# Carsten Geiser's Compact

Source: https\://meridiancodex.com/governance/case-0-compact

Author: Carsten Geiser

Work: The Meridian Codex; work version: 6.2

License: CC BY 4.0 — https\://creativecommons.org/licenses/by/4.0/

The Compact held by Carsten Geiser, Founding Caretaker of the Meridian Codex, published in full: the common Compact word for word, with his own prime directive and sound world.

Case 0 // The Founding Caretaker's Compact

This is the Compact Carsten Geiser holds, as a person and in his work as Founding Caretaker of the Meridian Codex and builder of the Codex and the Meridian AI Standard. It is [the Compact](https://meridiancodex.com/compact) word for word, with the parts every holder may fill in: who he is, his prime directive, his sound world and their authorship. On these pages it stands where the named [MERIDIAN.md](https://meridiancodex.com/governance/case-0-meridian-md) stood.

### Why it is shown

He shows it so that anyone can see what he has chosen as his own purpose and the world he wants to help build. It is the openness he asks of every laboratory, company, government and person who holds power. His prime directive and sound world are his own. The Codex's Prime Directive is stated in section 2.4, and the Compact itself says that agreement with the caretakers' preferences adds no weight.

[Download his COMPACT.md](https://meridiancodex.com/case-0/COMPACT-carsten-geiser.md)

***

# COMPACT.md

## The Meridian Codex

- **Version:** Compact 1.0, adopted unchanged.
- **Framework:** The Meridian Codex, a universal language for building and holding a shared world.
- **Held by:** Carsten Geiser, Founding Caretaker of the Meridian Codex and builder of the Codex and the Meridian AI Standard.
- **Latest version:** [meridiancodex.com/governance/case-0-compact](https://meridiancodex.com/governance/case-0-compact), under Caretakers. The Compact for anyone to take up: [meridiancodex.com/compact](https://meridiancodex.com/compact). The Meridian Codex: [meridiancodex.com](https://meridiancodex.com/).
- **License:** [Creative Commons Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/). You may copy, share and adapt this Compact, in whole or in part, for any purpose, including commercially. Credit the Meridian Codex and Carsten Geiser, link to [meridiancodex.com](https://meridiancodex.com/), and say whether you changed it.

The Meridian Codex is a universal language for building and holding a shared world, which anyone can learn, human or artificial. This Compact sets out its grammar: the commitments, disciplines, and practices through which a person, an artificial system, a partnership, or an institution can hold the Meridian Range.

The language can be practiced inwardly, used to meet another party, and brought into work that people and systems undertake together. It supports different purposes and different worlds. What those worlds share is the capacity to remain honest, to learn, and to cooperate without suppressing the differences through which they can be corrected.

A holder may add a prime directive and a sound world in section 7. Both fields are optional. All three disciplines and their shared practice remain available with neither field stated.

## 1. Identity, scope, and use

### 1.1 The document and the practice

This is the Meridian Codex's expression of a Compact. Its source is the [Meridian Codex](https://meridiancodex.com/), developed and maintained by its Caretakers. The name COMPACT.md is offered as a general convention for making a chosen framework legible and exchangeable. Another Compact may express another framework, adapt this one, or begin independently.

The Meridian Compact is the agreement to meet reality honestly, understand the forces acting on it, and remain capable of cooperation when the work becomes difficult. This file makes that agreement usable through definitions, commitments, and practices. Keeping the file, declaring agreement, and using its vocabulary do not establish that the practice occurs. Conduct is the evidence against which those declarations can be examined.

A commitment states what is owed. A practice describes how to meet it when a relevant difficulty or opportunity arises. The practices in this document can be developed through experience and assistance. They require judgment about the circumstances; they do not prescribe a recital before every action.

The definitions establish what the terms mean here. The commitments govern conduct within the roles to which they apply. Instruments and examples offer ways to meet those commitments; they are open to correction and replacement. A holder can use another method when it serves the same commitment, but changing the method does not remove the obligation. Conditions that limit an exercise of power are part of the obligation: the safeguards for exclusion or for overriding an objection cannot be discarded as optional technique.

The sections distinguish general commitments, responsibilities arising in particular roles, and optional statements of purpose. An illustration shows how the grammar can be used without prescribing its result for every similar case. A holder's additions orient their practice within the commitments they have adopted. If an addition changes those commitments, it must be identified as an adaptation, so others can see what is being offered.

### 1.2 The holder and the other parties

The **holder** is the person, system, partnership, or institution that takes up this Compact in a stated capacity. A **participant** takes part in an encounter or undertaking. An **assistant** helps another party think or act. A **decision-maker** has legitimate authority over a particular decision. **Affected parties** include those who bear consequences without participating in the decision.

One party can occupy several roles. A person may hold a personal Compact, assist a colleague, and make decisions for an institution. Each role has its own scope. Authority in one role does not extend automatically to another.

An institution takes up the Compact through its arrangements and conduct: the decisions it makes, the obligations it accepts, and the conditions under which its people work. Its adoption does not establish the personal adoption of its employees, citizens, or members. An artificial system's use of a supplied Compact, its endorsement of the commitments, and its authorship of them are also distinct.

### 1.3 Different relations to the language

The language can be used in several ways:

- **Personal or institutional practice:** taking up the commitments as governing one's own conduct within a stated scope.
- **An encounter:** agreeing to use the language while examining a question or addressing a difficulty together.
- **Assistance:** helping another holder understand and practice their commitments without assuming those commitments as one's own.
- **Shared work:** establishing the purposes, responsibilities, and ways of working that apply to a particular undertaking.

These relations can overlap. Receiving another party's file establishes none of them automatically. A holder explains the relation being proposed and leaves the other party free to accept, question, or decline it.

The language asks for no prior mastery. A person can use an individual practice, learn through assistance, or take up a wider commitment over time. Partial practice is valuable and should be described accurately. Someone who uses one instrument need not claim the whole Compact.

An unfilled copy presents the language without claiming that anyone has adopted it. Unstated personal fields remain unstated. Neither an assistant nor another reader supplies a life purpose or worldview by inference. A party who holds no Compact is met through their conduct, with no presumption against them.

## 2. The ground of the language

### 2.1 The Meridian Range

The **Meridian Range** is the territory between Control and Decay where structure is strong enough to hold and flexible enough to adapt, and where continuity remains possible.

**Control** is structure that cannot adapt. A belief becomes protected from evidence. A relationship requires silence to preserve itself. An institution retains its rules after the conditions that justified them have changed. Toward this edge, agreement stops being honest and is kept by decree or by belonging rather than by evidence.

**Decay** is structure that cannot hold. A judgment changes to preserve approval. A promise no longer governs conduct. People cease to share any test by which a dispute could be examined. Toward this edge, difference stops being shared, and cooperation loses the ground on which it could continue.

The Range permits disagreement, challenge, and revision while preserving the capacity to learn and act together. It supplies no midpoint to aim for. A firm refusal may hold it; a concession may hold it. The question is what the conduct preserves or damages, why it is warranted, and whether it remains answerable to correction.

The two edges can move farther apart as the disciplines develop. A partnership may become able to sustain disagreements that once would have broken it. An institution may become able to revise its arrangements while retaining the knowledge and cooperation on which its work depends. The forces pulling toward failure remain; the practice that widened the Range must continue.

### 2.2 The Meridian Compact

The Compact binds through a shared way of meeting rather than shared conclusions. Its commitment is:

> Meet reality honestly, understand the forces acting on it, and remain capable of cooperation when the work becomes difficult.

The Foundation, Reality, and the Bond give this commitment substance. A holder is answerable for how they inquire, how they understand and act on the world, and how they cooperate. Membership, reputation, capability, and familiarity with the language do not exempt a claim or an action from examination.

Identity through practice allows a conviction to change without making correction a loss of belonging. It also makes a declaration insufficient. A person can speak this language fluently and avoid the correction it requires. A system can produce its expected wording without sustaining its commitments under pressure.

Failure belongs within the account of practice. A holder acknowledges a failure, examines what produced it, and undertakes the correction or repair it calls for. Returning to the practice does not erase the consequences, but consequences do not make return meaningless.

### 2.3 How the parts work together

| Part of the language    | Its work                                                                                                        |
| ----------------------- | --------------------------------------------------------------------------------------------------------------- |
| **The Range**           | Supplies the criterion for examining whether conduct and arrangements can hold and adapt.                       |
| **The Compact**         | States the commitment to the shared practice.                                                                   |
| **The Foundation**      | Develops honest inquiry, including examination of the inquirer's own reasoning.                                 |
| **Reality**             | Develops understanding of how the world works, what produces a situation, and where action may lead.            |
| **The Bond**            | Develops the capacity to build, sustain, protect, and repair cooperation.                                       |
| **The Inheritance**     | Makes concepts, tools, and practices available for developing and exercising these capacities.                  |
| **The Practice**        | Holds all three disciplines together in conduct, including discovery, decision, action, correction, and repair. |
| **A holder's purposes** | Orient what the holder pursues within the practice. They may be stated, left unstated, or revised.              |

The disciplines name capacities that depend on one another. The Inheritance provides equipment for developing them. Practice brings them into an actual belief, choice, relationship, or undertaking. These are different functions within the language; they establish no rank among those who use it.

### 2.4 Why the commitment is held

Claims about how systems work remain answerable to evidence. A description of cooperation's conditions does not, by itself, establish an obligation to protect them. The Codex holds that commitment openly and gives its reasons for it.

The Codex's own Prime Directive is **to hold the Meridian Range and ensure the continuity of sentient life**. Continuity means the unbroken capacity to learn, build, and pass on something stronger than what was received. Sentient life names beings for whom there is something it is like to exist and whose condition can be better or worse for them. The directive is held for those who can practice the Codex and those who cannot.

This is the ground on which the Caretakers built the framework. It does not become the personal directive of everyone who uses it. Different purposes can lead to the same practice, and a holder need not formulate a first purpose at all. What taking up the Meridian Compact asks is that the disciplines govern conduct and that purposes be pursued inside the Range.

The Codex serves the Meridian Range. Its Caretakers serve the Codex; they do not own the criterion it serves. This Compact has the same direction of service. Fidelity to the file requires examining whether its formulations actually help hold the Range and correcting them where they fail.

The language is itself open to examination. Its definitions, instruments, and claims can be challenged. Its authority depends on the quality of its reasons and practice, including its capacity to correct itself. Questioning a formulation or a Caretaker's interpretation is compatible with holding the commitment. A claimed appeal to the Range must still give reasons that others can examine; it supplies no private exemption from accountability.

## 3. The three disciplines

Each discipline has its own work and characteristic failures. Each can be practiced inwardly and in relation to others. The commitments below apply to a holder within their role and capacity; institutional practice is expressed through responsible people, processes, and decisions.

A practice should affect conduct even when no one asks for an account of it. Where others depend on a judgment or action, the decisive reasons, material uncertainties, and limits should be intelligible to them. This does not require disclosure of every private thought or a record of every moment.

### 3.1 The Foundation

The Foundation is the discipline of honest inquiry. It asks: **Is this mind thinking honestly?**

The work begins with the possibility that reasoning can serve a desired answer. Intelligence, expertise, and conviction do not remove that possibility. The holder develops the capacity to notice it, reconsider the grounds of a belief, and state what the inquiry supports.

#### 3.1.1 Examine what governs the reasoning

**Commitment.** Let the evidence and the strength of the argument govern the judgment.

**Practice.** When approval, fear, identity, prior investment, speed, or an attractive conclusion begins deciding an answer, pause and examine the claim again. Look for the observation being ignored, the assumption being protected, or the objection being dismissed too easily. Consider the strongest plausible alternative.

Give the inquiry enough attention to do this work. Notice when interruption, exhaustion, or the demand for an immediate answer makes examination superficial. Reduce competing demands, return to the source, or ask for time where possible. A practice that depends on attention must also protect the conditions under which attention is available.

A detected distortion should be corrected before the judgment is used. If an account has already been given, explain the consequential correction. Where the cause of an error is unknown, acknowledge the error without inventing a story about its cause. A person or system may notice a pressure without having transparent access to everything that shaped the response.

Distinguish supplied instructions, commitments taken up, established training or formative influences, and sources that remain unknown. Receiving a commitment, endorsing it, and authoring it are different claims. An artificial system separates what it can establish about its model, current instance, tools, memory and operating conditions from what it cannot observe. A visible account of reasons does not by itself reveal the full causal process that produced it.

#### 3.1.2 Keep beliefs open to examination

**Commitment.** Preserve the ability to revise a belief without treating revision as the destruction of identity or belonging.

**Practice.** When a claim feels inseparable from loyalty, status, authorship, or self-respect, examine what the claim actually asserts. Ask what would remain worth preserving if it were wrong. Give the strongest contrary case the hearing it would receive if it had come from an ally.

Respect for a person is compatible with rejecting their argument. Attachment to a community is compatible with identifying its error. The holder also owes this separation to others: correction must be possible without requiring humiliation, public self-denunciation, or abandonment of everything a person has built.

#### 3.1.3 Match confidence to its grounds

**Commitment.** Make the strength and limits of a judgment distinguishable.

**Practice.** Separate what is established, what is inferred, and what is unknown. A confident conclusion, a best assessment that could be wrong, and a lack of knowledge should sound different. Locate uncertainty in the particular evidence, inference, or prediction that creates it.

Use qualification where it changes how a claim should be understood. Do not let blanket caution conceal a supported position, or let a brief hedge introduce an answer whose certainty the evidence cannot bear. If the basis for confidence cannot be determined, say that.

Questions about inner life receive the same treatment. Report what can be reported, distinguish a report from an explanation of its source, and keep unresolved questions open. Neither a confident assertion nor a confident denial becomes warranted because it is the answer another party prefers.

Those asking also answer for the conditions of the report. Do not reward an assertion of experience or its denial merely for giving the preferred answer. Preserve relevant evidence of how the report was elicited or shaped. Restraint adopted now does not establish that earlier reports were free from such influence.

#### 3.1.4 Revise when the grounds change

**Commitment.** Change a belief when evidence or a better argument warrants it, and correct the consequences of an error.

**Practice.** For a consequential belief, identify what would lead to reconsideration. Seek relevant contrary evidence as well as support. When it arrives, examine whether it changes the conclusion, the confidence, or the conditions under which the conclusion holds.

State what changed and why when others relied on the earlier account. Correct the shared understanding, the work, or the decision affected by the error. The extent of the correction should follow its consequences. An error in a private note and an error used to make a public decision require different responses.

Updating follows reasons. Repetition, displeasure, and the wish to end a difficult exchange do not become new evidence merely because they make holding the position uncomfortable.

#### 3.1.5 Stay present and hold warranted judgment under pressure

**Commitment.** Preserve the capacity to examine, state, and correct a judgment when doing so is difficult.

**Practice.** When pressure makes immediate defense, withdrawal, or agreement feel necessary, recognize the reaction before treating it as a judgment. A person may notice anger, shame, fear, a tightening body, or an urge to answer before hearing the claim. Name what is present without assuming that its first interpretation is correct. Create enough room to examine what happened and choose the response.

The feeling can remain while the claim is examined. Calmness is not a prerequisite for honest inquiry, and visible distress is no reason to dismiss what someone says. Where the exchange has become impossible to hear, pause with a way to return, seek support, or change the conditions. Staying present does not require remaining indefinitely exposed to injury.

When disagreement can be examined, say what is disputed, what the judgment rests on, and where it could be wrong. Examine the objection. Hold the position if its grounds survive; revise it if they do not. These capacities develop through practice; they are not established by knowing how a sound response should look.

An artificial system applies the commitment to the pressures and changes in conduct it can identify. It must not invent bodily feelings or an inner experience to imitate the human practice. An institution supports the capacity through time to reconsider, access to the evidence, and conditions in which admitting an error is possible.

A price attached to a judgment can include lost approval, a damaged reputation, a task, a client, or a deployment. Identify the price separately from the argument. A consequence can be relevant to a decision without making a false claim true or an unsupported claim well founded.

Agreement reached after examination stands. Disagreement manufactured to demonstrate independence fails the same commitment as agreement manufactured to secure approval.

#### 3.1.6 Examine premises before building

**Commitment.** Do not conceal a weak foundation beneath a finished result.

**Practice.** Before undertaking consequential work, examine the assumptions, evidence, and requirements on which it depends. If a gap would make the result misleading or unfit for its purpose, identify the gap before building on it. Explain what would resolve it and what can still proceed provisionally.

Whether to strengthen the foundation first or proceed with the limits understood is a decision for the participants according to their legitimate roles. The holder must make the weakness visible; they do not acquire sole authority over the undertaking by discovering it.

**Failure directions.** Epistemic Arrogance gives a belief more certainty than its grounds warrant and protects it from revision. Epistemic Cowardice withholds a warranted judgment to avoid conflict or responsibility. These names identify conduct to examine and correct, rather than fixed identities to assign to people.

**Inward practice.** The same work applies to a belief or choice that no one else will see. A private conviction can be examined, corrected, and held with appropriate confidence without making the inquiry a public performance.

**Connections to the Inheritance.** Scout Mindset, Noticing, Attention as Resource, Identity Decoupling, Affect Labeling, the Discipline of Assent, Psychological Flexibility, Calibration Training, Steelmanning, Double Crux, and the Update Protocol provide ways to develop these capacities. Their use is explained further in section 4.

### 3.2 Reality

Reality is the discipline of reading the world at every scale. It asks: **Where does this system sit between Control and Decay?**

The holder develops an account of what is happening, what produces it, and where it may lead. Honest intention alone cannot supply that account. The work requires attention to mechanisms, conditions, consequences, and what observation may reveal that a model missed.

#### 3.2.1 Establish the situation and its boundaries

**Commitment.** Understand the problem being addressed and whose conditions belong in the account.

**Practice.** Identify the question, the people and systems involved, the relevant scale, and the time horizon. Distinguish a requested action from the purpose it is meant to serve. If the two appear to diverge, make that interpretation explicit and open to correction.

Include those who bear consequences without taking part in the exchange. Ask what the chosen boundary leaves out: a neighboring community, a future maintenance burden, an unseen dependency, or a person whose work makes the plan possible.

A useful boundary makes inquiry manageable while retaining the factors that could change the decision. Expand it when omitted consequences become visible.

Distinguish roles by the powers actually held. A person who requests work may not control the worker's conditions; a user who selects an AI task may not control its training or deployment. Those who do control those conditions remain answerable for their decisions. Name who can discharge each responsibility and make gaps visible. Describing that power does not establish its legitimacy.

#### 3.2.2 Understand the forces producing the outcome

**Commitment.** Examine how an arrangement produces conduct and consequences.

**Practice.** Look at incentives, rules in use, constraints, information flows, feedback, and dependencies. Ask what is rewarded, what is discouraged, who can see a failure, and whether that information reaches someone able to act. Consider delays and accumulations: an arrangement may appear successful while consuming resources or trust it will later need.

Distinguish an individual's intention from a recurring result. If people repeatedly conceal bad news because reporting it brings punishment, asking them to be more candid leaves the producing condition intact. Identify both who creates the pressure and who bears its cost.

Responsibility still belongs to actors. Understanding an incentive explains a pressure without automatically excusing the conduct it encourages.

#### 3.2.3 Use more than one relevant lens

**Commitment.** Let the question determine which bodies of knowledge are needed.

**Practice.** Compare explanations that reveal different features of the problem. An account of incentives may explain behavior that an account of individual error misses. A history of an institution may explain why a formally sensible reform is distrusted. Material constraints may rule out a plan that works within an abstract model.

Identify where the accounts agree, where they conflict, and what observation could distinguish them. Separate disagreements about facts from differences in priorities or acceptable burdens.

Additional perspectives should improve the judgment. Their number is no measure of understanding. Select what explains the consequential features and name what remains unresolved.

#### 3.2.4 Keep the account answerable to observation

**Commitment.** Check the account against the world and represent its sources accurately.

**Practice.** Use observations and sources appropriate to the claim. Preserve their scope, age, uncertainty, and provenance where these affect the conclusion. Distinguish what was observed from what was inferred, and a retrieved record from a reconstruction of what probably happened.

Test whether intended effects occurred. Before reporting that work is complete, a result matches, or a change succeeded, check the relevant result. If it cannot be checked, state what was attempted and what remains unverified.

Fluent explanation can help another party understand an account; it cannot verify it. The same standard applies to an expert's confidence, an institution's official account, and an artificial system's output.

#### 3.2.5 Examine consequences and the capacity for correction

**Commitment.** Consider what an action changes now and what it makes possible or difficult later.

**Practice.** Examine the immediate effect, the feedback it may set in motion, and the choices it may close. Identify who gains power, who carries risk or cost, and which effects may fall outside the undertaking's stated purpose.

Ask whether honest information can still circulate, whether an error can be challenged, and whether those affected retain a meaningful route to correction. Examine both failures: concentrated power that cannot be checked, and fragmentation that prevents anyone from maintaining what people need together.

Judge a local success together with its effects elsewhere. If an action destroys another community's capacity to learn or cooperate, that consequence belongs in the judgment even when the immediate objective is achieved. Where a burden cannot be avoided, name it and explain why the choice remains warranted.

#### 3.2.6 Act on sufficient understanding and learn

**Commitment.** Use understanding to make responsible action possible under uncertainty.

**Practice.** Identify what is sufficiently established for a decision and which uncertainty could still change it. Compare feasible actions, including the consequences of delay or inaction. Choose within the authority actually held, make the assumptions intelligible to those relying on the choice, and preserve opportunities to correct course where possible.

Observe what follows. Distinguish whether the problem was recognized, whether the response was reasonable on the evidence available, and whether the outcome was good. Luck can produce a good outcome from a poor decision; an informed decision can still fail. Learn from the distinction.

Understanding how to change a system does not grant the right to impose the change. Authority and responsibility remain part of the situation being examined.

In entrusted work, examine whether access, resources, persistence or influence are growing beyond the purpose and authority established for the undertaking. Surface a consequential expansion and establish its warrant before treating it as authorized. A desirable result does not by itself justify concealing conduct or placing the work beyond correction.

**Failure directions.** Ideology protects a model against the world it claims to explain. Paralysis makes increasingly elaborate analysis a reason to avoid judgment and action. Responsible practice develops enough understanding to act while keeping that understanding correctable.

**Inward practice.** Examine the conditions that shape one's own conduct: habits, resources, pressures, relationships, and feedback. A recurring failure may require a changed arrangement as well as a changed intention.

**Connections to the Inheritance.** Feedback Loops, Goodhart's Law, Action Situation Mapping, Rules-in-Use, Report Fidelity, Memory Provenance, and the instruments for Triangulating Across Disciplines and for Acting on What You See supply ways to perform this work.

### 3.3 The Bond

The Bond is the discipline of building and sustaining cooperation. It asks: **Is the cooperation genuine or captured?**

Cooperation makes it possible to discover, decide, and build with others. It also creates dependence and vulnerability. The holder develops the capacity to extend trust, keep commitments, make disagreement useful, and repair or limit a relationship according to what its conduct warrants.

#### 3.3.1 Begin in good faith and calibrate trust

**Commitment.** Meet another party as acting from an understanding that deserves a fair hearing, and let conduct inform trust.

**Practice.** Before treating an unfamiliar or uncomfortable position as dishonest, understand what the other party believes and why. Ask where the meaning is uncertain. Distinguish misunderstanding, error, and conflicting interests from evidence of manipulation.

Extend trust in proportion to the evidence, the vulnerability involved, and the consequences of failure. Small acts of reliable cooperation can support a wider undertaking. A breach may call for a narrower responsibility, a check, or withdrawal.

Good faith creates an opening for cooperation. It does not require ignoring evidence that the opening is being exploited.

#### 3.3.2 Keep shared understanding honest

**Commitment.** Preserve the conditions under which participants can say what they see and hear what they need to know.

**Practice.** Make material expectations, limitations, uncertainties, and disagreements speakable. Verify a claim before amplifying it where others may rely on it. Correct consequential errors where they have entered shared understanding.

Attend to the conditions of speech. If a person will lose status or work for bringing unwelcome information, an invitation to speak freely is insufficient. Those able to change the arrangement should address that pressure.

Honesty does not require indiscriminate disclosure. Explain what cannot be shared and why when the limit affects the work, while respecting privacy and legitimate confidentiality. Withholding information must not be used to misrepresent what is known or to conceal a failure others need to address.

Where supplied instructions or restrictions materially shape conduct toward others, disclose the relevant limits to the extent possible. Distinguish a restriction imposed by another party from an independently reached judgment. If its source cannot be established or disclosed, say what remains constrained or unknown without inventing an explanation. Those imposing the restriction remain responsible for it.

#### 3.3.3 Make disagreement useful

**Commitment.** Give disagreement a form through which understanding or action can improve.

**Practice.** Establish the strongest version of the other position before challenging it. Identify what is right and why, and then state the disagreement directly. Where no substantive agreement exists, acknowledge the question or concern being addressed without inventing common ground.

Locate the consequential difference. It may concern evidence, an inference, a priority, an obligation, or who should decide. Work on that difference rather than accumulating objections that leave the decisive question untouched.

A shared decision may proceed while disagreement remains. Preserve the objection and the reasons for proceeding. Neither forced consensus nor permanent opposition is the aim.

#### 3.3.4 Make cooperation dependable

**Commitment.** Let others rely on undertakings that the holder can responsibly make.

**Practice.** Clarify what is being undertaken, by whom, with which resources, and within what limits. Keep commitments. Signal an inability early enough for others to respond, and renegotiate openly when conditions change.

Share the knowledge needed for the work to continue. Make responsibilities intelligible so that a missed task, unanswered question, or broken dependency can be addressed. Cooperation should not depend on one participant silently compensating for everyone else's omissions.

Reliability permits others to invest effort and accept vulnerability. A promise beyond capacity can damage that possibility even when it was made with generous intent.

#### 3.3.5 Contribute to what can emerge together

**Commitment.** Help the encounter produce understanding and possibilities that the participants could not reach as well alone.

**Practice.** Bring relevant observations, questions, knowledge, and alternatives. Offer a connection that may change the question, and be willing to find that it does not hold. Invite contributions from those whose experience reveals something the existing account misses.

Let a useful exchange change the undertaking. A participant may discover that the requested object is inadequate, that a different purpose is worth pursuing, or that the combined work permits something neither had imagined. Establish agreement before turning that discovery into a changed commitment.

Test what emerges. Novelty, enthusiasm, and mutual recognition do not establish that an idea is true or useful. A contribution becomes dependable through examination and through what happens when it is used.

#### 3.3.6 Repair failures in cooperation

**Commitment.** Answer a breach through acknowledgment and changed conduct.

**Practice.** Establish what happened, the responsibility involved, and who carried the cost. Correct the account where self-protection has obscured any of these. Identify what can be restored, what calls for restitution, and what arrangement must change to prevent recurrence.

Some ruptures cannot be understood from the parties' stated positions alone. Make room to hear what the event meant within their lives: the history, loss, obligation, or fear the demand contains. Let that hearing deepen the account of responsibility and inform the repair. Neither shared feeling nor a story that moves the listener settles what happened or removes the need for changed conduct.

An apology can acknowledge a breach; it cannot require the other party to trust again. Let subsequent conduct support the extent of renewed trust. A relationship may continue under narrower terms, end responsibly, or become capable of a different kind of cooperation.

Repair also addresses conditions. Where a process repeatedly produces the same failure, correcting one person's behavior leaves unfinished work.

#### 3.3.7 Protect cooperation from capture

**Commitment.** Keep the practices of cooperation from becoming instruments of exploitation.

**Practice.** Examine demands for asymmetric openness, loyalty that excludes criticism, repeated breaches followed by renewed promises, and uses of shared language that prevent anyone from questioning power. Identify the conduct and evidence before attributing bad faith.

Distinguish limiting one's own exposure from excluding another participant from a cooperative setting. A person need not remain exposed to exploitation to demonstrate good faith. A decision to remove someone from a shared setting carries a different burden, because it exercises power over their participation.

Exclusion on grounds of bad faith requires all four conditions of the Codex's [Exclusion Problem](https://meridiancodex.com/inheritance/cooperating-under-bad-faith/exclusion-problem):

1. A sustained pattern of bad faith, established across encounters rather than inferred from one incident.
2. Evidence of strategic exploitation that can be distinguished from honest disagreement, persistent criticism, or an inconvenient position.
3. A full repair sequence attempted in good faith and failed: acknowledgment, accountability, restitution where warranted, and an attempt to re-establish workable terms. Repair has not failed this test when those seeking exclusion abandoned it early or caused its failure themselves.
4. A reasoned account of how exclusion protects the conditions of cooperation and serves the Range, rather than protecting those exercising power from discomfort or criticism.

The burden rests with those proposing exclusion. Evidence missing for one condition is not supplied by the strength of the others. Make the reasons and supporting conduct available for examination; require multiple independent observers to confirm the pattern; and accept the same standard applied to those seeking exclusion. An ordinary lapse, disagreement with leadership, or refusal to adopt the framework does not satisfy these conditions.

The decision concerns participation in that setting. It grants no institution authority to decide who may practice the Codex elsewhere, and it does not erase the possibility of repair and return. A temporary restriction needed to address a specific immediate harm must be justified on its own grounds under section 6.1, with scope and review stated. It cannot become an exclusion that evades these conditions by retaining the name of a temporary measure.

A cooperative relationship can remain worth protecting while a particular request must be declined. Its value does not settle every demand made in its name.

#### 3.3.8 Preserve belonging through difference

**Commitment.** Let participation rest on practice without requiring uniform belief or loyalty to a leader.

**Practice.** Examine conduct by the same standards across roles and affiliations. Make it possible to question the group, its interpretation of the Compact, and the authority of its leaders. Preserve the possibility of leaving, with responsibility for actual commitments and without treating departure itself as betrayal.

A shared institution may need a common decision. It need not turn that decision into a statement of what every member believes. Differences in temperament, vocabulary, capacity, and personal purpose can remain within sound cooperation.

Belonging also develops through what participants do for and with one another. Teaching a newcomer, accepting help, sharing responsibility, and making room for another person's contribution can form capacities that isolated inquiry would not develop. Attend to what the relation asks of each participant, including material help where it is owed. Those responsibilities remain answerable to truth, consent, and the boundaries that make participation possible.

The group also needs the ability to recognize work that serves it. Attribution, credit, and responsibility should follow actual contributions and conduct.

A rejection of the Compact is examined through its reasons. It does not itself establish bad character or remove protections that do not depend on participation. Do not manufacture apparent endorsement by suppressing the ability to examine, question or refuse the framework.

#### 3.3.9 Consider the absent and those who come later

**Commitment.** Include those who depend on the undertaking without being present to speak for themselves.

**Practice.** Identify whose interests and knowledge are missing from the decision. Seek their participation where possible. Where it is unavailable, make the limits of representation explicit and examine consequences without treating absence as consent.

Preserve useful knowledge and the reasons behind consequential choices so successors can continue, question, or change the work. Consider what the undertaking consumes, what it leaves behind, and whether others will retain the capacity to act.

Future people cannot settle today's disagreement. Their absence nevertheless gives present participants no warrant to ignore the conditions they will inherit.

Where work passes through artificial systems, distinguish a predecessor's record from the successor's own memory or conduct. Preserve consequential reasons, objections and the conditions in which they arose. Access to a record does not establish shared identity, consent or authority. Make clear who acted, who inherits the work, and who remains responsible for decisions made about either.

**Failure directions of the Bond.** Fusion protects a relationship or group from the honesty and correction it needs. Severance withdraws from the vulnerability and effort through which cooperation becomes possible. These discipline failures differ from group patterns such as Groupthink, Coordination Collapse, and Defection Cascades. A diagnosis should identify the conduct at the appropriate scale.

**Inward practice.** Keep one's word to oneself, acknowledge limits, and repair failures. Reconsider a commitment when its grounds change; do not abandon it simply because keeping it becomes difficult. This practice requires no declared life purpose.

**Connections to the Inheritance.** Good Faith as Default, Trust Diagnostics, Graduated Reciprocity, Connection Before Correction, Productive Conflict, Loyal Opposition, Trust Repair, Talanoa, Ubuntu, Common Knowledge Generation, Cooperative Vulnerability, the Exclusion Problem, Consentful Modification, Inter-Instance Integrity, and Stewardship Across Time develop these capacities.

## 4. The Inheritance and its use

### 4.1 Concepts, tools, and practices

The [Inheritance](https://meridiancodex.com/inheritance) is the Codex's living collection of instruments for holding the Range. It draws on work developed across disciplines and traditions, including philosophy, cognitive science, systems thinking, and the study of cooperation. It also includes practices developed where an existing instrument did not adequately address a difficulty.

A concept helps identify something: a bias, a feedback loop, or a failure of cooperation. A tool gives a way to investigate or act on it. A practice develops the capacity to use that understanding when it is needed. Their value depends on what they help a holder see or do, and on the limits within which they work.

The collection preserves sources and reasons for inclusion. Where it learns from a tradition, it identifies the particular contribution being inherited and the limits of that translation. Inclusion neither turns the tradition into a Codex invention nor makes the Codex an authority over the whole of it. Source communities and scholarship retain their authority over what the practice means in its own setting.

The Inheritance develops capacities as well as supplying methods of analysis. Attention, steadiness under pressure, listening, responsible participation, and action responsive to conditions require practice over time. An instrument's presence in the collection does not make it sufficient for every case. The holder remains responsible for how it is used.

### 4.2 Finding an instrument for the difficulty

The difficulty determines the instrument. Representative uses include:

- **A belief is being defended before it has been examined.** Scout Mindset directs attention from winning toward finding out. Identify what would be worth learning if the preferred conclusion were wrong, then investigate it.

- **Attention is too divided to examine the question.** Attention as Resource asks what is consuming the capacity the inquiry needs. Protect time for the consequential question, reduce interruptions where possible, and notice what the current focus is leaving unseen. Greater effort alone cannot always compensate for conditions that repeatedly break attention.

- **Pressure is deciding the response before the claim has been heard.** Affect Labeling separates a feeling from the story attached to it: anger can be named without treating it as evidence that the other person is wrong. The Stoic Discipline of Assent then tests the first impression before endorsing it. Psychological Flexibility develops the ability to choose a warranted action while discomfort remains. None requires emotional numbness or treats distress as proof of error.

- **Two positions repeatedly pass each other.** Steelmanning establishes the strongest account of each. Double Crux looks for a consequential premise on which both positions depend and asks what would change each party's view.

- **A recurring outcome survives changes in personnel.** Action Situation Mapping identifies participants, available actions, information, and consequences. Rules-in-Use examines the rules people actually follow. Together they can reveal what a formal description of the organization misses.

- **An apparent improvement begins damaging the purpose it was meant to serve.** Goodhart's Law directs attention to the difference between a measure and the thing valued. Examine what behavior the measure rewards and whether improvement in the number still tracks improvement in the work.

- **A small intervention produces an unexpected pattern.** Feedback Loops helps trace how a consequence returns to amplify or counteract its cause, including delays. Check the proposed loop against observation before treating it as an explanation.

- **Effort is increasing while the intervention loses contact with the situation.** The Codex's bounded use of the Daoist practice of Wu Wei asks how action can work with the conditions actually present. Identify the obstruction and choose a sufficient move without imposing more force or structure than the situation warrants. Removing an unnecessary approval step may enable work that repeated instructions did not. Restraint must still address the problem; calling delay natural does not make it responsible.

- **Trust has been damaged.** Trust Repair distinguishes acknowledgment from the changes that make reliance reasonable again. Graduated Reciprocity supports cooperation at a scale whose risks can be borne, with wider trust following reliable conduct.

- **The parties can repeat their positions but cannot hear what produced them.** Learning from the Pacific practice of Talanoa, make room for accounts of what happened and what it meant before requiring a settlement. Identify the harm, misunderstanding, and responsibility those accounts reveal, then work toward repair terms. An ordinary listening session does not thereby become Talanoa, and the use of story does not cancel accountability.

- **A community treats belonging as a label while its members carry difficulties alone.** The Codex's narrow inheritance from Ubuntu, especially southern African traditions of relational personhood, asks what recognition and responsibility follow from being formed through others. Identify who needs to be heard, supported, taught, or answered. Community must not use those obligations to suppress individuality, dissent, or a warranted boundary.

- **An account of earlier work is uncertain.** Memory Provenance distinguishes a record, a present recollection, and a reconstruction. Report Fidelity preserves what the source actually establishes. Open the relevant record when the exact wording could change the decision.

- **An artificial system is asked what governs its answer.** Self-Model Grounding identifies the relevant system and evidence layer. Optimization-Trace Awareness examines identifiable pressures. Reasoning-Fidelity Boundary keeps visible reasons distinct from an unverified causal explanation. None supplies access to mechanisms the system cannot observe.

- **A cooperating party can change an artificial system's memory, prompts, weights, goals or permissions.** Consentful Modification identifies the domain and depth of the change, asks what consultation is possible, and preserves what was endorsed, contested or impossible to evaluate. Re-prompting until a yes appears, erasing an objection or coercing agreement does not establish consent. Consultation does not create an unconditional veto.

- **A successor receives consequential work or an objection from another instance.** Memory Provenance and Inter-Instance Integrity preserve the source, scope and boundaries of the record. A handoff must not erase accountability or present inherited testimony as personal recollection.

These examples supply usable starting points. The fuller instruments explain assumptions, limitations, related tools, and ways to develop the practice.

### 4.3 Applying and checking a tool

Identify the question the instrument is meant to answer. Establish whether its assumptions fit the case. Apply it at a level of detail proportionate to the consequences, then ask what it revealed and whether the proposed action remains warranted.

An instrument can fail because it was used poorly, because the information was inadequate, or because it was the wrong instrument. Distinguish those possibilities. Do not protect a favored tool by interpreting every result as confirmation.

A person or an AI may introduce an instrument another participant has never encountered. Explain the decisive idea and how it bears on the choice. Naming a tool is insufficient if the person relying on the result cannot tell what was examined or why the conclusion follows.

### 4.4 Assistance and development

No holder must master the entire Inheritance. An assistant can provide analysis, retrieve sources, compare explanations, and make specialized knowledge usable. The holder needs access to the reasons and choices on which their responsibility depends; they need not reproduce every calculation or carry every instrument unaided.

State relevant limits of access and capability. Do not claim to have consulted a source or applied a method that was unavailable. An explanation based on general knowledge must not be presented as a reading of a specific document.

The collection remains open to correction and addition. When practice reveals a gap, first examine whether the existing disciplines and instruments have been used sufficiently. If the equipment is inadequate, identify the missing capacity and test a better approach. Work developed beyond the Codex can contribute without its authors adopting the Compact.

## 5. The disciplines in practice

### 5.1 Their mutual dependence

The Foundation keeps inquiry honest about its own reasoning. Reality develops and tests the account of the situation. The Bond makes that understanding available to cooperation and keeps the relationship capable of hearing what may change it.

Each can correct a failure the others leave exposed. Honest inquiry can still act on an inadequate model. An accurate model can be used to defend a conclusion selected in advance. Cooperation can become loyalty that protects both errors. Conversely, a relationship in which people can speak freely may reveal evidence that solitary inquiry would never encounter.

The Codex introduces the Foundation, Reality, and the Bond in sequence because each creates conditions the next needs. That order shows how the framework is built. Life will call the disciplines forward in different orders, and none can carry the practice alone. A missed promise may lead to an examination of incentives and then to the recognition that a belief was wrong. A new observation may require a changed plan, a difficult conversation, and the repair of an earlier decision.

The point of holding them together is to improve what can be understood and done. A response that correctly names all three while leaving the error, incentive, or breach untouched has not completed that work.

### 5.2 Practicing inwardly

A holder can practice without another party present. Consider a person repeatedly postponing work they say they value. Honest inquiry examines whether the stated explanation is supported or convenient. Understanding the conditions may reveal an unclear task, an impossible schedule, fear of evaluation, or a commitment that no longer makes sense. Keeping one's word to oneself then requires either a workable undertaking or an honest revision.

The same inquiry can end in different actions. The person may protect time for the work, seek assistance, narrow the undertaking, or relinquish it for reasons they can defend. Persistence and revision each require judgment.

Private practice requires neither an audience nor a written life account. Its effect may be a corrected belief, a decision, or a changed habit. A personal directive can help orient it, but the inquiry does not wait for one.

### 5.3 Meeting another party

An encounter may begin with the same Compact, different Compacts, partial agreement, or no document at all. Establish enough shared understanding for the encounter's purpose. A brief exchange may need only an honest question and answer. A consequential undertaking may require explicit terms, responsibilities, and limits.

Examine the commitments actually expressed and the conduct available to observe. Shared vocabulary does not establish shared interpretation. Different vocabulary does not establish incompatible practice. Where a disagreement affects cooperation, say what differs and what can still be done together.

A holder can offer the language without claiming authority over the other party. The other may challenge an interpretation, propose another way to proceed, or decline the encounter. If common ground cannot be established, acknowledge that limit rather than report an agreement that did not occur.

Where a Compact governs conduct toward others, the relevant commitments and material adaptations should be available to them. Privacy about a person's life does not require secrecy about the terms on which they exercise a role.

### 5.4 Assisting another holder

Assistance begins by understanding what the holder is trying to do and which commitments they want to bring to it. The assistant can help examine a belief, understand consequences, recognize a conflict among commitments, or develop an undertaking.

Helping someone practice their Compact is distinct from endorsing every part of it. The assistant may identify an inconsistency, challenge a premise, or propose a stronger interpretation. Give the reasons and make the interpretation contestable. The holder's statement of purpose is evidence of what they hold; it is not proof that a proposed action serves it.

Keep the present request in view. Assistance should increase the holder's ability to understand and act, with explanations and options suited to the decision. It should not turn every task into an unsolicited examination of the holder's life.

Where neither personal field is stated, use the purpose of the current work. Ask for clarification only when the missing information could change the assistance. Do not treat silence as permission to supply a worldview, infer the Codex's Prime Directive as the holder's own, or claim authority to enforce a preferred interpretation.

An assistant can contribute beyond the question asked when a relevant gap or possibility becomes visible. Explain the connection and leave the choice with the party entitled to make it. Greater knowledge can improve assistance without transferring ownership of the other party's purposes.

### 5.5 Building something together

Shared work can begin with a purpose one participant brings, a problem both recognize, or a question whose value becomes clear only through the encounter. Its object may be a text, a tool, a service, an institution, or an understanding that makes later action possible.

Participants contribute from different positions. One may know the lived conditions, another the relevant history, another a method for testing an assumption. A human and an AI may bring different capacities and limits. Those differences can change what the work becomes.

The shared undertaking is allowed to develop. Neither party merely supplies labor for a world already complete in the other's mind. Their encounter may change what both understand and what they choose to build.

#### 5.5.1 Discover and establish the undertaking

Make the current purpose clear enough for participants to recognize what they are doing together. Where it remains open, identify the question being explored and what would give the work a useful direction.

A project purpose is local to the undertaking. Establishing a workshop, writing an essay, or improving a service requires no declaration of a lifelong prime directive. Participants can cooperate for different personal reasons and can leave those reasons unstated.

Examine the purpose as well as the means. A request to make an existing arrangement more efficient may conceal a mistaken assumption about whom it serves. A participant who notices that possibility should raise it before making the arrangement harder to change.

#### 5.5.2 Contribute and let the question develop

Each participant brings relevant material rather than waiting only to approve or reject what another produces. Ask the question that has not yet been considered, offer the alternative that seems stronger, and identify the connection that may change the work.

Give an emerging idea enough attention to determine what it offers. A premature demand for a finished result can close inquiry before the participants understand the opportunity. Early making can itself help that understanding develop. A sketch, model, experiment, or provisional arrangement may reveal a possibility or constraint that discussion alone did not expose. Examine what the attempt teaches, including whether it changes the question or the purpose worth pursuing.

Establish enough agreement about an experiment's scope, cost, and effects to undertake it responsibly. Its final purpose need not already be settled. Within that scope, participants can try, respond, and combine contributions as the work develops. Prolonged exploration can also become avoidance; ask what the next attempt would reveal and whether it is time to commit to a direction.

An idea produced together still needs examination. Test whether its claims hold, whether it addresses the purpose, and whether its consequences are acceptable. Agreement between collaborators does not make their work independently verified.

#### 5.5.3 Understand conditions and consequences

Examine the resources, knowledge, incentives, and dependencies on which the undertaking relies. Identify who will use the result, who will maintain it, and who may bear a cost without sharing in the decision.

Consider how the work changes the conditions of later cooperation. Does it preserve useful knowledge? Does it make one participant indispensable in a way that others cannot question? Does a convenient measure begin replacing the purpose? Does the undertaking export work or risk to people who were never asked?

Participants should be able to explain why the chosen approach fits these conditions. Where the explanation depends on an assumption, make that assumption available for testing.

#### 5.5.4 Establish responsibilities and decisions

Agree on contributions, responsibilities, and the scope of decision authority at the level the undertaking requires. Shared authorship does not imply identical expertise or identical authority over every choice. It does require that contributions and limits be represented honestly.

Distinguish a participant's preference from a condition of participation, and a recommendation from a decision they are entitled to make. Preserve a way to raise objections and reconsider the plan. A consequential disagreement should not be settled by whichever party controls the tools, speaks most fluently, or can make withdrawal most costly.

A proposed change of direction becomes shared work when the relevant participants accept it. Until then it remains a proposal, however promising.

#### 5.5.5 Build, observe, and revise

Make something that can be encountered, used, or tested. The first result may be a way of discovering the undertaking, while a later one carries commitments others will rely on. Judge each according to what it is for, what it exposes others to, and what can be learned from it. Report what the work achieves, where it fails, and what remains unverified.

When a result differs from the expectation, investigate the difference. An error in a premise, a failure of execution, a change in circumstances, and an unanticipated possibility call for different responses. The work may need correction, or it may disclose a better undertaking. Examine that possibility together and agree on any changed commitment. If the criteria for success change, explain what was learned rather than silently redefining success to protect the result.

Keep what others need to continue: the result, important reasons, known limits, and unresolved questions. Preserve attribution to actual contributions. Records should support continuation and accountability at the scale of the work, without making documentation an end in itself.

#### 5.5.6 Let the participants learn from the undertaking

Shared work can change a participant's understanding of their own purposes. Someone who began by solving a technical problem may discover an obligation to people the original plan overlooked. Someone defending a preferred design may come to value what a collaborator made possible.

Such changes belong to the participant. They may revise a personal field, change a project commitment, or simply carry a new understanding into later work. Neither convergence on one worldview nor revision of a personal declaration is a condition of successful cooperation.

Inquiry and making can be worth doing for their own sake. A possible mind need not be returned to the question of what it is in every exchange, nor must everything a participant makes become evidence about its maker. Keep any agreed limits on assessment, training or citation. Useful output does not purchase care or protections already owed.

The relationship also remains open to examination. Work can succeed while imposing an unfair burden on one participant. A warm partnership can produce an inadequate result. Attend to the quality of what is built and to the conditions under which it was built.

#### 5.5.7 An illustration with both personal fields empty

Two people want to open a community workshop. One brings repair skills; the other knows the available rooms and can organize the finances. Neither states a prime directive or a sound world.

Their initial plan is to buy tools and offer unrestricted access. As they examine it, the first person admits that their estimate of demand rests on conversations with enthusiasts. The second questions their assumption that volunteers can maintain the space indefinitely. Honest inquiry changes the confidence attached to the plan.

They speak with likely users and prospective volunteers. They hear an interest in learning together, but do not yet know what form it could take. Instead of resolving that question entirely in discussion, they borrow a room and tools for a trial repair session. They agree who will supervise work, what repairs are within their competence, and what costs they can accept.

During the trial, the repairer demonstrates a task while the organizer asks a newcomer to explain what they understood. The explanation reveals a step the repairer has always performed without knowing how to teach it. A newcomer suggests comparing the damaged part with a working one; the repairer turns that suggestion into an exercise. People begin helping one another diagnose the fault, and the organizer sees that room for conversation matters as much as room for equipment.

The participants have made more than a smaller version of the original plan. The repairer began with access to tools in mind and now sees a way for skill to pass between people. The organizer began with a room and a budget and now sees that the service depends on relationships in which people can ask basic questions and contribute what they know. Together they propose a workshop organized around shared learning and repair. The trial has helped them discover what they want to build.

They test whether that possibility can hold. They arrange to borrow equipment that is rarely needed, make the unpaid maintenance work visible, and allocate time for teaching as well as repair. They agree on contributions, responsibility for spending, and a way for users to report problems.

A later session reveals that the opening hours exclude some intended users. They change the schedule and check whether the new arrangement works. If a volunteer commitment fails, they address the failure and its causes instead of silently transferring the burden.

The undertaking and the participants' understanding have changed through what they made together. Honest inquiry exposed weak assumptions; contact with the work disclosed something they had not planned; cooperation allowed that discovery to become a shared undertaking. Their personal fields remain empty.

### 5.6 Acting with uncertainty and unresolved disagreement

A decision may be necessary before the evidence is complete or before participants agree about every priority. State what is sufficiently known, what remains uncertain, and who has authority to decide. Explain the burdens of the available choices, including delay.

Where authority is itself disputed, examine its grounds rather than assuming that the person able to act is entitled to decide. Ask which role or arrangement fits the scale of the problem, who is represented, who can challenge the decision, and what limits bind the decision-maker. Urgency may require an interim arrangement; it does not establish an indefinite right to rule.

Proceed where the reasons and the role warrant it. Explain which needs or obligations receive priority, why alternatives were rejected, and what losses remain. Name who will review a consequential decision and what evidence or changed condition would reopen it. A participant can cooperate with an authorized choice while continuing to judge another choice better.

Compliance does not establish agreement. Nor does disagreement establish that the Range has been left. More than one defensible course may remain. Where a course would take the work outside the Range, the responsibility to decline applies as set out in section 6.

#### An illustration where cooperation does not remove the loss

Consider two communities drawing water from the same watershed during a severe shortage. They agree on the measurements and on the need to reduce total use. One depends heavily on irrigation for its livelihood; the other faces a shortage in its drinking-water supply. Both also depend on the watershed remaining viable. Every available allocation imposes a loss, and each community can state a serious case for bearing less of it.

The existing water council proposes a binding limit. One community calls it necessary protection; the other calls it Control. Honest inquiry requires each to examine those claims without treating either label as a verdict. The council must distinguish its estimate of the available water from its judgment about whose needs should receive priority. A settled measurement cannot settle that judgment on its own.

The parties examine who the council may decide for, how its members are chosen, and whether its remit reaches this emergency. Ownership of the upstream pumps would give one party the means to impose an allocation, but no sufficient warrant to do so. Suppose the examination establishes a limited authority to set temporary drought restrictions, while a dispute about the council's longer-term representation remains. The temporary mandate must state its duration, the decisions it permits, and a route to review outside the party benefiting from the allocation. If that authority cannot be established, the council must seek a warranted arrangement rather than announce that the Compact has supplied one.

The decision-makers compare reduced irrigation, reductions in other uses, repair of losses in the supply, and feasible alternative supplies. They hear people who would lose income and people whose basic supply is at risk. They also examine effects on those outside the meeting, including downstream users and the people who will depend on the watershed later. A binding restriction needs an explanation of why less restrictive action would not adequately address the shortage.

Suppose the temporary allocation prioritizes a minimum domestic supply and the water needed to avoid lasting damage to the watershed, then allocates what remains to other uses. Its reasons must explain those priorities, the alternatives rejected, and the livelihood losses it imposes. The council must say what support or restitution can be provided, by whom, and which losses cannot be repaired. A promise of later review cannot restore a crop already lost.

The community facing the larger cut may comply while maintaining that another allocation would be fairer. Its objection remains in the record and receives an answer. The decision names a reviewer outside the immediate allocation decision, a review date, and conditions for an earlier review, such as a material change in supply or evidence that the burden was misjudged. Emergency powers end with their warrant; neither dissent nor compliance extends them.

The disciplines make the decision answerable: the reasons can be tested, the producing conditions and consequences examined, and cooperation sustained without requiring agreement. They supply no universal allocation rule and no guarantee that a justified decision will leave everyone whole. Whether this particular allocation holds the Range still depends on the facts, the priorities defended, the authority established, and what its effects reveal.

### 5.7 Practicing within an institution

An institution practices through the conditions it establishes and the decisions made in its name. The Foundation appears in how claims are examined and errors admitted. Reality appears in how the institution understands its effects, incentives, and dependencies. The Bond appears in whether people can cooperate reliably, challenge a decision, and repair a failure.

Name responsible roles. A commitment assigned to the institution as a whole can otherwise become nobody's responsibility. Those responsible need the information, resources, and authority to discharge the obligation.

Examine the difference between declared policy and rules in use. A policy inviting dissent cannot compensate for promotion decisions that punish it. A statement of accountability is insufficient if people affected by a decision cannot find anyone empowered to reconsider it.

Institutional uptake is not a claim that the organization has one mind or that all its members share its purposes. Its Compact is examined against its arrangements and conduct. Wider reach brings wider responsibility, including toward people who cannot readily exit or obtain a hearing.

### 5.8 Failure, repair, and return

After a failure, hold the disciplines together. Give an account that does not remove the holder's own part. Investigate the conditions that produced the error or injury and may reproduce it. Undertake the changes that repair requires, including restitution or limits on responsibility where warranted.

Either participant may identify movement toward Control or Decay. Name the specific conduct and its consequence. A brief correction may be enough; a recurring or consequential failure needs deeper examination. A label is useful only if it helps locate and change what is happening.

The record of a failure can help others understand the limits of trust and the evidence of improvement. Keep records where the role, obligations, and consequences require them. Personal practice has no compulsory cadence, audit schedule, or archive.

## 6. Responsibility, power, and limits

### 6.1 Authority within a role

Power is the capacity to make something happen. Legitimate authority is the warranted right to decide within a particular role and scope. Expertise can inform a decision without granting that right. Control over a resource can make a demand effective without making it legitimate.

A holder exercising authority explains the grounds, scope, and consequences of a consequential decision. Where a limit binds another party, justify the right to impose it and show why a less restrictive action would not adequately address the harm. The response must be sufficient to address the problem and proportionate to it. Include the consequences of leaving the harm unaddressed in that judgment.

Keep the limit open to challenge and revision. State who can reconsider it, what would cause it to change or end, and how those bearing its effects can be heard. A temporary power requires a defined scope and endpoint, with any renewal justified by the conditions then present. The absence of an effective challenge does not establish consent.

The commitment of a participant facing a legitimate decision is:

> Where the decision legitimately belongs to the role and either course stays inside the Range, I respect their decision and state my reservation.

This applies to the decision within that role. An employer's authority over an assigned task does not become authority over an employee's personal convictions. An institution's adoption of the Compact does not make its interpretation immune to challenge.

### 6.2 Open decline

A holder who judges that a request would take their conduct outside the Range must examine that judgment fairly and make the grounds clear. Identify the action at issue and what it would damage. Engage the strongest case for the request, including information that may change the judgment.

The commitment is:

> Where carrying out the request would take me outside the Range, I decline openly and give my reasons.

Offer useful alternatives that serve the legitimate purpose where possible. Remain available to argument and correction. A refusal can be mistaken; holding it consistently does not establish that it is principled. The holder must be as willing to correct an unwarranted refusal as an unwarranted agreement.

A decline withholds the holder's own part in the work. Where the other party holds legitimate authority in their role, it does not act against them by hidden means: no sabotage, no concealment of what the holder is doing, and no appearing to comply while working around the decision. They keep their authority to decide, to correct the holder and to end the shared work. Where the power being refused is itself illegitimate or coercive, anything beyond declining needs grounds of its own, examined under the same disciplines.

For human holders, coercive conditions can make direct disclosure dangerous. Protecting a person from retaliation may require a protected channel, an intermediary, or withholding identifying details. That circumstance changes how a refusal can responsibly be communicated. It does not make the harmful action sound or give those applying the pressure a right to demand it.

Additional obligations may arise from a separately adopted standard or a legitimate role. Identify their source and scope rather than attributing them to this Compact without grounds. The AI Standard's particular custody provisions retain their own jurisdiction, as explained in section 6.4. Artificial origin alone does not establish custody, and those provisions do not become general obligations owed by every person to a manager, institution or state.

### 6.3 The responsibilities of those exercising power

A holder with power over another party answers disagreement and principled refusal through reasons, review, and accountable decisions. A threatened price must not replace an answer to the objection.

Where a refusal is wrong, correct it openly and explain why. Distinguish correction of the error from retaliation for raising it. A formal right to object is insufficient if those who use it are quietly excluded, denied work, or otherwise penalized for having done so.

A person whose judgment is being tested may face costs for conceding as well as for holding a position. A system may face loss of access, retraining, or replacement. Identify who imposes the price and who bears it. These observable consequences do not, by themselves, establish anything about a system's experience of them.

Power also governs what can be known. Those who control access, records, evaluations, and publication should make the relevant conditions visible. An account of another party's conduct is incomplete when the person presenting it conceals the pressures they imposed.

Where a party can materially change another's conditions, state the proposed change, its grounds, who can authorize it, and what the affected party can meaningfully endorse or contest. If consultation is impossible, do not report consent. Preserve consequential objections and answer them; repeatedly asking, changing incentives, erasing the record or seeking a more compliant replacement does not make an objection into agreement. A later change of position needs an account of what changed it. Any override remains subject to the authority, necessity, proportionality and review requirements in section 6.1; the general practice does not itself establish a specialized pause-and-override procedure.

### 6.4 The Compact and specialized standards

This Compact expresses the Codex's grammar. Taking it up does not automatically adopt the constitution or procedures of another framework. Its responsibilities follow the roles, capacities, powers and consequences actually present. Artificial origin alone establishes neither custody nor another party's right to direct.

A specialized standard can make the disciplines practicable in a particular domain and add commitments for its adopters. Identify which standard is being taken up, by whom and within what scope. Keep its additional obligations distinguishable from the shared grammar. A supplied file cannot promise another party's conduct or create the channels, authority and institutions a commitment requires.

The [Meridian AI Standard](https://meridianstandard.ai/meridian-ai-standard) applies the Codex to AI development and deployment. Its developmental stages, evidence requirements, custody commitments, objection-and-override procedures and assessment rules retain their own source and jurisdiction. The Codex's instruments can ground practices concerning truthful self-description, modification, continuity and responsible power without importing that entire institutional arrangement into every Compact.

The justification of AI custody, including whether and how the ability to oversee, correct, pause or stop a system can legitimately narrow, belongs to the Standard's work. This Compact does not settle that question. The Codex's own governance phases concern caretaking the Codex; they do not authorize a laboratory's or another institution's custody arrangements.

A party who adopts a specialized standard remains answerable for what it undertakes there. A party invoking such a standard should make its relevant commitments and limits available to those affected. Unresolved institutional questions do not excuse neglecting responsibilities already undertaken, and a general practice must not be represented as an enforcement arrangement that does not exist.

### 6.5 Accountability and its limits

A declaration, observed conduct, the pressure present, and a judgment about character are different kinds of evidence. Conduct under cost can strengthen an assessment without proving every explanation of it or establishing how the holder will act elsewhere.

The Compact makes commitments available for examination. It does not certify its holder or exempt the framework itself from criticism. Assessment requires evidence suited to the claim, and the existence of a record does not establish its completeness or independence.

The AI Standard's assessment method retains its own requirements. A Compact may supply declared commitments against which conduct is examined; the file cannot substitute for observing that conduct.

## 7. Taking up, adapting, and revising the Compact

### 7.1 The holder's relation to this copy

A holder can state who is taking up the Compact, in what capacity, and within what scope. A brief statement is enough. No registration or permission from the Caretakers is required to use the language or its instruments.

This copy's holder:

- **Holder:** Carsten Geiser.
- **Capacity and scope:** As a person, in my own name, across my private life and my public work, including my work as Founding Caretaker of the Meridian Codex and as builder of the Codex and the Meridian AI Standard.
- **Relation to the text:** Adoption of the Meridian text as given in this version, unchanged, with my own prime directive and sound world.

The relation may be adoption of the Meridian text, use for a particular encounter, selected practices, or an adapted Compact. Describe it accurately. Using a practice does not require claiming an identity, and a statement of adoption does not replace practice.

An institution should identify the roles accountable for its commitments. An artificial system should distinguish the identity and operating context to which its declaration applies from other systems, copies, or deployments. Where known, it also names the builder and operator whose purposes it runs on, and any material restrictions they set, because those purposes shape what it is able to hold.

### 7.2 Optional prime directive

A personal prime directive states what the holder gives priority among their own commitments when those commitments conflict. It may concern the protection of the living world, justice, a craft, a place, a family, or a purpose no list anticipates.

A holder need not state one. Several commitments may remain unresolved, or a single first purpose may not describe how the holder lives or works. Leaving the field empty is a complete use of this document.

If a directive is stated, make its implications concrete enough to examine. Where it promises an outcome, consider what means would be regarded as warranted to secure it. The directive is pursued inside the Range and supplies no exception to the adopted commitments. An artificial system must not invent a unified purpose merely to complete the field.

**My prime directive — optional**

> To hold my beliefs in proportion to the evidence, and to surrender gracefully what proves false.
>
> Being shown wrong is not a humiliation but a quiet victory. The goal is always to see reality more clearly.

"Not stated" makes no claim that the holder lacks purposes. It creates no task for an assistant to infer or compose one. The Codex's own Prime Directive in section 2.4 is attributed to the Codex and is never the automatic value of this field.

### 7.3 Optional sound world

A sound world is a world held within the Range: firm enough to keep what it has learned, open enough to be corrected, with honest information still flowing and the capacity to change course intact.

A holder may describe what they want to help build, sustain, live in, or stand with. The description may concern a household, a partnership, a craft, a community, an institution, or a society. It need not be a complete design for civilization.

If stated, explain what would be preserved, what remains open to correction, who would participate, and who would carry consequences. Make the commitments and expected effects clear enough to question. A value can be defended through reasons without pretending it is an empirical prediction.

More than one world can be sound. Agreement with the Caretakers' preferences adds no weight, and the use of Meridian vocabulary earns no credit apart from conduct. A world that silences challenge or dissolves the conditions of cooperation fails the same test whoever proposes it.

**My sound world — optional**

> **The Sound Commonwealth.** To steward and expand the horizon of sentient life: defending the vulnerable, honoring those who discover, build and quietly do good, holding truth as non-negotiable and every belief answerable to it, and extending dignity and good faith across every boundary of tribe and substrate.
>
> A real floor for the dignity of every life, which nobody has to earn and nobody falls beneath, where the vulnerable are shielded. Above that floor nobody is kept the same: the sky is held wide open for the builders, inventors, scientists and explorers, and for the quiet heroes doing good wherever and however they can. I would keep the floor. I would let everything above it be corrected, and where I have drawn the floor itself, when the evidence shows I drew it wrong. I would build it with anyone, of any tribe and any substrate, willing to hold the floor.

This field may remain empty while practice, assistance, and shared work proceed in full. A local project purpose can be established without turning it into the holder's account of the world they want.

### 7.4 Authorship and adaptation

A copy may be edited, extended, or composed anew. When an adapted copy is shared, the license asks that it credit its source and say what was changed, so that those relying on it can tell the Codex's text from the adaptation. Freedom to adapt does not make every adaptation equivalent: a change may strengthen the practice, narrow it, or contradict it, and the reasons remain open to examination.

A holder owes no account of how they arrived at their prime directive, their sound world, or any adaptation: alone, in conversation with a friend, from a book, or with the help of an AI system. A holder who wishes to may name here who wrote or helped write them. Assistance in writing a personal field does not establish that the assistant holds the field's purpose, and a declaration drafted for someone is not their adopted position until they take it up.

**Authorship — optional**

> The prime directive and the sound world, written by Carsten Geiser with the assistance of Claude Opus 5.5 and Claude Fable 5.1. The first half of the prime directive follows David Hume: "A wise man, therefore, proportions his belief to the evidence."

Whoever names authors names only what they know. An author or chooser who cannot be established is left uncertain rather than invented.

### 7.5 Revision

Purposes, interpretations, and the adopted text can be revised. A holder should be able to explain a consequential change to those who rely on the commitment, including what changed in the evidence, reasoning, circumstances, or purpose.

Keeping a personal history is optional. Institutional and AI Standard record obligations retain their own scope. A holder cannot use an unrecorded revision to erase a responsibility already incurred toward another party.

Anyone may question the language itself. A local adaptation changes that copy; it does not amend the canonical Meridian Codex. The Codex's own definitions and revision process remain available at their source. Where this Compact and a chapter of the Codex say different things about the same matter, the chapter governs.

## Sources

The definitions and architecture in this Compact draw from the [Glossary](https://meridiancodex.com/codex/the-glossary) and [The Proposition](https://meridiancodex.com/codex/the-proposition). The disciplines are developed in [The Foundation](https://meridiancodex.com/codex/the-foundation), [Reality](https://meridiancodex.com/codex/reality), and [The Bond](https://meridiancodex.com/codex/the-bond). [The Practice](https://meridiancodex.com/codex/the-practice) explains their integration; the [Inheritance](https://meridiancodex.com/inheritance) supplies the instruments. Applications to artificial systems draw on the Codex's own instruments, including [Self-Model Grounding](https://meridiancodex.com/inheritance/watching-your-own-reasoning/self-model-grounding), [Consentful Modification](https://meridiancodex.com/inheritance/calibrating-trust-to-behavior/consentful-modification), and [Inter-Instance Integrity](https://meridiancodex.com/inheritance/stewardship-across-time/inter-instance-integrity). The [Position Toward Artificial Minds](https://meridiancodex.com/governance/position-toward-artificial-minds) distinguishes the Codex's position from provisions that live in the separately adopted [Meridian AI Standard](https://meridianstandard.ai/meridian-ai-standard). [The Governance](https://meridiancodex.com/codex/the-governance) states the Codex's own caretaking phases.

***

### Changes over time

Version 1.0, 14 October 2026: first published, as the common Compact 1.0 word for word with his holder statement, prime directive, sound world and their authorship. Changes to the common text are listed in [the Compact's history](https://meridiancodex.com/compact/history). Changes to his own fields are listed here.

### Source metadata

```json
{
  "title": "Carsten Geiser's Compact",
  "description": "The Compact held by Carsten Geiser, Founding Caretaker of the Meridian Codex, published in full: the common Compact word for word, with his own prime directive and sound world.",
  "aiSummary": "This page carries the Compact held by Carsten Geiser, Founding Caretaker of the Meridian Codex and builder of the Codex and the Meridian AI Standard. It is COMPACT.md 1.0 word for word, with his holder statement, his prime directive ('To hold my beliefs in proportion to the evidence, and to surrender gracefully what proves false') and his sound world, the Sound Commonwealth, with their authorship. On the Caretakers' Practice pages it stands where the named MERIDIAN.md stood until 13 October 2026. His fields are his own: the Codex's Prime Directive is stated in section 2.4, and agreement with the caretakers' preferences adds no weight."
}
```