People set the rules. No one is above them.

1. A constitution above every actor

Proposal v0.1

Start with protected rights and a short list of powers the Commonwealth may use. Any power not listed stays outside its authority. Require public ratification before these rules take effect.

Purpose

Keep a leader, a majority, a donor, or an AI system from granting itself power.

Tradeoff

Strict limits can slow useful action. Rights can conflict, and their meaning can be disputed.

Open question

Which rights should be beyond amendment? What approval thresholds and legal forms could protect them across countries?

• Protect privacy, fair treatment, free disagreement, voluntary participation, and the right to leave.
• Require notice, reasons, a fair hearing, and an appeal when a decision affects someone’s standing or rights. This is what we mean by due process.
• Define terms, removal, recall, and review for every office. A recall is a process that can remove an elected official before a term ends.
• Give changes several stages: public draft, review, a waiting period, and constituent approval. Ordinary votes cannot set protected rights aside.
• Make emergency powers narrow, capped, public, and temporary. They end automatically on a stated date; renewal needs a fresh lawful decision.
• The founder has no permanent veto, special voting weight, or right to rule. Helping start the work does not create a lifetime office.

2. Verified constituents

Proposal v0.1

A future verification process would check that each eligible person is real and counted once. Each would have one equal base unit of civic standing.

Purpose

Make participation fair without letting fake accounts or wealth multiply a person’s vote.

Tradeoff

Strong checks can expose private information or exclude people who lack documents, connectivity, or safe access.

Open question

How can verification protect privacy, resist coercion, include people across countries, and offer a fair appeal?

A constituent would be a person eligible to take part under future adopted rules. A confirmed email supporter is not a verified constituent. Email confirmation checks an inbox. It proves neither unique personhood nor eligibility, and creates no voting rights. The support registry remains inactive.

See the proposed standing and verification rules.

3. A Representative Assembly

Proposal v0.1

Elect several representatives in each constituency using a proportional method. Set limited terms, publish conflicts, record official votes, and provide recall or removal procedures.

Purpose

Let different views have a voice while giving a clear, accountable body the job of making ordinary policy.

Tradeoff

Representation adds distance between people and decisions. Organised groups may still gain an advantage.

Open question

How should constituencies, seat numbers, term limits, and recall rules be set without giving one region or group permanent control?

A constituency is a group of voters that elects a shared set of representatives. Proportional means seats should broadly reflect voters’ choices, rather than giving every seat to one winning side. The Assembly would not run its own elections. It could not override protected rights through an ordinary majority vote.

See the starting proposal for elections.

4. Temporary citizen panels

Proposal v0.1

Choose a temporary panel at random from eligible constituents, with steps to include different backgrounds and places. Let it study evidence, question experts, hear strong opposing cases, and publish majority and minority reports.

Purpose

Give everyday people time to examine hard questions, beyond donors, professional politicians, or a viral argument.

Tradeoff

Panels take time and money. People who accept invitations may differ from people who decline. Selection and briefing can shape the result.

Open question

Which questions need a panel, which groups must be included, and what support makes service possible for people with work or caring duties?

The panel would review major proposals, not replace the full constituent body or assume general lawmaking power. Its report must show where people disagreed. This draws on OECD guidance on representative deliberation; that guidance does not endorse this proposal.

5. Independent review and oversight

Proposal v0.1

Separate review of rights and procedure from ordinary policy. Keep election administration, custody of funds, audit, investigations, complaint support, and AI evaluation apart wherever practical.

Purpose

Make it possible to challenge decisions without asking the same people to judge their own work.

Tradeoff

Several bodies can cost more, disagree, or create slow handoffs. Reviewers can also be captured.

Open question

How should reviewers be appointed, funded, rotated, and removed while remaining independent?

A constitutional review body would hear rights and process disputes, election challenges, conflicts of interest, and administrative appeals. It could require correction or stop an unlawful action. It would not gain a general power to write policy.

An inspector general would investigate misuse. An ombuds office would help people raise complaints. Independent auditors would inspect money and records. AI evaluators would test models without being the team whose work they certify.

Match each decision to a limited authority

Proposal v0.1

Publish which body may decide, what checks it must pass, and where an appeal goes. Use the assignments below as a starting point.

Purpose

Prevent a body from choosing its own powers or skipping review when a decision becomes difficult.

Tradeoff

Boundaries can overlap. Several required checks may slow a decision or leave bodies disagreeing about who should act.

Open question

What exact thresholds, jurisdictions, composition rules, and procedures would resolve a dispute between bodies?

Changing rules without handing over control

Proposal v0.1

Require a higher approval threshold and a waiting period for changes to protected rights. Require several stages and approval across defined groups for constitutional amendments.

Purpose

Give people time to inspect changes and prevent one bloc from rewriting the system for itself.

Tradeoff

Strong safeguards can make needed repairs hard or let a minority block progress. A vote alone cannot settle every legal or rights question.

Open question

Which rights must remain beyond amendment, and how should cross-group approval balance equal standing with protection from domination?

Ratification means formal approval by the eligible constituent body. Proposed thresholds are not yet adopted. See the constitution amendment walkthrough for an example of the checks.

A published process for every major decision

Proposal v0.1

Use this ten-step lifecycle, with the responsible body and public records named at each stage.

Purpose

Make it possible to see how an idea became a decision, and where to question it.

Tradeoff

A complete process takes effort. Urgent or routine decisions may need shorter paths with clear limits.

Open question

Which steps are required at each level of risk, and how can the process stay usable rather than become red tape?

AI helps; people keep authority

Proposal v0.1

Allow AI to summarize with traceable sources, translate, flag contradictions, compare scenarios, develop opposing cases, and help check that procedures were followed.

Purpose

Help people understand more evidence without letting a tool become a ruler.

Tradeoff

A helpful summary can still omit a view or get a fact wrong. Using several models does not remove shared blind spots.

Open question

What independent tests, human review, and correction process are enough before each use?

AI may not guess unexpressed consent, hide serious dissent, certify its own work, cast votes, or make sovereign decisions. Every important summary needs source links, visible uncertainty, and a way to challenge it. Accountable humans must authorize and remain responsible for lawful decisions. Agents may carry out bounded administrative tasks under their supervision.

An administration accountable to the people it serves

Proposal v0.1

Give each agent a limited job, explicit permissions, a named human supervisor, and records that can be inspected. Its instructions follow the people's adopted rules, not a model provider's business priorities.

Purpose

Test whether routine administration can be easier to use, more responsive, and less costly without losing accountability.

Tradeoff

Automation can speed up a bad rule or repeat an error at scale. Model providers, training, and software design can carry biased or misaligned incentives. Logs and oversight also have privacy and cost tradeoffs.

Open question

Which tasks can agents handle reliably, and what evidence would justify each permission?

• People set the job. Publish its scope, who is responsible, what tools it may use, and where an appeal goes.
• Bound the work. Agents cannot grant themselves permissions, infer consent, cast votes, authorize spending, transfer funds, or certify their own work. Humans retain authority over rights, money, policy, and appeals.
• Keep a way to intervene. People can inspect, correct, suspend, or replace an agent. Protect private information when publishing records.
• Test the benefit. Compare completion time, total cost including human review, error rates, access, and unresolved complaints. Faster alone does not mean fairer.

Start with voluntary, bounded trials before considering more responsibility. No administrative agent service is operating through this site. Adoption in public administration would require lawful authority and accountable public decisions; the Commonwealth has no power to impose it.