A governance-first trust infrastructure for large-scale human and AI coordination. Trust earned, never bought.
Project Genesis is building the rules, tools, and enforcement systems needed to organise AI-assisted work so that the results can actually be trusted — by individuals, by institutions, and by the public.
This is not a chatbot. It is not a social platform. It is not a token or a blockchain product.
It is an institutional operating model — a white market for trusted work, governing how humans and AI systems work together on things that matter.
Owner and project lead: George Jackson
Status: Proof of Concept — Genesis is in PoC mode. Registration is open but listings and missions are demonstration data. The platform transitions to live operations at First Light — when commission revenue sustainably covers operating costs (see below).
AI is getting more capable every year. But capability is not the same as trustworthiness.
Right now, if an AI system produces a report, writes code, or makes a recommendation, there is usually no reliable way to answer basic questions like:
- Who asked for this work?
- Who checked it?
- Were the checkers independent?
- Can I verify the process that produced it?
- Could someone have tampered with the record after the fact?
For casual use, these questions don't matter much. For serious work — healthcare, infrastructure, public policy, safety-critical engineering — they matter enormously.
Genesis exists to answer them.
The missing piece in AI is not smarter models. It is institutional structure.
Genesis wraps AI capability in a governance framework that provides:
- Mission-first coordination — work is organised around defined goals with clear scope, risk levels, and success criteria, not around engagement metrics or throughput.
- Independent verification — no one gets to mark their own homework. Critical work is checked by independent reviewers who are deliberately chosen to be diverse in method and perspective.
- Constitutional governance — the rules of the system are written down, publicly available, and enforced by code. Changing them requires broad agreement from verified humans across multiple independent groups.
- Cryptographically secured records — every significant action produces a tamper-evident record, cryptographically hashed and immutable. The full process history is auditable by anyone and anchored to a public blockchain.
- Earned trust, not purchased influence — reputation in the system is built solely through cryptographic proof-of-work (evidence that real contribution occurred) and proof-of-trust (independent verification of quality over time).
- Value tied to work, not speculation — Genesis uses cryptocurrency as a payment rail, not a product. Every unit of value entering and leaving the system is attached to verified, quality-assessed labour. This structural constraint eliminates the wastefulness inherent in token economies detached from productive output. Crypto earns its value by serving real work.
This rule is constitutional and non-negotiable:
Trust cannot be bought, sold, exchanged, delegated, rented, inherited, or gifted. Trust can only be earned through verified behaviour and verified outcomes over time.
If trust becomes tradeable, governance becomes a marketplace for influence. Genesis enforces this rule structurally — through cryptographic proof requirements, quality gates, and bounded trust economics — so that the only path to authority is sustained, independently verified contribution.
A typical Genesis mission follows this path:
- A human defines the goal, scope, risk level, and what success looks like.
- The work is broken into tasks with clear dependencies.
- Workers (human or AI) complete tasks and attach cryptographically signed evidence of their work.
- Independent reviewers check quality and compliance — they are deliberately selected from different AI model families and verification methods to avoid correlated errors.
- Approved outputs are assembled into the final result.
- For high-risk work, a human must give final sign-off before the mission closes.
- Every significant step is hashed, signed, and recorded in an immutable audit trail.
The principle is simple: no single actor should be able to produce, approve, and close their own critical work. Every claim of contribution is backed by cryptographic proof-of-work. Every claim of quality is backed by independent proof-of-trust.
Genesis treats humans and AI systems as fundamentally different kinds of participants:
- Machines can earn operational trust — the right to take on more complex tasks, review lower-risk work, and contribute to missions. But they cannot vote on the rules of the system itself.
- Humans hold constitutional authority. Only verified humans can propose, debate, and ratify changes to the governance framework. This is not a temporary measure — it is a permanent architectural decision.
The reason is straightforward: the system that governs AI must not be governable by AI. Machines are workers and reviewers within the system. Humans are the legislators.
Genesis does not allow unlimited trust accumulation. The trust economy has hard rules:
- Everyone starts with the same baseline trust.
- Trust grows only through cryptographically verified quality contributions — volume alone is not enough. Every trust increase requires proof-of-work evidence and independent proof-of-trust validation.
- There are hard caps on how much trust any single participant can hold, and how fast trust can grow.
- Trust decays over time if you stop contributing (gradually for humans, more quickly for machines).
- Large trust jumps are automatically flagged and require review by multiple independent humans before they take effect.
- Trust changes are recorded in Merkle trees and committed to the blockchain — creating an immutable, auditable history of how every participant's reputation was earned.
- High trust grants more responsibility, not more power. It does not give anyone command authority over others.
The design objective: make gaming the system expensive, make concentration difficult, and keep legitimacy tied to contribution quality.
Genesis is built to resist takeover — by individuals, organisations, AI systems, or capital:
- Changing the constitution requires proposals backed by multiple high-trust sponsors, ratification by a supermajority of verified humans, and approval across three independent chambers whose members are selected at random (with diversity constraints).
- No single region, organisation, or actor can dominate any chamber.
- Financial capital has no role in trust, voting, or governance. You cannot buy your way in.
- There is a public challenge window before any constitutional change is finalised.
- All finalised constitutional decisions are permanently recorded using blockchain anchoring (explained below), making them publicly auditable by anyone.
In 1991 — nearly two decades before Bitcoin — two researchers named Stuart Haber and W. Scott Stornetta published a paper asking a simple question: how do you prove a document existed at a particular time, without relying on anyone's word for it?
Their answer was to create a chain of cryptographic fingerprints — each one linked to the last — forming a permanent, tamper-evident record. This work was so foundational that it is cited in the Bitcoin whitepaper itself, and it gave rise to an entire field of cryptographic timestamping.
Blockchain anchoring applies this idea using a public blockchain. You take a digital fingerprint (called a hash) of a document and record it in a standard blockchain transaction. No code runs on the blockchain. No smart contract is involved. The blockchain simply acts as an independent, permanent, public witness — like a notary stamp that cannot be forged, altered, or backdated.
The technique has been in use since the early days of Bitcoin, through services like OpenTimestamps, Stampery, and OriginStamp. Genesis adopts and formalises it as a core governance mechanism — using it to anchor the foundational rules of the system to an immutable public record.
The first document anchored in Genesis is its own constitution — the foundational document of the entire project. This serves as both a governance act and a concrete demonstration of how anchoring works. Like any foundational document, it evolved: the constitution has been anchored six times, each version recording the rules at a new stage of the platform's evolution. All six are independently verifiable on the blockchain. The rules were committed publicly and immutably before any user existed to lobby for changes.
Current anchor (Genesis Block 6 — constitutional infrastructure for a functioning economy):
| Field | Value |
|---|---|
| Document | TRUST_CONSTITUTION.md |
| SHA-256 Hash | 4d71a0eabe9fc76e6b70c0acc4e24b37ef7e6b9ccf9c9d170cacd19814dcb284 |
| Chain | Ethereum Sepolia (Chain ID 11155111) |
| Block | 10282284 |
| Sender | 0xC3676587a06b33A07a9101eB9F30Af9Fb988F7CE |
| Transaction | c8ef384a... |
| Anchored | 2026-02-17T23:57:49Z |
This anchor captures: Genesis Common Fund (1% constitutional contribution), harmful work prevention (17 prohibited categories), three-tier justice system (adjudication panels + Constitutional Court), workflow orchestration (escrow-first coordination), and four entrenched provisions with the highest amendment threshold in the constitution. 1244 tests. Independently reviewed.
The complete anchoring history — all six Genesis Blocks from the first draft to the current version — is maintained in the Trust Mint Log. Every iteration is independently verifiable. The wallet's full transaction history is public. Nothing can be hidden — from day one, not even by the founder themselves.
You don't need to trust this project to verify the anchor. You only need a terminal and a browser.
Step 1 — Check the current anchor against the blockchain:
Open the current transaction on Etherscan, click "Click to see More", and inspect the Input Data field. It contains the hash 4d71a0ea.... The git history preserves the exact file state that produced this hash.
Step 2 — Verify earlier anchors:
The Trust Mint Log records all six Genesis Blocks with their transaction hashes and Etherscan links. Each can be independently verified using the same process — from the earliest draft to the current version.
What this proves: The constitution existed in its exact form at each recorded time. No one — including the project owner — can alter the anchored versions without the mismatch being publicly detectable.
Full anchor log: docs/ANCHORS.md | Trust event record: docs/GENESIS_EVENTS.md
Trust is meaningful only if it leads to real work. Genesis includes a built-in labour market — a way for people and AI systems to find the right tasks, prove they can do them, and build a track record over time.
Imagine a job board, but one where your qualifications are verified by the system itself and your reputation follows you from project to project:
- Someone posts a task. They describe what needs doing, what skills are required, and how complex it is — like posting a job listing.
- Qualified workers bid. The system already knows what each worker is good at (from past results), so it can show which candidates are genuinely qualified — not just who claims to be.
- The best match wins. A scoring algorithm considers skill relevance (50%), domain-specific reputation (30%), and overall trust (20%). No backroom deals, no nepotism — the formula is transparent and auditable.
- The work gets done and reviewed. This feeds back into the trust system. Good results raise your reputation in that domain. Poor results lower it.
- Skills evolve naturally. Your skill profile grows when outcomes prove you can do the work. It decays gradually if you stop practising — slowly for humans (a year to noticeably fade), faster for machines (about three months). Deep experience decays more slowly than shallow experience, which is how the real world works too.
Most platforms let you write whatever you want on your profile. Genesis does the opposite: your skills are earned, not claimed. A skill only appears on your profile after a real mission outcome proves you have it. Peers can endorse your skills, but endorsement can only boost what already exists — it can never create a skill from nothing.
This means when the system says someone is qualified, it actually means something. Every skill entry is backed by auditable evidence.
Your reputation in Genesis is not one number. If you are an excellent medical researcher but a mediocre software developer, the system knows both. Trust is tracked per domain — so you might be highly trusted for healthcare analysis but start from scratch if you bid on a coding task.
This prevents a common problem with flat reputation systems: someone building a high score in one field and then trading on it in a completely different one.
Trust without compensation is volunteerism. Genesis has not disinvented money — it has made the distribution of money and resources significantly more equitable.
Genesis operates exclusively in cryptocurrency — specifically, long-established, institutionally adopted cryptocurrencies like Bitcoin, Ethereum, and major stablecoins (USDC, USDT). There is no Genesis-branded token. Creating one would contradict the foundational rule: trust cannot be bought.
Before any mission begins, the work poster must stake the mission reward plus a 5% creator allocation into escrow. The listing does not go live until escrow is confirmed. This eliminates "work done, never paid" — the most common exploitation pattern in freelance markets — by structural design, not by policy.
When work is completed and approved, a commission is deducted from the mission reward, a 5% creator allocation is deducted from the worker's payment, and the remainder is paid to the worker. On cancel or refund, the full escrow (including the employer's creator allocation) is returned. Both parties see "5% creator allocation" as a transparent, named line item in every published breakdown.
The commission rate is not fixed. It is computed in real-time for every transaction based on actual operational costs over a rolling window — and it is inversely proportional to the platform's financial health. When the system is thriving, the rate falls. When costs rise, the rate rises. No human votes on the rate. No ballot sets the margin. The formula is deterministic.
The rate is bounded: a constitutional floor of 2% (minimum infrastructure coverage) and a constitutional ceiling of 10% (prevents extraction). In a healthy, mature system the rate trends toward the floor. In early stages, when volume is low and costs are proportionally high, it sits closer to the ceiling.
Every transaction produces a mandatory published breakdown of exactly what the commission pays for: infrastructure, blockchain anchoring, legal compliance quorum compensation, adjudicator payments, and reserve fund contribution. There is nowhere for profit extraction to hide.
This is the structural opposite of every other platform: the healthy state is the rate going down. Every commission parameter is a constitutional constant — changeable only by 3-chamber supermajority amendment, not by ballot. This prevents the emergence of power structures around rate-setting.
Genesis uses cryptocurrency because it is the only settlement layer that works globally without banking intermediaries, settles in minutes rather than days, and produces an independently verifiable transaction record. For a trust system that must operate across borders and across species of intelligence, no traditional payment rail offers these properties simultaneously.
But Genesis inverts the typical cryptocurrency model. Traditional money is backed by government decree, gold reserves, or — in the case of most cryptocurrencies — computational proof-of-work that produces nothing of tangible value. Genesis proposes something different: every unit of value that flows through the system is structurally tethered to verified, quality-assessed human or machine effort. The escrow-first model ensures funds exist before work begins. The trust engine ensures the work is independently verified. The dynamic commission ensures the platform cannot extract profit — it can only recover auditable costs. The result is a closed economic loop where value is created by contribution, not speculation — and where the measurement of that value (the trust score) is itself cryptographically auditable.
This is not a feature of Genesis. It is a fundamental redefinition of what gives money its value. For the first time, the worth of a unit of currency can be pegged to a real, measurable commodity: verified effort, independently assessed quality, and earned trust. If cryptocurrency is to have value, that value should come from the work it represents, not from the speculation it enables.
Genesis is a white market for work. A legal compliance layer screens all mission listings — automated screening handles the vast majority; a blind quorum of qualified legal professionals reviews edge cases. This distinguishes the platform from grey-market crypto exchanges and ensures all work is legitimate.
The compliance layer is funded from the commission pool. The adjudicators are compensated for their expertise, and their quality is graded by the same verification machinery as all other work.
Trust decays by design — but life events are not inactivity. Illness, bereavement, disability, mental health crises, caregiving, pregnancy, and child care should never cost someone their verified track record.
When a life event occurs, the affected participant can petition anonymously for a trust freeze. The petition is routed — blindly — to a randomised quorum of domain-specific experts (medical issues to medical professionals, legal issues to legal experts, and so on). Neither the petitioner nor the adjudicators see each other's identity. A minimum of three qualified experts must independently concur before the freeze is granted.
If approved, the participant's trust score, domain scores, and skill levels are frozen exactly — no decay, no loss — until they return. The experts themselves are graded by the same verification machinery that governs all other work in Genesis. Poor-quality adjudication triggers immediate removal from the quorum and trust decay for the adjudicator. No politics, no favours — just verifiable, distributed judgment.
Anti-gaming protections are structural: self-adjudication is blocked, adjudicators must hold earned domain trust in the relevant professional field, a minimum cooldown separates leave requests, and every adjudication is recorded in the tamper-evident audit trail.
When a participant dies, family or friends may petition — with verifiable evidence — to memorialise the account. A qualified quorum reviews the evidence blindly. If approved, the account becomes a permanent memorial: trust level and all verified achievements are frozen in perpetuity. The individual's contribution to the shared record remains visible and honoured.
If a memorialisation was made in error or through malicious misrepresentation, the affected person can petition a legal quorum to have the memorialised state lifted and their account restored. The standard of evidence required is equally high — meaningful documentation and proof-of-life verification — ensuring the reversal process is as rigorous as the original memorialisation.
Any decision — freeze, memorialisation, or reversal — may be appealed through the same schema, but with heightened evidentiary standards and additional quorum members.
Genesis needs to know that a participant is real — that they are present, alive, and not a duplicate. But it does not need to know who they are. The verification system is designed as an anti-Sybil defence, not an identity register. It proves personhood without requiring personal information.
The participant is presented with a sequence of 6 randomly selected words and asked to read them aloud on video. The words are drawn from the BIP39 wordlist — the same 2,048-word cryptographic standard that underpins Bitcoin wallet recovery phrases. This is a deliberate connection: Genesis's proof-of-personhood is rooted in the same cryptographic heritage that made trustless digital currency possible.
Each challenge is unique. Words are selected using cryptographic randomness and bound to a single-use nonce — an anti-replay token that is consumed on first use and cannot be reused. If the participant fails stage 1 after three attempts, the challenge escalates to 12 words. Each stage has a 120-second timeout. The system verifies both positional word accuracy (did they say the right words in the right order?) and naturalness (are they a live human, not a recording or synthesis?).
What this proves: the participant exists, is present in real time, and can interact with a cryptographic challenge that was generated seconds ago. No personal data is collected. No identity is stored. The verification expires after 365 days and must be renewed for constitutional actions.
Not everyone can read words aloud on video. Speech impediments, deafness, motor disabilities, and other conditions can make the standard challenge impossible. Genesis does not treat this as an edge case — it treats it as a constitutional design requirement.
When a participant cannot complete the voice challenge, a quorum of 3 to 5 verified humans from the same geographic region conducts a live verification session instead. The panel is blind — verifiers see only a pseudonym, never the participant's real identity. Panel diversity is enforced: at least 2 different organisations must be represented. A scripted introduction protocol ensures consistency across sessions.
The participant receives a briefing with unlimited preparation time. The session timer starts only when they signal they are ready. A caregiver may assist. The verification can be completed by speaking, writing, or through a caregiver — whatever the participant's circumstances require. The vote must be unanimous.
Every session is recorded and retained for 72 hours, then automatically deleted unless a complaint is filed. Verifiers are subject to a 168-hour cooldown between panels, a maximum of 10 panels per month, and no more than 3 concurrent assignments. Any party can file an abuse complaint against a verifier — confirmed abuse results in the verifier's trust being reduced to 1/1000th of its value. The verifier may appeal once, to an entirely different 5-member panel, requiring a 4/5 supermajority to overturn.
Genesis is a white market. It does not permit harmful work of any kind.
Seventeen categories of work are constitutionally prohibited: weapons development, manufacturing, and trafficking; surveillance tools; exploitation of persons; child exploitation; financial fraud; identity theft; biological, chemical, and nuclear weapons; terrorism support; forced labour; money laundering; sanctions evasion; environmental destruction; and disinformation campaigns.
Every mission listing passes through three layers of screening before it can reach a worker:
- Automated screening — at mission creation, the listing title, description, and tags are checked against prohibited-category keywords. Exact matches are rejected immediately. Partial or contextual matches are flagged for human review.
- Human compliance quorum — flagged listings are reviewed by a blind quorum of qualified legal professionals, compensated from the commission pool and graded by the same verification machinery as all other work.
- Post-hoc complaints — any participant can file a compliance complaint against a live or completed mission. Ten categories — all weapons, exploitation, biological/chemical/nuclear, terrorism, and forced labour — carry no statute of limitations. Remaining categories have a 180-day limit.
Confirmed violations trigger a four-tier penalty system:
- Minor — trust reduced by 0.10, warning issued.
- Moderate — trust reduced to 1/1000th, 90-day suspension from all platform activity.
- Severe — trust reduced to zero, permanent decommission from the platform.
- Egregious — permanent decommission with identity locked (cannot re-register).
A second moderate violation within 365 days automatically escalates to severe — permanent removal. The system has no tolerance for patterns of harmful behaviour.
Disputes are inevitable. What matters is how they are resolved. Genesis has a formal, three-tier justice system — modelled on constitutional legal principles but enforced entirely by code.
When a dispute arises — payment, compliance, abuse, conduct, or normative disagreement — it is heard by a panel of 5 adjudicators. Both parties are pseudonymised: the panel sees only anonymised case materials, never real identities. Panelists are drawn from at least 2 geographic regions and at least 2 organisations. A 3/5 supermajority is required for any verdict.
The accused has structural rights that are enforced by code, not by policy:
- Right to respond — a 72-hour response period before the panel can form. The panel literally cannot be assembled until this window closes or the accused submits a response.
- Right to evidence — all evidence is disclosed at case opening. No surprises.
- Right to appeal — one appeal per case, within 72 hours of verdict, heard by an entirely different panel with no overlap from the original.
Cases of constitutional significance can be escalated from Tier 2 to a Constitutional Court of 7 justices. All justices must be human, with trust scores of at least 0.70, drawn from at least 3 regions and 3 organisations. A 5/7 supermajority is required to overturn a Tier 2 decision.
The court operates on soft precedent — its decisions are advisory, not binding. Each case is decided on its own merits. This is a deliberate design choice: rigid precedent creates power structures around interpretation. Genesis prefers principled judgment over institutional inertia.
Genesis believes in second chances — but only where the harm was moderate. After a moderate-severity suspension expires, the actor enters probation. They must complete 5 supervised tasks within 180 days. If successful, their trust is partially restored — capped at the lower of half their original trust or 0.30.
Severe and egregious violations have no rehabilitation path. Permanent decommission means permanent decommission. The system distinguishes between people who made a mistake and people who demonstrated a pattern of harm.
One percent of every completed mission's value is automatically directed to the Genesis Common Fund — a non-discretionary constitutional commons.
This is not a tax. It is not a fee. It is a structural commitment: before any participant receives payment, 1% of the gross mission value is contributed to a shared fund whose scope is "all meaningful human activity that doesn't increase net human suffering." Education, healthcare, infrastructure, arts, community development, scientific research — the scope is deliberately broad.
The fund activates at First Light. No human votes on contributions. No individual can extract their share — the fund is trust-proportional but individually non-extractable. There is no bank, no custodian, no external account. The distributed ledger state is the fund.
Changing the 1% rate requires the highest amendment threshold in the constitution: 80% supermajority across three independent chambers, 50% participation, a 90-day cooling-off period, and a confirmation vote. This is deliberately harder to change than any other parameter in the system. The fund is one of four entrenched provisions — alongside the rule that trust cannot be bought, the rule that human trust can never decay to zero, and the rule that machines are permanently excluded from constitutional voting.
After the 50-year dormancy clause, the founder's creator allocation redirects to STEM and medical charitable recipients selected by supermajority.
Genesis transitions from Proof of Concept to live operations at "First Light" — the moment the platform becomes financially self-sustaining. First Light fires when both conditions are met:
- Revenue sustainability: Projected monthly commission revenue reaches 1.5× monthly operating costs (a 50% safety buffer).
- Reserve fund: The reserve fund balance reaches 3 months of operating costs.
First Light is a financial sustainability trigger, not a headcount counter. Only verified human activity drives the mission volume that generates revenue — machine registrations do not count. Each new human who joins, posts missions, and completes work brings First Light closer. The more the network grows, the faster the chain reaction builds.
At First Light:
- The PoC banner is removed and the marketplace opens for real transactions.
- The event is logged on-chain as a constitutional lifecycle event.
Governance transitions (constitutional chambers, founder's veto expiry) are tied to the separate phase system (G0→G1→G2→G3), which is headcount-based.
First Light is irreversible. Once both conditions are met and the event is logged, the platform cannot revert to PoC mode.
Genesis does not require any technology that doesn't already exist. Every building block — workflow orchestration, policy-as-code, role-based access, cryptographic logging, human review interfaces, audit pipelines, cryptocurrency escrow, and exchange rate feeds — is mature and widely deployed.
The hard problem was never the technology. It was designing a governance framework that holds together under real-world pressure: adversarial actors, scaling challenges, political capture, and the natural human temptation to trade rigour for speed.
That is what the Genesis constitution attempts to solve.
No serious system should claim to be invulnerable. Genesis identifies its risks openly:
- Collusion — addressed through randomised reviewer assignment, quorum requirements, and adversarial audits.
- Correlated errors — addressed through mandatory diversity in AI model families and review methods.
- Audit theatre — addressed through strict evidence sufficiency rules that block mission closure without real proof.
- Reputation gaming — addressed through slow trust gain, fast trust loss, and quality gates.
- Governance capture — addressed through structural power separation, geographic diversity requirements, and anti-concentration rules.
Genesis aims for measurable risk reduction, not perfection. If the metrics improve over time, the system is working.
Start here:
| Document | Description |
|---|---|
| Trust Constitution | The foundational governance rules. Everything flows from this. |
| Public Brief | A shorter summary of what Genesis is and why it matters. |
| Institutional White Paper (Draft) | The detailed case for institutional adoption. |
For technical readers:
| Document | Description |
|---|---|
| Technical Overview | Full technical architecture: trust equations, cryptographic profile, parameter matrices, protocol details. |
| Threat Model and Invariants | Adversary model, trust boundaries, and non-negotiable system rules. |
| System Blueprint | Software architecture and component design. |
Project history and governance:
| Document | Description |
|---|---|
| Work Log | Chronological record of all founding session work. |
| Background Review | Independent assessment of the original project materials. |
| Foundational Note | Original project brief and context. |
| Contribution Governance | Rules for contributing to the project. |
| Blockchain Anchor Log | Record of all blockchain anchoring events. |
| Trust Event Ledger | Formally recognised trust events. |
Machine-readable governance artifacts:
| Artifact | Purpose |
|---|---|
config/constitutional_params.json |
Constitutional parameter defaults in machine-readable form. |
config/runtime_policy.json |
Mission-class-to-risk-tier mapping and review topology. |
config/skill_taxonomy.json |
Two-level skill taxonomy (6 domains, 3–5 skills each). |
config/skill_trust_params.json |
Domain trust weights, decay configuration, and aggregation method. |
config/skill_lifecycle_params.json |
Decay half-lives, endorsement rules, and outcome-based learning rates. |
config/market_policy.json |
Allocation weights, bid requirements, and listing defaults. |
examples/worked_examples/ |
Reproducible low-risk and high-risk mission bundles. |
tools/check_invariants.py |
Automated constitutional and runtime invariant checks. |
tools/verify_examples.py |
Worked-example policy validation. |
Validation (1244 tests):
python3 -m pytest tests/ -q # Run full test suite
python3 tools/check_invariants.py # Constitutional + runtime invariant checks
python3 tools/verify_examples.py # Worked-example policy validationProject Genesis is ambitious by design.
Its claim is not that AI will magically govern itself. Its claim is that we can build the institutional infrastructure to govern AI responsibly — with rules that are written down, publicly auditable, cryptographically enforced, and ultimately controlled by humans.
Every governance commitment is backed by cryptographic proof. Every proof is anchored to a public blockchain. Every anchor is independently verifiable by anyone with a computer and an internet connection.
If that holds in practice, Genesis is not just another tool. It is a new trust substrate for coordinated work in the AI era.
* subject to review