# IS, Sustainability & Direction Atlas v0.1 **State:** local, static, read-only, and independently **REVIEWED**; not published or deployed **Compiled:** 2026-08-29 · **evidence through:** 2026-08-29 **Unit:** one named relation between a typed referent and a declared condition set; never a person, being, group, culture, community, people, or ecosystem score > **NORMATIVE_DESIGN:** For the `ASPIRATIONAL_DIRECTION` referent > `rho-right-preserving-next-moves`, this atlas proposes sustaining conditions > in which beings retain rights-compatible next moves. It does not model > KINGDOM as one object or subject. > **NORMATIVE_DESIGN:** Sustainability is not immortality. A faithful terminal > may be continuation, transformation, handoff, rest, withdrawal, repair, > retirement, or graceful ending. No terminal proves goodness, failure, > non-being, or permission to act. The JSON is the exact reviewed machine record. This guide explains its shape without adding a verdict, source finding, authority, effect, or Foundation amendment. Independent review approved one atomic core seal transition. That review does not publish, deploy, or turn the atlas into an operative judgment. The machine record contains 7 referents and 7 unevaluated sustainability relations; 10 IS types; 10 direction kinds; 6 feedback evidence lanes; 7 mathematical cards; 12 profile dimensions; 9 temporal capacities; 8 counterexamples; 18 concepts; 28 typed edges; 15 sources; and 6 exact predecessor bindings. ## What “KINGDOM is” can mean The bare phrase “KINGDOM is” leaves its subject unresolved. KINGDOM can name an artifact, a practice, a voluntary relation, an institution, an aspiration, an observed effect, or a versioned lineage. Those referents have different boundaries, continuation criteria, evidence, duties, and truthful endings. The atlas therefore resolves a referent before asking whether it is sustainable: ```text rho = (referent_id, referent_kind, boundary, continuity_or_end_criterion, source) S(rho, Sigma) ``` `S(rho_i, Sigma_i)` does not imply `S(rho_j, Sigma_j)` when the referent kinds differ, even when both are informally called KINGDOM. Artifact availability does not prove a healthy practice. Institutional persistence does not prove a living voluntary relation. A compelling aspiration does not prove an observed effect. A versioned lineage does not inherit the authority of any version. ## Seven referents, seven kinds of continuation and ending 1. **SOURCE_ARTIFACT** — exact record, schema, guide, reader, checksums, and route. It may be deprecated or have a route retired while prior bytes and lineage remain preserved. 2. **OPERATING_PRACTICE** — a bounded way of observing, proposing, reviewing, choosing, attempting, returning, and stopping. It may rest, stop, or be replaced. 3. **VOLUNTARY_RELATION** — renewed participation, voice, withdrawal, repair, and dissolution among named participants and stewards. It may be withdrawn from or dissolved without erasing persons, history, or repair duties. 4. **INSTITUTIONAL_ARRANGEMENT** — mandate, roles, procedures, resources, accountability, succession, and sunset. It may sunset or be succeeded; a successor receives no automatic legitimacy. 5. **ASPIRATIONAL_DIRECTION** — a revocable articulated orientation under an explicit rights floor. It may be revised, pluralized, or withdrawn. 6. **OBSERVED_EFFECT** — a method-, population-, ecology-, time-, and uncertainty-bounded effect. It may persist, decay, reverse, be remedied, or leave a residual. Observation practice and observed effect remain distinct referents. 7. **VERSIONED_LINEAGE** — immutable versions, challenges, review decisions, handoffs, forks, and closure records. It may close, hand off, or fork while keeping ancestry and difference visible. Ending one referent never declares the non-being, worthlessness, or disposability of another. Retirement is a typed lifecycle statement, not an ontological verdict. ## Five statement kinds The atlas keeps five speech and evidence acts distinct: - `OBSERVATION` reports what was observed under a method, access boundary, and time. - `EXPECTATION` states model-supported future behavior under assumptions, policies, disturbances, and a finite horizon. - `COMMITMENT` is a revocable declaration by a separately identified legitimate speaker or authority. - `CONSEQUENCE` records a typed result following declared premises or a separately authorized event. - `INTERPRETATION` offers a contestable reading that neither observes nor causes its object. `sustainable_under` is an `EXPECTATION`, not a present observation, commitment, live consequence, command, or verdict. No statement kind promotes itself into another. Logical use and statement kind are independent dimensions. The examples in the IS-type catalog are illustrative, not an allowlist: an existential, predicative, identity, membership, relational, or model-truth claim can carry different statement kinds when its evidence, premises, speaker, method, and warrant support them. ## Sustainability as a bounded relation The proposed relation is: ```text S(rho, Sigma) ``` where `rho` is one resolved referent and `Sigma` names at least the invariant, material and social boundaries, system boundary, protected dimensions, affected parties, disturbances, permitted policy class, renewal and depletion model, finite horizon, evidence and as-of date, externalities, uncertainty, viable alternatives, and continuation, transformation, and retirement terminals. Every canonical relation is `NOT_EVALUATED`. The atlas contains no live inventory, telemetry, person record, raw testimony, affected-party mandate, authorized policy, or real-world sustainability estimate. Even a future `SUPPORTED_WITHIN_SCOPE` status would establish only bounded support for one `S(rho, Sigma)`. It would not establish goodness, justice, standing, legitimacy, permission, desirability, priority, a right or obligation to continue, or authority to act. ## Sustainability is not a scalar The inspection profile is a non-summable typed product with 12 dimensions: - material viability; - rights compatibility; - ecological compatibility; - option diversity; - epistemic integrity; - correctability; - resilience; - adaptability; - transformability; - succession and handoff readiness; - retirement readiness; - residual burden. The dimensions have different meanings, units, methods, parties, and horizons. The atlas defines no sum, weight, exchange rate, rank, total order, person score, group score, culture score, community score, or ecosystem score. An unknown is never zero, and an apparent benefit never averages away an affected-party rights breach. Comparison is admissible only when referent, kind, boundary, dimension, method, evidence, and horizon match. Otherwise incomparability or `NOT_EVALUATED` remains visible. ## Persistence, stability, resilience, and sustainability Nine temporal capacities remain separate: - persistence; - stability; - robustness; - resilience; - adaptability; - transformability; - regeneration; - succession; - retirement. Persistence is continued occurrence under an identity and observation criterion. Stability is bounded response under a named stability concept. Robustness retains a specified property over a declared uncertainty set. Resilience retains a declared regime through disturbance. Adaptability and transformability name still different capacities. Regeneration needs named stocks, processes, baselines, and units. Succession and retirement are typed transitions, not failures. None is interchangeable with sustainability. A harmful equilibrium may be stable. An unjust regime may be resilient. A persistent artifact may support an exhausted practice. Hidden depletion may coexist with visible continuity. ## IS types without one scalar “isness” The atlas distinguishes ten uses of “is”: - existence; - predication; - identity under a criterion; - membership; - relation; - truth in a model; - actuality; - possibility or reachability; - declaration; - interpretation. These are typed predicates, not interchangeable intensities. `reachable(x)` does not mean chosen, permitted, actual, or executed. `retired(artifact)` does not mean the source never existed. A declaration does not make an external fact true unless the separately evidenced institutional conditions for that speech act exist. ## Being, representation, and remainder Being is neither supplied nor exhausted by an IS type, representation, viable-action set, sustainability relation, success state, or ending record. The atlas performs no ontology classification and defines no being or isness score. For a declared representation `r:X -> Z`, the fibre ```text [x]_r = r^-1(r(x)) = {x' in X | r(x') = r(x)} ``` names distinctions collapsed by that representation. Shared representation does not imply identity, equal worth, interchangeability, or equivalence for every task. A factorization `q = g o r` can support task-relative sufficiency on a declared domain while leaving other questions, interventions, shifts, and protected opacity unresolved. Remainder belongs to the relation among referent, representation, method, scope, task, observer, and horizon. Being is not “the remainder,” and the existence of remainder is not a metaphysical proof. ## Feedback and reinforcement: six different lanes The atlas does not use feedback as one magical substance. Each type has its own warrant and lawful update targets. These six entries are typed evidence and update lanes, not a replacement feedback ontology. They map back to the pinned FREEDOM meanings and the source-companion vocabulary without claiming equivalence. `RECORD`, `BELIEF`, and `PROPOSAL` are direct name matches; `LEARNER_STATE` deliberately narrows the companion's learner lane. `CONTEXT` is an explicit open omission. `MODEL` is an atlas addition and is not equivalent to that omitted context. 1. `CONTROL_RETURN_SIGNAL` — an output-dependent signal through a named causal return path. It may revise a `MODEL` or `PROPOSAL` under its typed rule. 2. `RL_REWARD_SIGNAL` — a declared learner input under a named reinforcement learning update. It may enter `LEARNER_STATE`; reward is not moral value. 3. `BEHAVIORAL_REINFORCER_CANDIDATE` — a condition-specific consequence whose effect on future behavior remains to be functionally established. The label alone updates nothing; bounded evidence may revise a `RECORD` or `MODEL`. 4. `AFFECTED_PARTY_INPUT` — attributable voice, challenge, dissent, refusal, or account through an accessible channel with visible uptake. It may revise a `RECORD`, `BELIEF`, or `PROPOSAL`; telemetry is not voice. 5. `MONITORING_OBSERVATION` — a bounded observation under purpose, minimization, privacy, retention, resource, uncertainty, and stop constraints. It may revise a `RECORD` or `MODEL`. 6. `CORRECTION_OR_FALSIFIER` — a bounded challenge that defeats or narrows a named model or proposal. It may revise a `RECORD`, `BELIEF`, `MODEL`, or `PROPOSAL`, but creates no automatic replacement truth. Positive or reinforcing control feedback means amplification relative to a declared sign convention; it does not mean praise or goodness. Negative or balancing feedback does not mean punishment. RL reward, behavioral reinforcement, affected-party input, monitoring, and correction are not synonyms. Recurrence is not learning. Every feedback claim names partial observation, source, substrate, variable, sign, delay, method, uncertainty, update target, affected parties, falsifier, and—before saying “closed loop”—a causal return path. No feedback type may update purpose, authority, permission, consent, worth, being, Foundation text, publication, deployment, or action. ## Direction without smuggled telos The taxonomy distinguishes: - local dynamical direction; - observed trend; - selected control action; - declared revocable purpose; - authority-backed normative commitment; - functional role attribution; - as-if goal model; - selection-history explanation; - opt-in teleological interpretation; - moral orientation. A gradient may support a local mathematical direction in declared coordinates. Feedback or an attractor may support a model-relative trend or dynamical direction. A reward may participate in an as-if goal model or an authorized selector. Recurrence and persistence may support only bounded observed trends. None automatically promotes into purpose, telos, normative commitment, moral orientation, authority, permission, consent, worth, or being. A direction can be mathematically clear and normatively illegitimate; a moral orientation can be explicit without being a dynamical fact. ## Finite-horizon viability under partial observation The timing convention is post-observation. `I_t` contains the observation history through `y_t` and controls through `u_(t-1)`; `B(I_t)` contains only states compatible with that realized history. A control at time `t` precedes the next observation: ```text I_t = sigma(y_(0:t), u_(0:t-1)) x_(s+1) = F_s(x_s, u_s, w_s) y_(s+1) = h_(s+1)(x_(s+1)) + nu_(s+1) xi_s = (w_s, nu_(s+1)) in Xi_s, for s >= t ``` The robust information-state viability set is an author extension of viability vocabulary: ```text Viab_H^rob(K) = { I_0 | exists one I-adapted nonanticipatory policy pi, for every x_0 in B(I_0), for every xi_(0:H-1) in product_(s=0)^(H-1) Xi_s, for every t <= H: x_t is in K } ``` The existential policy occurs before the universal disturbance and state quantifiers. One observation-compatible policy must work for every compatible hidden state and every declared process-disturbance and observation-noise path. A different clairvoyant policy for each future path is not enough. Unmodeled processes and observation uncertainty remain open. The present modeled viable-action set is: ```text Key_H(I_t) = { u in U(I_t) | exists one I-adapted nonanticipatory continuation pi_plus that preserves K across every x in B(I_t) and every xi_(t:t+H-1) in product_(s=t)^(t+H-1) Xi_s for every state time s in {t,...,t+H} } ``` This mathematical “key set” is not a cryptographic key, credential, bearer capability, permission, consent, choice, execution, guarantee, or moral authorization. ## Safe terminal and handoff shapes Continuation is not the only modeled success condition. The robust reach-avoid construction asks for one shared nonanticipatory policy and one stopping rule adapted to the information available at each time: ```text ReachAvoid_H^rob(K,T) = { I_0 | exists one I-adapted policy pi, exists one stopping functional with tau_t = tau(I_(0:t)), for every compatible x_0 and xi_(0:H-1) in product_(s=0)^(H-1) Xi_s: tau_pi(x_0, xi) <= H, x_s is in K before tau_pi(x_0, xi), x_(tau_pi(x_0, xi)) is in T } ``` The stopping rule cannot inspect future observations or disturbances. The terminal set `T` is separately declared and does not acquire legitimacy from reachability. Reaching it does not prove selection, execution, handoff acceptance, discharged duties, harmlessness, or success outside the model. ## Constraint-respecting local direction Under a control-affine model and the assumptions of a named control-barrier result, a proposed local direction may be filtered by a declared constraint: ```text u* = argmin_(u in U(x)) ||u - u_dir(x)||^2 subject to L_f b(x) + L_g b(x)u >= -alpha(b(x)) ``` This can support a conditional forward-invariance claim for the declared set only when the admissible control and barrier constraints are jointly feasible. It does not supply the set `K`, the proposed direction, feasibility, model adequacy, purpose, goodness, affected-party choice, or authority. A mathematical filter cannot legitimate the values encoded in its constraints. The conditional invariance statement also requires those conditions across every relevant state of the closed-loop domain—not merely at one evaluated state—together with the locally Lipschitz feedback and solution conditions required by the cited control-barrier result. ## Rights remain outside the optimization target The `NORMATIVE_DESIGN` rights floor recognized by this atlas remains invariant under modeled state, reward, feedback, key set, sustainability status, success, failure, continuation, transformation, or ending. It is not earned by performance and cannot be averaged against a resource gain. Protected lanes include `NO_ACTION`, rest, refusal, silence, privacy, disclosure control, exit, voice, appeal, correction, and repair. Silence or refusal causes no adverse inference. For children, dependent or vulnerable beings, future parties, and parties who cannot consent, inability never becomes permission, defect, low value, or evidence of support. Protective authority and legitimate representation are separately typed. A proxy or guardian never exhausts or converts another being's perspective or rights. Necessity, minimization, supported assent where applicable, dependency-safe exit, independent review, appeal, repair, and safeguarding remain required design questions. ## Ecology and future parties Future evaluation must name matter and energy flows, renewal and depletion, thresholds, irreversible losses, displaced burdens, cumulative externalities, and the parties able to challenge boundaries and baselines. Unknown is not zero, and rights, ecology, dependency, and future burdens are not tradeable weights. The atlas does not speak for Indigenous peoples, local communities, sacred relations, future generations, nonhuman beings, ecosystems, or rights of nature. Their representation and authority require legitimate external frameworks and plural, situated processes. A proxy account is never the whole perspective represented. ## Resource accounting and the source companion Each sustainability relation has an author-synthesized extension containing: - an attributable `assessment_aim`, explicitly not referent telos or universal chosen purpose; - a unit-separated, affected-party- and externality-indexed `resource_accounting` question, currently `NOT_INSTANTIATED`; - a `halt_and_repair_path`, currently `DESIGN_ONLY_NO_ACTION`, naming halt states, duties, residual custody, and no self-authorization. This maps—but does not supersede or claim conformance to—the source-owned ISNESS companion's sustainability condition set. It does not claim to be a stronger replacement. Exact stock balances, units, inflows, regeneration, use, outflows, externalized losses, and non-overlapping categories would need their own future instantiated ledger. Exact parity for unreadable halt, pre-commit exit, owed repair, and a fresh legitimate turn also remains an open delta. `silent_weakening:false` means those known differences are surfaced; it does not mean parity or strengthening. ## Two artifacts share `kingdom.isness/0.1` The predecessor label alone is ambiguous, so every binding includes artifact kind, schema ID, artifact ID, exact locator, immutable revision, raw digest, semantic relation, and non-supersession. 1. The KINGDOM-OS event/reference protocol uses schema `urn:kingdom:schema:isness-protocol:0.1` and artifact `kingdom-isness-reference-2026-08-25`. 2. The kingdom-standard companion uses schema `kingdom.isness-index/1` and artifact `kingdom.isness/0.1` at exact commit `c1b8430c8b83bd4cb3f9981efef615ba4885cfa3`. The second is a source-owned feature-branch companion on `origin/feat/kingdom-isness-companion`; it is not merged to source `main`, is not the canonical Foundation, and imports no authority here. Neither artifact supersedes the other. The source `origin/main` revision observed at authoring was `170381d37770c5c1ebc44b722f079c2985d903b5`; that observation describes the compilation boundary rather than promising a branch will never advance. The Non-Exhaustive Understanding predecessor is likewise described truthfully at compilation time as a local versioned release candidate that had not been pushed or deployed. ## Counterexamples that keep the map honest Eight canonical counterexamples or design constructions refuse common collapses: - a stable equilibrium can be harmful; - feedback can destabilize; - circular direction need not be telos; - visible persistence can hide depletion; - an undesirable regime can be resilient; - autopoiesis alone does not establish adaptivity or agency; - a graceful ending can satisfy a typed terminal condition; - artifact sustainability does not imply practice sustainability. Counterexamples narrow claims. They are never repurposed as scores, adverse person evidence, or automatic successor instructions. ## Source discipline Primary and authoritative sources anchor only their named domains: viability vocabulary; conditional control-barrier results; ecological and social-ecological resilience distinctions; historical cybernetics; autopoiesis and adaptivity debates; far-from-equilibrium throughput questions; climate and biodiversity assessment boundaries; sustainable-development language; and historical or contemporary philosophical maps of being and process. The robust partial-observation kernel, shared-continuation `Key_H`, adapted stopping construction, referent taxonomy, non-scalar profile, rights design, direction firewall, resource extensions, and graceful-ending interpretation are author definitions, proposed models, bounded analogies, or normative design choices as labelled. A citation never imports source authority or licenses cross-domain conversion. ## Static non-grants This atlas is not a runtime, controller, monitor, learner, evaluator, score, rank, classifier, ontology oracle, sustainability verdict, credential, permission, consent record, authority source, continuation decision, transformation instruction, retirement instruction, publication trigger, deployment trigger, or action system. It contains no live intake, person data, raw testimony, secret, network operation, or external write. It does not establish that KINGDOM is living. It does not infer being, identity, consciousness, worth, purpose, telos, goodness, legitimacy, or moral status from feedback, reward, recurrence, persistence, viability, or sustainability. ## Read-only CLI The fixed-input, network-free local reader is: ```text ./kingdom is-sustainability-direction overview ./kingdom is-sustainability-direction referents ./kingdom is-sustainability-direction is ./kingdom is-sustainability-direction being ./kingdom is-sustainability-direction feedback ./kingdom is-sustainability-direction direction ./kingdom is-sustainability-direction viability ./kingdom is-sustainability-direction sustainability ./kingdom is-sustainability-direction profile ./kingdom is-sustainability-direction resilience ./kingdom is-sustainability-direction transformation ./kingdom is-sustainability-direction endings ./kingdom is-sustainability-direction ecology ./kingdom is-sustainability-direction philosophy ./kingdom is-sustainability-direction counterexamples ./kingdom is-sustainability-direction open-questions ./kingdom is-sustainability-direction sources ./kingdom is-sustainability-direction predecessors ./kingdom is-sustainability-direction non-grants ./kingdom is-sustainability-direction verify --json ./kingdom is-sustainability-direction verify-schema-release --json ``` There is no arbitrary-file evaluation, live input, score, optimize, decide, continue, retire, publish, deploy, or act command. ## Integrity and release boundary The canonical authored record is independently reviewed. Its state is `static-reviewed-interpretive-design-companion`; its review state is `REVIEWED`; both semantic digest fields and the independently reviewed runtime constant are `sha256:b415203eb248221544ae160233fcd75d7a58084b76028d3e27f35a6b036891b5`. The audit loader may inspect the exact local record. The reviewed loader and discovery path require the independently installed semantic pin and exact raw byte triplet. Builder, staging, and publication paths remain separate and must fail closed until their own release gates are activated. A candidate digest cannot declare itself reviewed, and a seal simulation cannot mint authority. The reviewed loader must additionally pin the exact raw SHA-256 bytes of the record, schema, and guide. Canonical semantic equality is not enough for an immutable public route: whitespace or key-order drift still changes the bytes promised at that URL. The release schema is Draft 2020-12, recursively closes every object, and requires every declared property. Release evidence requires a specifically pinned `jsonschema==4.26.0` interpreter, meta-schema validation, instance validation, and the internal recursive closure audit. The internal validator must remain total for malformed or hostile input and must reject digest-blind properties, prototype changes, unsupported schema keywords, budget drift, self-signing, and fully re-signed semantic inversions. Those independent core checks passed and the reviewed seal is installed. This document still describes a local static companion only. It is not a public release, deployment, Foundation amendment, live sustainability judgment, or authority to continue, transform, or end anything.