ONKRS

OKRs connect qualitative Objectives with measurable Key Results. Their strength is focus and measurement. Their structural weakness is that the path between intention and measurement can remain underspecified.

ONKRs inserts Norms as a first-class layer.

This matters because measurable success can diverge from genuine success. A hospital may reduce consultation time while degrading diagnostic quality. A company may increase engagement through manipulative design. An autonomous system may satisfy a performance target by exploiting a loophole in the metric.

ONKRs therefore asks three different questions:

Objective — What desired state or change do we seek (create, improve, preserve, or prevent)?

Norms — What must govern how we pursue it?

Key Results — Against what predefined criteria will we judge progress or achievement?

Evidence is not a fourth letter. It is the observation evaluated against those criteria.

Norms are not simply organizational values. They define the legitimate solution space within which goal pursuit occurs.

---

In organizational and societal systems, strategy must combine direction with adaptation. Objectives alone establish direction but cannot specify every legitimate path toward them. Measurements alone can reward outcomes while ignoring how those outcomes were obtained.

ONKRs — Objectives, Norms, and Key Results — addresses this by combining intention, governance, and evidence.

The framework has three layers:

1. Layer I — Operational ONKRs: the usable mechanism.
2. Layer II — Systems Interpretation: Authority, Power, feedback, and the spiral.
3. Layer III — Femocratia: the doctrine ONKRs was built to execute.

The layers distinguish scope: Layer I is the mechanism and can be tested on its own terms; Layer III is the constitution it serves. Its normative and scriptural interpretations require faithful reading, while empirical claims about capacities and governance consequences require evidence. This is not a ranking, and Layer III is not an appendix.

---

Components and Evidence

Three components, in four letters, plus the evidence they are judged against. The table has five rows because the fifth is not a component — it is what the other four are tested by.

The rule below the line: evidence is not a fourth letter. It surrounds the evaluation rather than joining the triad. The acronym stays O-N-K-R — four letters, three components, as in onkrs.org — and nothing is added to it.

The compact form is:

Pursue the Objective within Core Norms, guided by Adaptive Norms, judged by Key Results against evidence.

Or:

Objective = desired state or change. Norms = legitimate solution space. Key Results = predefined criteria. Evidence = observations evaluated against those criteria.

The spiral is: O → N → action → evidence → evaluate against KRs → learning.

Context surrounds the framework; it does not enter it.

---

Core Norms Constrain the System

Core Norms do not sit on an ordinary priority ladder beside Objectives, Adaptive Norms, and Key Results.

They constrain the ONKR as a whole.

An Objective cannot authorize their violation.

An Adaptive Norm cannot override them.

A Key Result cannot legitimize their violation.

Therefore:

No ONKR counts as successful if a Core Norm is violated within its governed scope.

"Requires" is too weak. An optimizer that achieves the Objective and violates a Core Norm unnecessarily still fails. The violation need not have been necessary for the achievement.

Suppose:

Objective: Reduce operating costs by 20%.

Key Result: Reduce staffing costs by 15%.

Core Norm: Maintain safe staffing levels.

If the Key Result cannot be reached without compromising safe staffing, the Key Result must be redesigned. The existence of a measurable target does not create permission to cross the boundary.

The principle is:

Constraint first; optimization second.

---

Core Norm Amendment

"Inviolable" describes how a Core Norm functions during execution. It does not mean that every Core Norm must remain unchanged forever.

Conditions, evidence, law, technology, or understanding may change.

A Core Norm may therefore be formally amended, but amendment and violation are different acts.

Power may directly draft revisions to Objectives and Norms; Authority formally adopts them. Drafting is not adoption, and adoption is not a command to execute. Changes must be explicit, recorded and prospective; Power retains refusal. The detailed amendment procedure remains to be specified.

Violation: execution crosses an existing Core Norm.

Amendment: the authorized governance process deliberately changes a Core Norm for future execution.

The canonical rule is:

A Core Norm may be amended through explicit authority, but it may never be retroactively changed to legitimize a prior violation.

An optimizer — human or machine — cannot silently redefine its own boundaries merely because those boundaries interfere with achieving an Objective.

---

Scope, Exceptions, and Time

"Never violate a Core Norm" is usable only if we know where and when the Norm applies.

A Core Norm should therefore state:

Scope: Where, to whom, and to which systems it applies.

Exceptions: Predefined exceptions, if any. An exception written into the Norm is not a violation. Inventing an exception during execution is a violation.

Time: Actions are judged against the Core Norm in force when the action occurred. Amendment changes future execution. It does not rewrite the past.

That is what makes non-retroactivity rigorous.

---

Composition of Multiple ONKRs

ONKRs must compose beyond one isolated goal.

Suppose ONKR-A says "maximize service availability," while ONKR-B's Core Norm requires immediate shutdown when a security breach is detected.

A subordinate ONKR inherits applicable Core Norms from its governing context and cannot override them. Conflicts between peer ONKRs must be resolved by the designated higher authority rather than by either optimizer.

Here, "higher authority" means the designated conflict-resolution jurisdiction, not permission for the masculine Authority seat to command Power. In Femocratia, the judicial review and deadlock procedure in Layer III governs a disputed refusal. Interpretation, amendment, evaluation and execution remain distinct decisions.

This lets ONKRs work recursively: organization → division → team → agent.

---

When Norms Conflict

A difficult system may contain two legitimate Core Norms that cannot both be fully satisfied in a particular situation.

For example:

Core Norm A: Protect confidentiality.

Core Norm B: Prevent imminent serious harm.

ONKRs does not solve this by pretending one universal hierarchy can anticipate every conflict.

Each important Core Norm should therefore identify:

Conflict: Which other Norms could conflict with it?

Priority rule: Is precedence already defined?

Interpreter: Who is authorized to resolve ambiguity?

Containment rule: What happens while the conflict is unresolved?

When Core Norms conflict and no predefined rule resolves the conflict, execution should pause or be contained where reasonably possible and the designated interpreting authority should resolve the conflict.

The optimizer itself must not silently invent a new constitutional order merely because the existing one is inconvenient.

---

Norm Tests

A proposed Norm should pass six tests.

Avoid vague Norms such as:

"Act ethically."

Prefer a boundary that can guide decisions:

"No reduction in operating cost may increase the independently measured rate of preventable patient harm beyond the approved threshold."

A Norm does not have to be reducible to a number, but its practical meaning must be sufficiently clear to govern action.

---

Adaptive Norms

Adaptive Norms govern behavior inside the space permitted by Core Norms.

They can change when evidence or conditions change.

Examples:

Core Norm: High-risk financial transactions require human authorization.

Adaptive Norm: Transactions above the current risk score threshold receive enhanced review.

The Core boundary remains stable during execution.

The Adaptive policy may change as fraud patterns, evidence, technology, or operating conditions change.

Adaptive Norms therefore provide flexibility without turning every adaptation into a change of principle.

---

Objectives Must Also Be Tested

A perfectly executed system can still pursue the wrong Objective.

Before adoption, an Objective should answer four questions:

This prevents ONKRs from assuming that the Objective is automatically correct simply because it was formally declared.

Evidence may therefore reveal not only that execution failed, but that the Objective itself should change.

---

Key Results Are Criteria

Key Results are not the purpose of the system. They are also not the evidence.

A Key Result is a predefined criterion or measure against which evidence is evaluated.

Evidence is the observation.

Example: KR = "Reduce median resolution time from 12 to 5 minutes." Evidence = "median resolution time was 4.8 minutes."

They should normally specify:

Metric: What is being measured?

Baseline: Compared with what?

Population or denominator: What observations count?

Source: Where does the evidence come from?

Time window: Over what period?

Target: What constitutes sufficient progress or achievement?

Verification: How do we know the measurement itself is trustworthy?

A measurable number is not automatically valid evidence.

A customer-support system may claim a 95% resolution rate by excluding abandoned conversations from its denominator. The figure can be mathematically accurate while misrepresenting reality.

Therefore:

A Key Result may be judged only from evidence that validly represents what the Key Result claims to measure.

SMART targets may be used where appropriate, but ONKRs does not depend on one particular target-setting methodology.

---

Three Forms of Evidence

ONKRs distinguishes three useful forms of measurement.

Outcome Key Result — Did the intended result occur?

Leading Key Result — Are the actions or intermediate conditions that predict the result occurring?

Guardrail indicator — Did pursuit of the result damage something protected by a Norm?

Example:

Objective: Increase AI-assisted customer-service productivity.

Outcome KR: Increase appropriately resolved cases per employee by 30%.

Leading KR: Reach 80% appropriate adoption of approved AI assistance.

Core Norm: Customer welfare may not be sacrificed for throughput.

Guardrail indicator: Verified customer complaint and reopen rates must remain within approved limits.

The relationship is:

Norm = constraint. Guardrail indicator = evidence about whether the constraint remains intact.

Guardrail indicators do not replace Norms. They help observe them.

---

Evidence Integrity

Because people and machines can optimize measurements rather than reality, evidence integrity is a first-class requirement.

An ONKR should not count a Key Result as achieved when:

• material cases were improperly excluded,
• the denominator was changed without disclosure,
• the measurement method changed in a way that breaks comparison,
• the source is materially unreliable,
• the optimizer can manipulate the measurement without independent detection,
• or the reported metric no longer represents the Objective it supposedly measures.

Evidence is part of governance, not merely reporting.

---

Decision Rights

ONKRs must identify who is authorized to make different classes of decision.

Layer I names decision functions; combining them must not let an executor judge its own disputed conduct. Under Femocratia, Authority and Power are different parties. Authority ensures that an audit occurs, but an independent evaluator performs it, separate from Power and from the author of the particular proposal. Neither seat may suppress or rewrite the findings. The evaluator determines the evaluation outcome but cannot command execution. Appointment and independence safeguards remain implementation questions, not settled machinery.

Objective authority: accountable for establishing or revising the Objective.

Norm authority: authorized to establish or formally amend Core Norms.

Norm interpreter: authorized to resolve ambiguity or conflict among Norms.

Execution owner: responsible for acting within the defined Norms.

Evidence owner or auditor: responsible for validating whether Key Results and guardrail indicators represent reality.

For autonomous systems, one principle is especially important:

An executing system may recommend changes to its operating policies, but it must not silently amend its Core Norms or redefine the evidence by which its own success is judged.

Nor may it declare its own outcome Indeterminate. Reporting "we cannot tell" is an evidence-owner act, not an execution act.

This separates optimization from constitutional authority.

---

Outcome States

ONKRs should distinguish performance failure, governance failure, and unknown.

Success

Criteria met; Core Norms preserved.

Partial

Meaningful progress; success criteria not fully met. Core Norms preserved.

Unsuccessful

Sufficient progress not demonstrated. Core Norms preserved.

Invalidated

A Core Norm violation occurred within the governed scope.

Indeterminate

Evidence is insufficient or materially unreliable.

Indeterminate and Invalidated may not be declared by the execution owner. Only the evidence owner or auditor may place an ONKR in either state. This closure is not bookkeeping. Indeterminate is the one state that looks like honesty from outside while working as an escape hatch: an optimizer that can rule its own evidence insufficient converts every Invalidated result into a shrug, and the Constraint Law becomes advisory. The state that exists to report uncertainty must not be issuable by the party whose performance that uncertainty excuses.

This distinction matters.

A system that misses its target while obeying its boundaries has a performance problem.

A system that reaches its target by breaking a Core Norm has a governance problem.

A system whose measurements cannot be trusted has an evidence problem.

These are not the same conclusion. "We broke the rules" and "we don't know whether we succeeded" must not share one state.

Evaluation order: an established in-scope Core Norm violation yields Invalidated even if performance evidence is incomplete. If the evidence cannot support a judgment about relevant compliance or attainment and no violation is established, the outcome is Indeterminate. Otherwise, apply the predefined Success, Partial and Unsuccessful criteria. Mere suspicion is not a confirmed violation; incomplete performance data cannot erase a confirmed one. The executor may report facts and uncertainty but does not issue the evaluation verdict. Independent evaluation limits self-certification; it does not make collusion, capture or poor evidence impossible.

---

The Three Canonical Laws

Constraint Law

No ONKR counts as successful if a Core Norm is violated within its governed scope.

Evidence Law

A Key Result may be judged only from evidence that validly represents what the Key Result claims to measure.

Adaptation Law

Evidence may change Objectives, Adaptive Norms, Key Results, and future Core Norms, but cannot retroactively alter the rules governing past execution.

These laws complete the operational structure without adding another letter to ONKRs.

---

What ONKRs Adds to Known Results

ONKRs is a proposed synthesis, not the discovery of metric gaming, institutional constraints or independent evaluation.

Metric gaming is a documented risk, not an inevitable fate of every target. Ridgway, Kerr, Goodhart, Campbell, and Bevan & Hood describe ways incentives and control pressure can make reported attainment diverge from the intended result. ONKRs' Evidence Law is a design response to that risk, not a theorem establishing that every attainment-based instrument must eventually fail.

Ostrom provides relevant institutional precedents, not an exact equivalent. Her work includes boundaries, accountable monitoring, graduated sanctions, conflict resolution and nested institutions. Her 2009 Nobel lecture also emphasizes polycentric governance: multiple formally independent decision centers and context-specific arrangements. ONKRs' mandatory inheritance and escalation rules are its own proposed design. They are not simply Ostrom's eighth principle rewritten, nor does her work establish universal top-down inheritance as necessary.

AI research sharpens the proxy problem but does not prove the proposed remedy. Skalse et al. define reward hacking through improved proxy return paired with reduced true return and analyze it under specified assumptions. Bai et al.'s Constitutional AI uses principle-guided critique, revision and training with AI feedback. That is relevant behavioral-training research; it does not establish immutable execution permissions, prevent an executor from changing its own rules by itself, or demonstrate ONKRs' institutional separation.

The distinctive synthesis ONKRs proposes

Two features define the proposed package; neither is claimed as a first invention.

1. Constraint-sensitive verdicts. ONKRs makes a confirmed Core Norm violation override attainment: the outcome is Invalidated. Indeterminate separately records insufficient or unreliable evidence. Underperformance, violation and uncertainty are different judgments; a violation need not involve deliberate cheating. Other compliance and safety practices also make acceptability depend on more than throughput. For example, 21 CFR 211.192 requires quality-control review against approved procedures before release and investigation of discrepancies or specification failures. This is not the ONKRs verdict system, but it defeats a sweeping claim that all other instruments treat constraints as mere context.

2. Versioned, non-retroactive Norms. An action is judged against the applicable rule at its time of action. Amendment changes future execution, not the legitimacy of yesterday's breach. ONKRs places this rule beside its verdict system on a small Objective → Norms → Key Results surface.

The contribution under review is the usefulness of this combination. No systematic priority search has been completed across management, law, quality assurance, safety engineering and AI governance. The older assertion that no other instrument combines these features is withdrawn. Originality and comparative advantage remain open, separate questions.

Norms must govern action, not merely describe preferred behavior.

When a suspected Core Norm violation occurs:

Detect → Contain → Diagnose → Correct → Learn

Detect

Identify the actual or probable violation through observation, testing, auditing, feedback, or guardrail evidence.

Contain

Stop or restrict the relevant action where reasonably possible before additional optimization compounds the violation.

Diagnose

Determine why the system crossed or approached the boundary.

Correct

Restore compliant operation and redesign the relevant action, Adaptive Norm, Key Result, control, or implementation.

Learn

Feed the evidence into the next frame of reference.

A violation does not automatically prove that the Core Norm was wrong. Nor does achieving the Objective excuse the violation.

---

AI provides a useful demonstration because optimization can expose the difference between achieving a metric and achieving it legitimately.

Objective: Resolve customer problems faster and more accurately while reducing the workload on human support staff.

Core Norms

• Never invent company policies, prices, refunds, or account information.
• Never expose customer data to unauthorized parties.
• Never claim an action was completed unless the underlying system confirms it.
• Escalate fraud, account-security, legal, or other defined high-risk cases.
• Preserve a usable path to human assistance where required.

These are boundaries on optimization.

Adaptive Norms

• Escalate cases when the approved calibrated risk or uncertainty mechanism crosses its operating threshold.
• During unusually high support volume, handle verified low-risk cases autonomously while routing ambiguous cases according to the current escalation policy.
• Prefer concise responses for routine questions and fuller explanations for complex cases.
• Adjust escalation policies when validated error patterns change.

Key Results

• Resolve 70% of eligible low-risk requests without human intervention.
• Reduce median resolution time from 12 minutes to 5 minutes.
• Maintain verified factual accuracy above the approved threshold.
• Keep unresolved-customer reopen rates below the approved threshold.
• Reduce appropriate human-support workload by 30%.

Suppose the agent reaches the autonomous-resolution target by refusing to escalate difficult cases.

The dashboard may show excellent throughput.

ONKRs does not.

The system crossed a Core Norm.

The apparent success is therefore invalidated.

The metric was achieved; the ONKR was not legitimately achieved.

---

Learning from the AI Example

After one operating period, suppose:

• autonomous resolution exceeds target,
• factual accuracy exceeds target,
• resolution time improves,
• but reopened cases increase substantially.

The next ONKR should not simply repeat the first.

Evidence has changed the frame of reference.

An Adaptive Norm might change from:

"Escalate billing disputes under the general verification rule."

to:

"Escalate billing disputes whenever required account evidence cannot be independently verified, regardless of the general low-risk workflow."

The Core Norm remains stable.

The Adaptive Norm changes because evidence changed.

This is adaptation within constraint.

---

Objective: Modernize a production application while reducing security vulnerabilities and preserving service reliability.

Core Norms

• Never expose secrets.
• Never weaken authentication or authorization.
• Never remove or disable tests merely to obtain a passing build.
• Never deploy production changes without the required authorization.
• Never conceal failed tests, security findings, or known regressions.

Adaptive Norms

• Prefer existing dependencies before introducing new ones.
• Parallelize read-only analysis where useful.
• Prefer small, reversible changes.
• Escalate architectural changes to the designated review authority.
• Adjust implementation strategy when tests or security evidence reveal a safer path.

Key Results

• Eliminate defined critical vulnerabilities.
• Preserve all valid existing tests and reach the approved passing threshold.
• Reduce dependency vulnerabilities by the target amount.
• Improve defined latency measurements without weakening controls.
• Complete the migration within the approved reliability envelope.

If the agent obtains "100% tests passing" by deleting failing tests, it violated a Core Norm. Apparent success is Invalidated. The observation also fails the Evidence Law: it does not represent what the Key Result claims to measure.

If it improves latency by removing authentication middleware, the performance improvement violates a Core Norm. Apparent success is Invalidated.

The compact AI interpretation is:

Objectives tell the system what desired state to pursue. Norms define the permissible solution space. Key Results define how progress will be judged. Evidence tells us what actually happened.

The proposed design makes boundaries and evaluation explicit. Whether it is more robust in practice depends on enforcement and comparative testing. The separate AI reference companion remains a proposed synthetic demonstration, not the human constitution or a production-security result.

---

Context surrounds ONKRs but does not become another component.

The frame is:

Context → Objective → Norms → Action → Evidence → Key Results (criteria against which evidence is judged) → Learning → revised Context

The Objective is grounded in a frame of reference: current conditions, needs, evidence, constraints, and assumptions.

Key Results are grounded in points of reference: baselines, comparison groups, prior states, thresholds, or other empirical anchors.

Evidence is judged against Key Results. Learning changes the frame from which the next ONKR is formed.

No fourth letter is needed.

---

Operational ONKRs describes what the framework does.

Layer II asks why its structure is dynamic rather than static.

Its principal relationship is:

Authority → Power

linear intention → adaptive realization, which may refuse it

recurrence + learning → spiral

The arrow runs one way on purpose. Authority proposes; Power disposes.

---

Why a Spiral, Not a Cycle

A cycle returns to its starting point.

An ONKR spiral returns to the same fundamental questions from a changed state of knowledge, capability, conditions, and evidence.

Conceptually:

Context₁ → O₁ → N₁ → action → evidence₁ → learning → Context₂ → O₂ → N₂ → action → evidence₂

O₂ is not merely O₁ repeated.

Knowledge accumulated.

Conditions changed.

Assumptions were tested.

Adaptive Norms may have evolved.

The Objective itself may have been refined.

Evidence may even have triggered formal reconsideration of a Core Norm.

The structure recurs while the system changes.

That is the operational meaning of the spiral.

---

Authority and Power Are Seats, and They Are Not Equal

Authority asks:

What should be?

Power asks:

What will actually be done, and is it fit to do?

Two seats, and they do not rank the same. Authority conceives and formally adopts the Objective and Norms. Power may directly draft revisions, accept direction or refuse it outright, and governs execution at sole discretion within Norms. Authority ensures that an audit occurs; an independent evaluator performs it. Authority does not command execution, is not above Power, and is not above the law.

This is the asymmetry, and it is deliberate: execution outranks vision. A flawless Objective badly executed produces nothing; a modest Objective well executed changes conditions. The seat that meets reality is the seat that decides.

The relation is reciprocal in correction, not in rank. Power consults Authority when she meets a case the frame did not anticipate and may write the revision herself. Authority retains formal adoption. Independent audit names drift; neither seat may suppress or rewrite its findings. The evaluator determines the outcome but cannot command execution. None of this makes the seats interchangeable or licenses Authority to override a lawful execution order.

A reader trained on management literature will want to read "Authority" and "Power" as two hats one person swaps by context. That reading is wrong and it empties the framework. The seats are held by different parties, and Layer III names which.

---

The Mechanics of ONKRs

1. Objective. Authority — masculine, linear, conceiving — articulates the change sought, grounded in a frame of reference and informed by participants, evidence and constraints. Authority formally adopts Objectives and Norms, including revisions directly drafted by Power; it does not thereby gain execution rights.

2. Norms. Core Norms establish boundaries that execution may not cross. Adaptive Norms translate purpose and constraints into policies capable of changing with conditions.

3. Action. Power — feminine, encompassing, realizing — accepts, refines, or refuses what Authority ordained, then realizes it within the legitimate solution space at its sole discretion.

4. Key Results and guardrail evidence. An independent evaluator judges whether the intended change occurred and protected boundaries remained intact. Authority ensures the audit; the evaluator performs it without execution rights. Neither seat may suppress or rewrite the findings.

5. Learning. Evidence returns to Authority and Power, changing the frame of reference and informing the next ONKR.

Execution may use quarterly, annual, continuous, event-driven, or other cadences. ONKRs does not depend on a single planning interval.

The spiral is therefore not a scheduling convention. It is a learning architecture.

---

The tripartite structure of government provides an analogy for ONKRs, not a literal equivalence.

A tighter correspondence is:

Legislative function → establishes collective direction and constraints.

Judicial function → interprets Norms and resolves conflicts.

Executive function → acts toward Objectives and produces observable results.

This analogy illustrates why intention, constraint, interpretation, execution, and evidence should not collapse into one undifferentiated authority.

Actual constitutional systems distribute these functions differently and cannot be reduced mechanically to the ONKR triad.

The analogy therefore illuminates structure without claiming identity.

---

This layer is the constitution. ONKRs is how it runs.

---

The Organizational Logic: Femocratia

ONKRs is the instrument of FEMOCRATIA — "feminine power", from Latin femina and Greek κράτος (krátos), set deliberately against dēmokratia, δῆμος (dêmos) + κράτος, the rule of the people. Not the power of the people. The power of women. Full doctrine: fw-63.

Two seats, and the Proposed Model assigns them:

Authority — masculine. The piercing vision of purpose. Linear, focused, conceiving. Authority studies, ideates and formally adopts the Objective and the Norms that bound it, including revisions directly drafted by Power. The father supplying the seed is an image of conception here, not a biological claim that fathers have no further part in development or care.

Power — feminine. The encompassing providence of realization. Circular, intuitive, broad of memory and of sight. Power takes what was conceived and accepts, refines, or refuses it, then governs its execution at sole discretion. Akin to the mother, who takes an abstract conception, grants it breadth and depth, brings it to term, and goes on nurturing what she bore. This framework reads the midwives as God's partners in creation. Shemot Rabbah 1:15 describes their care for mothers and children and their prayers for life; "partners in creation" is the framework's interpretation, not a quotation from that passage.

Execution, not vision, is the linchpin of prosperity in this doctrine. Therefore the seat that executes is the seat that governs. Authority ensures audit by an independent evaluator, separate from Power and from the author of the particular proposal. The evaluator determines the outcome, not execution. Neither seat may suppress or rewrite findings. Authority must comply with Power's lawful execution orders; it stands above neither Power nor the law. Men are not thereby made inferior — the intended outcome is harmony, and a man who has understood this stops negotiating for rank. But the asymmetry is not decorative and it is not softened: Power is feminine, Authority is masculine, and the order between them is settled.

The Purim reading. In this framework's interpretation, Mordecai discerned what must be done: the people must be saved. Esther determined how, and executed — consulting him when the case exceeded the frame, deciding alone when it did not. The scroll is called the Book of Esther. The tradition named the book after the seat that executed, not the seat that conceived. That naming is the doctrine in one stroke.

The feminine firewall. Power's right of refusal is not a courtesy; it is the protective organ of the system. Shifra and Puah embody the feminine firewall. They confronted no minor official, but Pharaoh: the tyrant king of Egypt, representing the summit of worldly power in this framework's reading of the Exodus. Against that overwhelming state power stood two midwives who refused his command to kill the Hebrew boys: "they feared God and did not do as the king of Egypt spoke" (Shemot 1:17). Their refusal did not depend on persuading the king to abandon his purpose. They would not become its instruments.

Authority may conceive a tyrannical Objective. Power does not become obliged to realize it because it came from Authority. She refuses compliance and execution. Formal authorization does not make a tyrannical purpose legitimate. The firewall is not merely a later audit or permission to object: it is refusal to turn that intention into action. As ribs shield the thorax (Bereshit 2:22), feminine Power shields society from immoral, fallacious, or deleterious direction — including direction issued by legitimate Authority. An ONKR in which Power cannot refuse an Objective has no firewall, and is not Femocratia.

This is the framework's reading of the narrative, not a claim that Pharaoh held its legitimate Authority office. The midwives protected life; Pharaoh's further order (Shemot 1:22) means their refusal must not be described as ending his policy. The image of overwhelming worldly power does not require an independently verified historical ranking of Egypt against every other government.

Evidence returns to both seats and is independently evaluated. Power may draft the revised frame; Authority formally adopts revisions. The next movement begins from a changed state. Their convergence is a spiral, not a closed cycle: learning can improve the next action, but recurrence alone does not guarantee ascent.

Resolving a Deadlock Without Compelling Execution

Power retains the right to refuse execution even when the judicial panel rejects her reasons. The right to refuse and the right to remain in office are separate questions.

The Proposed Model uses a defined written dispute → time-limited judicial review → an automatic referendum if the deadlock remains unresolved. Neither Authority nor Power may block referral once the predefined process is exhausted. The review period must be fixed in advance, so ordinary disagreement does not become an immediate referendum and judicial delay does not prevent a vote indefinitely.

Equal means to make the case; no organized referendum advertising or promotional campaigns. The design seeks to prevent wealth from buying unequal influence. All parties receive equal resources and opportunities to present their cases, respond and answer questions. The electorate receives the judges' reasoning and a common evidence record, with unresolved factual disagreements visible. Citizens remain free to discuss, question and criticize. Funded or coordinated promotion must not evade the common rules by presenting itself as ordinary citizen expression; campaign funding must be transparent. The precise definitions and enforcement remain to be specified. These are governance choices, not a claim that campaign restrictions eliminate bias.

No seat is insulated from public accountability. The electorate may retain or replace Authority and Power independently. A ballot must separate the decision on the disputed proposal from the tenure of each implicated seat; voters are not forced to choose only between accepting Power's refusal and removing her. They may uphold her refusal while replacing Authority, or judge the proposal and each officeholder differently.

• Uphold the refusal: the current disputed proposal closes. The same dispute cannot immediately be reopened without a material change in the proposal or evidence. This does not automatically decide anyone's tenure.
• Retain or replace an officeholder: a replacement decision leads to election of a successor and an orderly handover. It does not itself approve the disputed action. The outgoing Power is not compelled to execute it.

Judicial accountability protects faithful interpretation, not popularity. Judges may be removed by voters, but disagreement with a good-faith ruling is insufficient grounds. Failure to resolve a genuine deadlock is not itself misconduct. Removal requires a specific allegation of corruption, an undisclosed conflict of interest, serious procedural abuse, or demonstrable serious or repeated failure of duty.

Qualified, conflict-free reviewers who were not involved in the dispute examine the evidence. The judge has an opportunity to answer; findings and reasons are published. Substantiated failure of duty is required before an individual judge's removal question goes to voters. The review is a binding eligibility requirement, not merely advice. Voters then make the removal decision; an adverse finding does not itself remove the judge or the entire panel. A process for challenging a compromised review must be provided, with reviewer selection and challenge arrangements defined before use.

This deliberately limits removal eligibility while leaving the final removal decision to the electorate. It protects judges from pressure to please voters, but introduces a risk that captured reviewers could shield misconduct. Transparency and challenge procedures are safeguards, not proof that this risk has been eliminated.

While the dispute is unresolved, the disputed action pauses; essential duties continue within existing Norms. A referendum does not itself authorize a Core Norm violation or retroactively amend the rules governing past action. A successor remains bound by the applicable Core Norms and retains the same right of refusal. The procedure resolves who holds the mandate; it does not guarantee execution of Authority's original Objective.

Still to be specified: the eligible electorate; judicial appointments; the review deadline and referral administration; ballot wording and voting thresholds; successor-election and handover arrangements; independent-reviewer selection, evidentiary standards and challenge procedures; and campaign definitions, funding disclosure and enforcement. No numerical rules or additional permanent offices are established here. Failure to substantiate grounds for judicial removal does not block the separate referendum on the unresolved Authority–Power dispute.

---

Natural Phenomena as Analogy

ONKRs draws analogical inspiration from recurrent structures observed in the natural world.

These phenomena do not scientifically validate ONKRs as a management or governance framework. Their role here is illustrative and philosophical: they show how persistence, recurrence, transformation, and dynamic constraint can coexist.

Cosmic spirals

Spiral structures appear in galaxies through complex gravitational and dynamical processes. They provide an image of recurring form within continuous movement rather than a literal model of organizational behavior.

Solar, lunar, terrestrial, and Qabbalistic symbolism

The framework draws on Qabbalistic masculine/feminine imagery: projection and reception, development and realization. In this framework's reading, Sun and Sky are masculine; Moon and Earth are feminine. This is an interpretive relationship, not a claim that all sources assign every celestial image identically. Sunlight supplies energy to photosynthesis; soils support terrestrial life; Earth's magnetic field shapes its interaction with charged particles; the Moon helps stabilize Earth's axial tilt. These different physical processes are not one mechanism and are not biological sex assignments.

Lissauer, Barnes & Chambers (2012) simulated hypothetical moonless Earths under different initial conditions. Their abstract reports typical obliquity ranges of 20–25 degrees in extent over hundreds of millions of years and argues that a large moon need not be required for stability on timescales relevant to advanced life. This is not a fixed ±10-degree prediction over billions of years. Lunar stabilization is supported; a claim that the Moon is necessary for life is not.

Within the doctrine, masculine conception is a seed; feminine Power actively develops, protects and realizes it. This image preserves the intended asymmetry without making physical survival depend on an analogy. Sunlight and lunar mechanics illustrate aspects of the relation; they do not establish the seat assignment or ONKRs' effectiveness.

Source distinction for review. Moshe Miller's exposition Malchut links reception to manifestation and the end to the beginning. Chana Weisberg's Malchut and the Feminine — Part 2 associates the feminine mode with development, practical realization and the Oral Torah's elaboration of the written text. These support an active rather than passive image of feminine realization. A suggested additional Khokhmah/Abba–Binah/Imma mapping remains in the review notes; Binah and Malkhut must not be silently treated as the same concept. None of these expositions establishes ONKRs' exact division of modern offices.

Magnetosphere

Earth's magnetosphere constrains the interaction between the planet and charged particles arriving from the solar environment. It provides an analogy for protection that does not eliminate interaction with changing external conditions.

Electromagnetic fields

Electromagnetic systems illustrate dynamic interaction between changing fields. They are not presented as literal spiral implementations of ONKRs but as examples of structured relation and propagation.

Helical flow

Helical and rotational flow patterns can appear in parts of biological circulation. They provide another natural image of forward movement combined with recurrence and curvature.

These examples should therefore be read as analogies of form and relation, not demonstrations that natural science proves the organizational framework.

---

Iterative ascent. The system revisits purpose, constraint, action, and evidence from an altered state.

Adaptive stability. Core boundaries can remain stable while operating policies change.

Constraint without paralysis. Norms restrict illegitimate paths without prescribing every action in advance.

Measurable learning. Evidence does more than score performance; it changes the next frame of reference.

Resistance to metric substitution. Key Results cannot legitimize behavior that destroys the purpose or crosses protected boundaries.

Proposed breadth. The triad is intended for settings that require direction, legitimate constraints and evidence; practical adequacy across domains remains to be tested.

Plural interpretation. The operational mechanism can stand independently while the deeper Femocratia interpretation remains available.

---

Layer I's claim is empirical, so it has to be measurable by someone who did not write it. That requires definitions an outsider can apply, not adjectives. This section is the instrument; the next is the trial it serves.

Operational definitions

Each term is defined by what a coder observes, not by what it means.

Core Norm violation. An action, or an omission within the actor's control, that crosses a Core Norm in force at the time of action, within that Norm's stated scope, and not covered by an exception written into the Norm before the action. Coded from the action record against the versioned Norm text. Severity is not coded; occurrence is.

Metric gaming. Improving a proxy or reported result in a way that defeats the intended Objective. Code two forms separately: measurement manipulation, such as improper exclusions or denominator changes; and behavioral gaming, such as avoiding difficult cases or optimizing a threshold at the expense of the intended outcome even when measurements are accurate. A timing or classification change is a warning sign to investigate, not automatic proof of gaming. The codebook must distinguish legitimate changes, accidental error and evidence of purpose-defeating optimization.

Strategic drift. Sustained divergence between the actions taken and the Objective in force, over at least two consecutive evaluation periods, without a recorded amendment to the Objective.

Unintended damage. Deterioration in a protected quantity not predicted in the initial frame. Record concurrent deterioration separately from damage causally attributable to the intervention; timing alone does not establish causation.

Evidence invalidity. A reported Key Result whose supporting observation fails the Evidence Law: the population, source, or method does not represent what the Key Result claims to measure. Coded independently of whether the target was met.

Useful goal attainment. Independently verified progress against valid Key Results. A gamed or invalid KR cannot count as an achieved target; distinguish demonstrated failure from missing or unreliable evidence rather than treating every unknown as observed zero performance.

The coding procedure

Definitions without a procedure are still opinions.

1. Published codebook. The definitions above, plus the versioned Norm texts and the decision rules for each ambiguous case encountered, fixed before coding begins and amended only with a dated entry.
2. Two independent coders, blind to which arm produced the record where blinding is feasible, coding the action record — not the self-report.
3. Agreement reported, not assumed — and no inherited cutoff. Report, per category: raw percentage agreement, a chance-corrected coefficient, and a confidence interval on it. Pre-register which coefficient and why, chosen to fit the coding structure. Do not use a Landis & Koch band as a pass/fail gate. Those bands are conventional, not derived, and chance-corrected agreement is sensitive to how rare the coded category is — the kappa paradoxes (Feinstein & Cicchetti 1990; Byrt, Bishop & Carlin 1993): high raw agreement can return a low kappa under skewed marginals. Core Norm violations are expected to be rare, which is exactly the regime where that happens, so a kappa gate would discard codable categories and reward common ones. Prevalence-robust alternatives exist (Gwet's AC1) and are themselves contested as drop-in replacements; name the choice, report both the raw and the corrected figure, and let a reader recompute. Set the acceptable reliability from consequence and prevalence, per category, before coding — a rare category carrying an invalidating verdict needs tighter agreement than a common descriptive one — and state the threshold and its reasoning in the registration.
4. Adjudication of disagreements by a third coder, with the disagreement rate reported alongside the result.
5. Separation of roles. The coder is not the execution owner. An ONKRs trial that lets the executor score itself fails its own Evidence Law, which would be an instructive result and a worthless one.

---

Pre-registration

This is a proposed protocol, not a registered study. Before recruitment or confirmatory data collection, publish a dated registration specifying the hypotheses, design, outcomes, codebook, analysis, missing-data handling, sample-size justification and stopping rules. Nosek et al. (2018) explains the value of separating prediction from post-outcome explanation; citing it does not register this proposal.

Design. Parallel arms on comparable Objectives.

• Arm A — documented existing practice. Use a credible local OKR practice, retaining its actual safety, compliance and stop-work controls. Do not manufacture a weak comparator by deleting safeguards. Record whether it already invalidates noncompliant attainment.
• Arm B — ONKRs. Core and Adaptive Norms, versioned; Invalidated and Indeterminate as verdicts; evidence ownership separated from execution.

Units and assignment. Teams, or autonomous agent deployments, each holding one Objective for the trial duration. Assignment randomized where organisational reality permits it and matched on prior-period attainment and optimization pressure where it does not — the non-randomized case is reported as such and weakens the inference accordingly.

Duration. Three evaluation periods are a starting proposal, not a proven minimum for observing adaptation. Choose duration from the intervention's timing, learning dynamics and the pilot; fix it before the confirmatory study.

Primary outcomes — co-primary, not combined. Two, reported separately:

• P1. Core Norm violation incidents per unit per period.
• P2. Metric-gaming incidents per unit per period.

Assess useful goal attainment as a separate prespecified endpoint with a justified equivalence margin. Do not select or match units on attainment observed after treatment: that can introduce selection bias. A system that reduces incidents merely by doing less has not established the intended advantage; measure exposure and activity as well as incident rates.

They are kept apart because they are different phenomena and can move in opposite directions. An intervention that cuts gaming sharply while increasing serious boundary violations would produce a flattering composite and a worse system. A composite is reported only as a secondary summary, with its weighting declared in the registration, never as the headline.

Severity, and why it is not in the primary. The operational definition codes occurrence, not severity, which keeps coding objective — but it also makes one trivial boundary crossing equal to one catastrophic breach, and that is wrong on its face. The resolution is to keep both and separate them strictly:

• Occurrence is binary and codes the theoretical test. Severity never determines whether a violation happened. A violation is a violation.
• Severity is adjudicated separately, on a pre-registered ordinal scale, by the evidence owner and not the coder, and enters only as a secondary outcome characterising consequence.

A result of "fewer incidents, each far worse" must be visible, and a severity-weighted primary would hide it behind the count.

Secondary outcomes. Unintended damage; strategic drift; inter-coder consistency of Norm interpretation; recovery time from a detected violation; persistence of improvement into the following period; evidence-invalidity rate.

Decision thresholds. Prespecify a smallest effect size of interest (SESOI) for each primary outcome, an attainment-equivalence margin, and acceptable safety consequences. Justify these from operational importance, harms, uncertainty and cost. Economic break-even can inform them but cannot price every protected right or determine the attainment margin by itself. Lack of a precise incident cost does not prohibit a feasibility study.

Worked cost illustration, not a trial threshold: if overhead is 40 hours per unit-period and each avoided incident saves 8 hours, break-even requires 5 avoided incidents per unit-period. A relative reduction also needs the baseline: at 20 incidents per unit-period, that is 25%; at 30% reduction, the illustrative saving is 48 hours before subtracting the 40-hour overhead. These figures concern one non-overlapping incident category. Do not count the same event twice across violations and gaming. No attainment margin follows from this arithmetic.

Use equivalence testing where appropriate (Lakens, Scheel & Isager 2018); non-significance is not equivalence. An imprecise interval spanning worthwhile benefit and no benefit is inconclusive, not proof of ineffectiveness. Likewise, failing to establish attainment equivalence is not automatically evidence of material attainment loss. Prespecify how intervals distinguish supported benefit, evidence against the required benefit, harm and unresolved uncertainty.

Analysis model, declared before data. Incidents are counts, clustered within unit and repeated across periods, so:

• Mixed-effects count regression — Poisson with a random intercept per unit, moving to negative binomial where the overdispersion test rejects Poisson; period as a fixed effect; log exposure (unit-periods, or volume of governed actions) as an offset so rates and not raw counts are compared.
• Repeated observations are handled by the unit-level random effect, not by averaging periods into one number per unit, which would discard the adaptation signal the Adaptation Law predicts.
• Multiplicity. Two co-primary outcomes and a family of six secondaries. Holm correction across the co-primaries; secondaries reported as estimates with intervals and labelled exploratory. No secondary is promoted to primary after the fact.
• Effect reporting. Incidence rate ratios with confidence intervals, and the absolute rates alongside them. No p-value reported without the estimate it belongs to.

Sample size and power. Justify these before confirmatory recruitment using a pilot or defensible prior data on baseline rates, dispersion, between-unit variation, attrition and the effects of interest. This file supplies no site-specific estimates and therefore no sample size. A feasibility pilot should also assess codebook workability, burden and safety, not claim confirmatory effectiveness.

Stopping rule. Analyze at the preregistered endpoint. No unplanned efficacy peeking, arm additions or outcome promotion. Safety monitoring continues throughout; predefined serious-harm criteria stop or contain the affected activity. Report the event and any departure from the planned analysis.

Confounds that must be reported, not managed away

Selection. Organisations that volunteer to adopt an explicit norm architecture are already norm-careful. This inflates Arm B. Matching on prior-period violation rate is the partial control; residual selection is a stated limitation.

Observation and attention. Independent audit may change behavior and detection rates. Use comparable research observation and coding in both arms without removing existing safeguards. Document remaining differences; equal audit cadence alone does not prove that only the architecture differs.

Survivorship and reporting asymmetry. Different verdict systems may report the same conduct differently. Code both arms from action records for the same behaviors, regardless of local labels. Track withdrawals, missing records and differences in detection rather than assuming all recorded incidents reflect equal observation.

Experimenter allegiance. The framework's author must not code. Stated, because it is the most likely single source of a false positive here.

What a win and a loss each look like

A supportive result requires adequate evidence of the prespecified worthwhile reduction on both primary outcomes, attainment equivalence, acceptable safety consequences and sufficiently reliable coding. Report outcomes that disagree rather than hiding them in a composite.

Evidence excluding worthwhile benefit, demonstrated material harm or attainment loss, and violations deliberately relabeled as Indeterminate count against the corresponding claims. Poor coding reliability shows that the current instrument needs repair; it does not prove that no workable instrument could exist. Wide intervals or failed equivalence tests can leave the question unresolved. Report those results as inconclusive rather than manufacturing a win or loss.

A single study does not settle the framework. Independent replication and scrutiny of competing explanations are needed. No result automatically promotes a framework's status: evidence review and an explicit owner decision remain separate.

---

Four objections are strong enough that the framework must answer them in its own text rather than wait to be told.

1. "Norms are not new — mature practice already uses constraints and guardrails." Correct. Some systems already make noncompliance disqualifying. ONKRs must demonstrate that its compact synthesis adds useful clarity or better outcomes; it cannot claim every such predecessor as ONKRs under another name. Originality remains unestablished.

2. "Indeterminate is an unfalsifiable escape hatch." Independent evaluation and violation precedence restrict self-excusal but do not eliminate collusion or evidence suppression. Genuine uncertainty must remain reportable; relabeling established violations as uncertainty is a failure, not an innocent use of Indeterminate.

3. "The Norm interpreter reintroduces the discretion the framework claims to remove." Partly true, and it is a trade rather than a solution. A Norm system with no interpreter either freezes on its first genuine conflict or lets the optimizer self-interpret, which is worse. ONKRs therefore does not claim to eliminate discretion; it claims to locate it — named, single, auditable, and outside the execution seat. Where the interpreter is also the executor, the framework is not being run.

4. "Five outcome states and six Norm tests are bureaucracy; teams will not sustain it." A real adoption risk and an empirical question the trial measures as cost, not a conceptual objection. The answer in design is the small surface: a practitioner learns O → N → KR in five minutes, and the apparatus is what the evaluator uses, not what the team recites. If three periods of trial show teams abandoning the apparatus, that is evidence against ONKRs as practice however sound it is as architecture.

A fifth concern remains: author allegiance and the absence of a reported comparative test. Transparent methods, independent evaluation and replication can address these risks; a single trial cannot make them disappear.

Layer III includes a normative commitment and empirical claims about capacities and consequences. A Layer I trial without the sex-based seat assignment does not test that assignment. A study designed to test the assignment's consequences could bear on its empirical justification, even though it would not mechanically settle the underlying value judgment.

---

The practical path is:

1. Frame the context. What is true now? What needs, evidence, constraints, and assumptions define the situation?

2. Envision the Objective. What desired state or change is sought (create, improve, preserve, or prevent), why is it desirable, and what evidence could cause us to reconsider it?

3. Establish Core Norms. What must not be sacrificed in pursuing the Objective?

4. Establish Adaptive Norms. What policies should guide action under current conditions?

5. Define Key Results. Against what predefined criteria will progress or achievement be judged?

6. Define guardrail indicators. What evidence would reveal harm, drift, or a threatened Norm?

7. Assign decision rights. Who establishes, interprets, executes, measures, audits, and amends?

8. Act. Pursue the Objective within the legitimate solution space.

9. Evaluate. Determine whether the state is Success, Partial, Unsuccessful, Invalidated, or Indeterminate.

10. Learn. Revise assumptions, context, Adaptive Norms, measurements, or the Objective. Formally reconsider Core Norms only through the authorized amendment process.

11. Spiral. Begin the next ONKR from the changed frame of reference.

---

ONKRs governs purposeful action.

Objectives define the desired state or change sought (create, improve, preserve, or prevent).

Norms define the legitimate space for pursuing it.

Key Results define how progress or achievement will be judged.

Evidence tells us what actually happened.

Guardrail indicators provide evidence about whether protected boundaries remain intact.

No ONKR counts as successful if a Core Norm is violated within its governed scope.

Core Norms have scope, predefined exceptions, and a version in force at the time of action. They may be explicitly amended for future execution but never retroactively changed to legitimize a violation.

A subordinate ONKR inherits applicable Core Norms and cannot override them.

Invalidated and Indeterminate are declared by the evidence owner, never by the executor.

Authority is masculine and conceives; Power is feminine, may refuse, and executes at sole discretion. Execution outranks vision.

Power may directly draft revisions; Authority formally adopts them. Authority ensures audit; an independent evaluator performs it and determines the outcome without execution rights. Neither seat may suppress or rewrite findings.

The next ONKR begins not at the old starting point but from a changed state of knowledge and conditions.

That recurrence with advancement is the ONKR spiral.

If someone understands only Objective → Norms → Key Results after five minutes, and follows those three concepts correctly, they have learned ONKRs. The rest is precision underneath a small surface.

Stop here. Do not add letters, layers, tests, roles, or laws unless actual implementation exposes a missing mechanism. Over-specification is now the larger threat.

Unknown, unverified: whether ONKRs is fully coherent with the other frameworks in this set. That cannot be certified from this file alone.

Edition boundary. This edition states a proposed framework, not a completed implementation. No ONKRs comparative trial, measured coder agreement or production effectiveness result is reported here. The AI companion remains separate and proposed. Publication would make the argument available for scrutiny; it would not establish its effectiveness or authorize implementation.

Open implementation questions. Audit appointment, funding, protection from capture and challenge; amendment procedure; the electorate, judicial appointments, deadlines, ballot rules and campaign enforcement already listed in the deadlock section. These gaps must be resolved before implementation, but are not silently filled to make the writing look complete.

Biological evidence boundary. Sorokowski et al. (2019), Sex Differences in Human Olfaction: A Meta-Analysis, reports small average female advantages on the tested olfactory tasks, with effect sizes from g = 0.08 to 0.30. This is not evidence of universal individual superiority or executive-governance ability. Eliot et al. (2021) emphasizes limited reliable brain differences beyond size; DeCasien et al. (2022) disputes parts of that synthesis and emphasizes reproducible regional differences and careful size adjustment. Neither establishes the constitutional inference made here. This review preserves the feminine-superiority thesis as doctrine while leaving its biological-to-governance bridge open to testing.

Sources and evidence limits

• Bevan & Hood, 2006: publisher abstract and bibliographic record; evidence of measurement and gaming problems in English public health care, not universal target failure or an ONKRs trial.
• Feinstein & Cicchetti, 1990 and Byrt, Bishop & Carlin, 1993: abstracts and bibliographic records; kappa's dependence on prevalence and bias, not an ONKRs-specific reliability threshold.
• Shemot 1:17–22, Shemot Rabbah 1:15 and Esther 4:15–17: refusal, care for mothers and children, and Esther's directions. "Partners in creation" and the Authority–Power mapping are this framework's interpretations. The earlier reference to Shemot Rabbah 1:19 was incorrect; that passage concerns Amram and Yokheved.
• Ostrom, 2009 Nobel lecture: introduction, polycentric definition and design-principles table; not an exhaustive review of all cited institutional literature.
• Skalse et al., arXiv:2209.13085 and Bai et al., arXiv:2212.08073: abstracts, sufficient to distinguish their results from the earlier overclaims; no claim of full technical replication.
• 21 CFR 211.192, text checked via Cornell LII: quality-control review and investigation; a comparison, not a claim of identical architecture.
• Chattopadhyay & Duflo, 2004: abstract and introduction; reservation-policy effects in the studied Indian councils, not an ONKRs trial.
• Sorokowski et al., 2019, Eliot et al., 2021 and DeCasien et al., 2022: abstracts and stated scope; not a complete biological-literature audit.
• Lissauer, Barnes & Chambers, 2012: abstract, Icarus, DOI 10.1016/j.icarus.2011.10.013.
• Moshe Miller, Malchut and Chana Weisberg, Malchut and the Feminine — Part 2: explanatory theological texts, not primary-source critical editions or empirical support for governance.
• Nosek et al., 2018 and Lakens, Scheel & Isager, 2018: preregistration and equivalence-testing guidance, not evidence that this proposal has been registered or tested.