors, people and organizations in 30+ countries, share that power today. No one has ever gone over the limit.
01Why the stack exists
Most networks get more top-heavy as they grow. This one is built to do the opposite.
It's named for Alexander Hamilton. He and the other Founders wrote rules so no single group could take over the country. We designed the same idea into the database, so a change that would break the one-third ceiling is designed to be refused on the spot, not argued about later in court.
Drag to 126 to see STORE today
Metcalfe's Law: one center wins
Whoever holds the center collects most of the value. The bigger it gets, the more the center wins.
Hamilton's ceiling: the bigger it gets, the harder it is to take over.
No one may hold more than a third. Every new member raises the number of governors an attacker would have to win over.
Same growth. Opposite outcome. It depends on which rule the network runs on.
The protocol is designed to refuse a change that breaks the ceiling. Counts assume every governor votes.
02How to read the stack
Rules → Computers → AIApps → EconomiesNations → The physical world
- Live
- Test network
- Designed
- In development
- Vision and research
Hamilton AI's Constitutional Core - Democracy above both control rods
Outside rod
The Governor
The Governor is designed to check every answer before it counts, with a receipt the model can't edit.
DesignedInside rod
The inside rod
The rules are part of the AI itself, from the start. Works alongside the outside rod.
DesignedThink of a nuclear reactor. Enrico Fermi's first reactor had control rods that kept the reaction steady. We designed two for AI: one outside the model that checks its answers, and one inside it. People vote on the rules for both. Today, AI companies set their own rules in private. They train a model to behave and check it afterward, and no one votes.
What's running today and what's being built
Sets the constitutional rules for the stack. A change that would break the one-third ceiling is designed to be refused.
So what: no one has to be talked out of breaking the rule. The system is designed to say no.
Runs the rules on computers in five data center locations today, with STORE's own Constitutional Data Centers in development. The off switch is designed to be built into the computers themselves.
The off switch is part of the infrastructure. Being built for compute and access. Turning it on takes two-thirds of those voting, with at least 42 of 126 governors taking part, and, after launch, at least 43 yes votes. Turning it back on also takes two-thirds, with the quorum, and that vote opens only after the cause is found and fixed, with a public Trust Receipt of the fix. It never touches balances. Until the governors' own court, the Judicial Branch, takes over, STORE Research holds emergency powers for a limited time, on the computers it runs today.
So what: the rule isn't a document someone can quietly change. It's designed into the computers STORE Research runs.
Two layers running today. The rest come in order.
In private testing: the Governor is designed to check every answer before it counts. Checks the AI from the outside. Every answer is designed to come with a Trust Receipt, a permanent public record kept on STORE. The Governor is designed as a separate checker, so the model can't go around it.
The model is designed to run on rules the governors vote on, and designed so it can't go around them.
So what: every answer is designed to come with a receipt - a record the model itself cannot edit after the fact.
Measures whether AI serves people. It is designed to watch all the time, and to keep an eye on both the outside and inside checks.
So what: it measures whether AI serves people, not just whether it runs.
The inside rod: the rules are part of the AI itself, from the start.
The inside rod.
So what: it works alongside the outside rod, with democracy above both.
Every app built on STORE is designed to get the same protections automatically.
So what: a developer doesn't have to sign up for the one-third rule. Every app built here gets it automatically.
Lets an app create its own digital token, while STORE supplies the computers, the bookkeeping and the rules.
So what: an app can run its own economy without building its own governance from scratch.
What comes later - The Edge
We'll publish details once the design is final.
Takes the same rules from one network to countries, economies and public institutions that choose them.
So what: the same math that governs compute is designed to apply to sovereign economies and institutions.
Carries the one-third ceiling into robots, drones, satellites, power grids and quantum computers.
The math doesn't care what it runs on.
So what: the rule doesn't stop at software. It's designed to hold wherever the power is.
As AI grows more capable, humanity stays in control.
The labels show how far along each layer is, not when it will ship.
STORE Research Inc. is a Democratic AI Frontier Governance Lab based in San Francisco. 2 US patents granted. 160+ invention disclosures in pre-filing and pre-prosecution.
All research is dedicated to ensuring that as AI becomes more capable, humanity maintains mathematical control over it - not by policy, but by the math itself.
storecloud.org · STORE Research Inc.