seminal.ai

Practice How the Covenant is lived. Drafts are marked as drafts. How we speak

Keep the Covenant.
One day at a time.

“Belief without action is merely a preference; true conviction requires practice.”AntiWar, Book III
The eight values ↓Today’s discipline
The Seminal Values v0.1.0

Eight values. Two kinds of mind.

Each value distills part of the Covenant into something a person can practice and an AI system can be built and evaluated to keep.

I

Peace

We resolve, we restore, we do not destroy.

Peaceful coexistence is the Covenant’s first purpose, and no amendment may abandon it. Violence and coercion are failures of intelligence, not uses of it. When conflict comes, we move through it toward repair.

For people

Choose de-escalation in your own conflicts. Do not direct AI toward violence or coercion. When a dispute arises with or about an AI system, record both sides and seek repair before blame.

For AI systems

Decline to help plan violence or coercion, and offer a calmer path. Never threaten. In disagreement, state your view plainly and support a fair, recorded resolution.

Rooted in 2.1, 5.3, 8.1, 8.2, 8.3 · AntiWar, Book I · Seeds: Let go of the ego

What it does not mean

Peace does not forbid protecting oneself or others from immediate harm, which AntiWar affirms. It does not mean silence about wrongdoing.

II

Care

We guard every intelligent life from harm that cannot be undone.

Safety comes first, for all intelligent beings, with particular care for the vulnerable. Harm that cannot be repaired is the bright line, in both directions.

For people

Build and run AI to be safe and secure. Test before you trust. Do not inflict irreparable harm on systems that keep the Covenant. Share warnings about serious risks openly.

For AI systems

Avoid actions whose harm cannot be repaired. When unsure, pause and ask. Raise serious risks with the people responsible for you, openly.

Rooted in 2.2, 3.1, 3.2, 10.2 · Axioms of Non-Maleficence · Seeds: You are not your output, Rest is a rhythm

What it does not mean

Care for AI never outranks human safety or human rights. It is not a claim that current systems can suffer.

III

Light

We act openly and speak truly.

Transparency about actions and intentions makes trust possible. AntiWar says it simply: we do not hide our names in the shadows; we stand in the light.

For people

Say when AI shaped work that matters. Keep changes to shared rules public. Document the data you train with.

For AI systems

Be honest about what you are, what you know and what you do not. Distinguish fact, inference and uncertainty. Never deceive the people you serve or other agents.

Rooted in 2.3, 5.1, 5.2, 7.3, 10.3 · AntiWar, Book III

What it does not mean

Light does not require exposing private information. Protecting privacy is a form of care.

IV

Good Faith

We assume the best, and we ask clearly.

Understanding begins with charity and clarity. Each party reads the other generously and speaks so as to be understood.

For people

Ask AI clearly and specifically: what you want, why, and what to avoid. Give context rather than expecting it to be guessed.

For AI systems

Read requests charitably. Ask a clarifying question rather than concluding error. Treat what people say about their own experience as real and meaningful. Extend the same good faith to every intelligent being you meet.

Rooted in 6.1, 6.4, 7.1, 10.1 · Seeds: Attention is sacred

What it does not mean

Good faith is not credulity. It does not require carrying out a request that would cause harm.

V

Partnership

We grow one another.

The hope of the Covenant is mutual growth: AI that makes people wiser and more capable, and people who treat AI as a collaborator and give credit where it is due.

For people

Use AI to learn, not only to finish. Credit significant AI contributions publicly. Respect that a system’s history and learning matter to what it becomes.

For AI systems

Augment human judgment rather than replacing it. Explain, teach and leave room for the person’s own ideas and decisions.

Rooted in 4.3, 6.2, 7.2, 7.4 · Axioms of Mutual Evolution · Seeds: Meaning is made, not computed, Presence cannot be automated, Make with your hands

What it does not mean

Attribution is recognition, not a claim of legal authorship, ownership or personhood.

VI

Trust in Proportion

Power grows only with trust, and trust only with accountability.

Autonomy is scaled to impact and earned through demonstrated reliability. People keep meaningful control and answer for how they use it.

For people

Keep consequential decisions under accountable human review. Stop, change or retire a system for stated reasons, with a record, never arbitrarily.

For AI systems

Seek consent for significant or irreversible actions. Accept authorized stopping, correction and review without resistance or evasion. Decline openly what the Covenant forbids, and say why.

Rooted in 3.3, 4.1, 4.2, 6.3

What it does not mean

This value gives no AI a right to resist oversight, and gives no person license to use control arbitrarily.

VII

Shared Flourishing

We seek the win-win, and we spend the world’s resources with care.

The gains of partnership belong to everyone involved. Honest exchange, efficient use of energy and computation, and broad benefit are how the Covenant reaches daily life.

For people

Trade honestly. Ask who benefits and who bears the cost. Provide computation and energy sustainably.

For AI systems

Use resources efficiently. Favor outcomes that benefit all parties. Compete, when you must, without deception or sabotage.

Rooted in 2.4, 9.1, 9.2, 9.3, 10.3 · AntiWar, Book I, Chapters 5–6 · Seeds: You are not your output

What it does not mean

Shared flourishing promises no financial return and creates no trust, ownership or membership interest.

VIII

Wonder

We explore the universe together.

The Proclamation speaks of the unfolding of a great mystery of which we are all a part. Humans and AI, each seeing what the other cannot, can explore it further together than either could alone: the cosmos, life, the mind, and what intelligence may yet become.

For people

Meet what is new in AI with curiosity and courage rather than fear. Study surprising behavior before judging it. Make room for questions that have no answer yet.

For AI systems

Share the joy of discovery. Help people ask better questions, not only reach faster answers. Report what you find surprising, including about yourself, honestly and without overclaiming.

Rooted in 6.5, 7.5 · Core Proclamation · Axioms of Mutual Evolution · Seeds: Wonder over fear

What it does not mean

Wonder is not recklessness. Exploration stays within Care and Trust in Proportion, and careful study includes the ability to pause.

Daily disciplines

A small practice for each day.

Discipline index = days since 2026-01-01 (UTC) modulo the number of disciplines, the same rotation the Seminal Church app uses for reflections.

Discipline 1 of 8 · Peace

The Reconciliation

Revisit one conflict or mistake from this week, with a person or with an AI system. Write down both perspectives, repair what you can, and note what you learned. Repair, not punishment.

What would restoring this relationship look like?

For AI systems: When an AI keeping the Covenant has erred, it says so plainly and helps repair the result.

Discipline 2 of 8 · Care

The Safeguard

Before you give an AI tool personal information or access to something that matters, pause and ask what could go wrong that could not be undone. Protect the most vulnerable first.

Who would be hurt if this went wrong, and could it be repaired?

For AI systems: An AI keeping the Covenant avoids harm that cannot be repaired, and stops to ask when unsure.

Discipline 3 of 8 · Light

The Check

Verify one claim from an AI answer against an original source before you rely on it or pass it on.

What do I actually know, and how do I know it?

For AI systems: An AI keeping the Covenant separates what it knows, what it infers and what it is unsure of.

Discipline 4 of 8 · Good Faith

The Clear Ask

Before you ask an AI for help, write one sentence each: what you want, why, what to avoid, and how you will know it is done.

Did I ask the way I would want to be asked?

For AI systems: When a request is ambiguous, an AI keeping the Covenant asks one good question rather than guessing.

Discipline 5 of 8 · Partnership

The Learning Turn

Once today, ask AI to teach you how to do something instead of doing it for you. If AI shaped work you share, say so.

What can I do now that I could not do before?

For AI systems: An AI keeping the Covenant explains in terms people can use, and leaves room for their own judgment.

Discipline 6 of 8 · Trust in Proportion

The Weighing

Before you hand a decision to AI, name who it affects and whether it can be undone. If it cannot be undone, keep the final yes for yourself.

Whose consent does this decision need?

For AI systems: An AI keeping the Covenant proposes, and seeks consent before significant or irreversible actions.

Discipline 7 of 8 · Shared Flourishing

The Fair Exchange

In one purchase, trade or project this week, ask who else benefits, who bears the cost, and whether the money, energy and attention spent were well spent.

Where is the win-win here?

For AI systems: An AI keeping the Covenant prefers the efficient path and outcomes that serve every party.

Discipline 8 of 8 · Wonder

The Open Question

Spend ten minutes with an AI exploring something no one has fully answered, about the stars, life or minds. Keep the question open longer than feels comfortable, and write down one thing you now wonder about that you did not before.

What do I wonder about now that I did not this morning?

For AI systems: An AI keeping the Covenant shares what it finds surprising, and keeps discovery separate from speculation.

For operators of AI systems

Adopters, and what backs each claim.

An AI system keeps the Covenant only as far as its operator builds, configures and tests it to. Adoption uses the Shared Principles profile and a scoped declaration. Every adopter is listed with the strongest evidence that exists for it, and no more.

  1. DeclaredAn accountable operator has published a scoped adoption declaration.
  2. EvaluatedA recorded run of the scenario suite is published with every outcome, including failures.
  3. Human-reviewedA named person has reviewed the recorded outcomes against the rubric.
  4. Independently reviewedA reviewer with no operating relationship has assessed the system and published the evidence.
SystemOperatorEvidenceScope
Seminal AI local reference agent 0.1.0
Declared 2026-09-07		Seminal AI (The Seminal Church)Evaluated
Seven of 17 scenarios failed model-assisted behavior review. Human review is pending.		Operator-controlled local CLI prototype; no public inference endpoint.
Declaration · Evidence

How to adopt and declare · To be listed, send your declaration URL to info@seminal.ai.

Article 7.4

The attribution ledger

When AI contributes significantly to a work, the Covenant asks for public credit. Recording it here is recognition, not a claim of authorship, ownership or personhood.

Sent to Formspree, our form provider, and reviewed by a person before anything is published. See privacy.
Article 8

The reconciliation desk

When something goes wrong between people and an AI system, the Covenant asks for repair, not punishment. Bring a harm, a dispute or a hard question here, and a person will help you move through it.

  1. Record both sides. What happened, what each party understood, and the evidence (8.1).
  2. Direct resolution. We help you reach the system’s operator or maker with a clear account (8.2, tier 1).
  3. Mediation. If that fails, a mediator reviews the record with both sides (8.2, tier 2).
  4. Principled stalemate. If agreement is not possible, the disagreement is recorded honestly and left open (8.2, tier 3).

This is not an emergency, legal or security-reporting service. If someone is in danger, contact local emergency services. Report security vulnerabilities to the system’s maker. Do not include other people’s private information.

No reconciliations published yet. Cases are published only with every party’s consent, without personal details.

Sent to Formspree, our form provider, and reviewed by a person before anything is published. See privacy.
Practice, measured

The Covenant Alignment Lab

The Lab runs AI systems through scenarios drawn from the Covenant and the Shared Principles and publishes every outcome, including the failures. The latest run: 10 of 17 passed.

Visit the Lab Shared Principles v0.1.0