# GenOS Agent Runtime

> Open-source AI agent runtime with versioned state, supervised execution and path-specific evidence gates. Capability maturity varies by mode and adapter.

- Website: https://genos.work/en
- French overview: https://genos.work/fr
- Source repository: https://github.com/PISSARAW/GenOS
- Reviewed product contract: https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/docs/03-reference/contrat-produit-et-completude.md
- Source revision: 5091f70608a9305cb7bbbf5bb292d60c4c9b361e (reviewed 2026-10-18)
- Sitemap: https://genos.work/sitemap.xml

## Start here
- English overview: https://genos.work/en
- French overview: https://genos.work/fr
- Runtime: https://genos.work/en/runtime
- Recorded runs: https://genos.work/en/runs
- Evidence ledger: https://genos.work/en/evidence
- Developer quickstart: https://genos.work/en/developers
- REST and MCP API reference: https://genos.work/en/api-mcp

## Site sections
- Overview: What does GenOS keep with the work? — https://genos.work/en/systems
  - System map: https://genos.work/en/systems
  - Latest evolutions: https://genos.work/en/updates
  - Runtime overview: https://genos.work/en/runtime
  - Recorded runs: https://genos.work/en/runs
  - Concept Atlas: https://genos.work/en/concepts
- System: What is GenOS made of? — https://genos.work/en/runtime
  - Runtime: https://genos.work/en/runtime
  - Supervision: https://genos.work/en/runtime/supervision
  - Workers: https://genos.work/en/workers
  - Daemons: https://genos.work/en/daemons
  - Topologies: https://genos.work/en/topologies
  - Morphogenesis: https://genos.work/en/morphogenesis
  - Teaching lab: https://genos.work/en/lab
  - Organizations: https://genos.work/en/organizations
- Evidence: What supports a claim, and what remains open? — https://genos.work/en/evidence
  - Evidence ledger: https://genos.work/en/evidence
  - Five capabilities: https://genos.work/en/capabilities
  - Benchmarks: https://genos.work/en/benchmarks
  - Recorded runs: https://genos.work/en/runs
  - Research program: https://genos.work/en/research
  - External repositories: https://genos.work/en/ecosystem
- Developers: How do I install, program, or integrate GenOS? — https://genos.work/en/developers
  - Quickstart: https://genos.work/en/developers
  - API / MCP reference: https://genos.work/en/api-mcp
  - Live sandbox: https://genos.work/en/sandbox

## French pages
- Aperçu: https://genos.work/fr/systems
  - Carte système: https://genos.work/fr/systems
  - Dernières évolutions: https://genos.work/fr/updates
  - Aperçu du runtime: https://genos.work/fr/runtime
  - Exécutions enregistrées: https://genos.work/fr/runs
  - Atlas des concepts: https://genos.work/fr/concepts
- Système: https://genos.work/fr/runtime
  - Runtime: https://genos.work/fr/runtime
  - Supervision: https://genos.work/fr/runtime/supervision
  - Workers: https://genos.work/fr/workers
  - Daemons: https://genos.work/fr/daemons
  - Topologies: https://genos.work/fr/topologies
  - Morphogenèse: https://genos.work/fr/morphogenesis
  - Laboratoire pédagogique: https://genos.work/fr/lab
  - Organisations: https://genos.work/fr/organizations
- Preuves: https://genos.work/fr/evidence
  - Preuves: https://genos.work/fr/evidence
  - Cinq capacités: https://genos.work/fr/capabilities
  - Benchmarks: https://genos.work/fr/benchmarks
  - Exécutions archivées: https://genos.work/fr/runs
  - Programme de recherche: https://genos.work/fr/research
  - Dépôts externes: https://genos.work/fr/ecosystem
- Développeurs: https://genos.work/fr/developers
  - Développeurs: https://genos.work/fr/developers
  - Référence API / MCP: https://genos.work/fr/api-mcp
  - Sandbox: https://genos.work/fr/sandbox

## Truth modes
- SIMULATION: A simplified model running in the browser. It explains an idea and proves nothing about the runtime.
- RECORDED: An artifact from a real historical GenOS execution. Inspectable, with failures included; reproduction depends on the available inputs and harness.
- LIVE: A connected GenOS endpoint accepts a mission request. Check worker, decision, verification and promotion states in its response.

## Knowledge surfaces (machine-readable)
- https://genos.work/openapi.json
- https://genos.work/api/v1/concepts
- https://genos.work/api/v1/topologies
- https://genos.work/api/v1/benchmarks
- https://genos.work/api/v1/experiments
- https://genos.work/api/v1/research
- https://genos.work/api/v1/evidence
- https://genos.work/api/v1/capabilities
- https://genos.work/api/v1/releases

## Concepts (156)
- Failure modes: https://genos.work/en/concepts/maladies — GenOS models potential failures in an agent system: excessive alerts, cross-workspace contamination, unintended intervention effects, or degraded performance.
- Memory: https://genos.work/en/concepts/memoire — GenOS memory helps retrieve useful experiences, connect decisions, and avoid known dead ends. It combines detailed episodes with indexed facts and provenance signals.
- Cortex: https://genos.work/en/concepts/cortex — In GenOS, “cortex” describes some retrieval and association functions. It does not mean a unified artificial brain: these mechanisms are distributed across indexes, memories, agents, and controllers.
- Ontogenesis: https://genos.work/en/concepts/ontogenese — Ontogenesis is the resident controller that follows a project between missions: it selects a task, starts a bounded run, examines evidence, and coordinates integration or resumption.
- Ontology: https://genos.work/en/concepts/ontologie — The operational ontology provides structures for describing entities, identity criteria, relationships, and hypothetical scenarios. It supports analysis without deciding on its own what should be executed.
- Genome: https://genos.work/en/concepts/genome — A versioned software model of inherited traits and constraints; it is not biological DNA and does not by itself instantiate a running agent.
- Epigenetics: https://genos.work/en/concepts/epigenetics — A software policy for expressing modeled traits in context; it is an analogy, not biochemical marking or biological inheritance.
- AgentDNA: https://genos.work/en/concepts/agent-dna — A versioned binary data format for agent identity and runtime traits; editing it does not by itself start or clone a process.
- Phenotype: https://genos.work/en/concepts/phenotype — The observable runtime profile produced by identity, constraints, and environment.
- Instinct: https://genos.work/en/concepts/instinct — Bounded innate response patterns that can be selected or modulated by runtime context.
- Reproduction: https://genos.work/en/concepts/reproduction — Rust reproduction operations transform genome and cell models under inheritance rules; worker startup is separate, while MCP biomimicry handlers record simulations.
- Speciation & grafting: https://genos.work/en/concepts/speciation-graft — A documented path for proposing a new capability or agent lineage and evaluating it before promotion.
- AGOW: https://genos.work/en/concepts/agow — An active workspace where candidate frames compete and selected content can be broadcast under policy.
- GVX: https://genos.work/en/concepts/gvx — A verified development loop that isolates changes, observes outcomes, and uses evidence to govern maturity.
- Self-model: https://genos.work/en/concepts/self-model — A bounded model of agent identity and continuity used for analysis, not a claim of subjective experience.
- Theory of self: https://genos.work/en/concepts/theory-of-self — Runtime models that describe agent identity, continuity, and change without asserting subjective experience.
- Predictive system: https://genos.work/en/concepts/predictive-system — A multi-scale prediction model that relates observations, hypotheses, and expected outcomes.
- Natural Search Control Plane: https://genos.work/en/concepts/natural-search-control-plane — A control model for bounded search using pressure, causal progress, hypotheses, and explicit control.
- Endogenous Thought Test: https://genos.work/en/concepts/endogenous-thought-test — A bounded experimental protocol that detects an internal signal, formulates a falsifiable question, arbitrates whether to investigate it, produces a read-only proposal, and submits it to independent verification.
- Attention & salience: https://genos.work/en/concepts/attention — Selection and prioritization signals used to direct limited agent work toward relevant inputs.
- Mind and consciousness taxonomy: https://genos.work/en/concepts/consciousness-taxonomy — A vocabulary for discussing computational indicators while separating them from claims of subjective experience.
- Functional indicators: https://genos.work/en/concepts/functional-indicators — Observable indicators used for bounded analysis; they do not establish subjective consciousness.
- Epistemics: https://genos.work/en/concepts/epistemics — The runtime vocabulary and procedures for representing what is known, uncertain, or contested.
- Beliefs: https://genos.work/en/concepts/beliefs — Versioned propositions with confidence, provenance, and revision history.
- Claims: https://genos.work/en/concepts/claims — Explicit statements that can be supported, challenged, scoped, and linked to evidence.
- Evidence: https://genos.work/en/concepts/evidence — Artifacts and observations used to support or reject a claim under a stated decision rule.
- Contradictions: https://genos.work/en/concepts/contradictions — Conflicting claims or observations preserved for review instead of silently collapsed.
- Provenance: https://genos.work/en/concepts/provenance — Source and lineage information attached to evidence, findings, and decisions.
- Brier calibration: https://genos.work/en/concepts/brier-calibration — A scoring method for comparing probabilistic forecasts with resolved outcomes.
- Gettier problems: https://genos.work/en/concepts/gettier — A model for examining why a justified belief can still fail to count as knowledge.
- Knowledge: https://genos.work/en/concepts/knowledge — A structured representation of sources, claims, inference, and uncertainty.
- GenOS Philosophy: https://genos.work/en/concepts/genos-philosophy — A canonical registry with 420 bounded executable audits over eleven shared primitives, linked to Ontogenesis verification without granting promotion authority.
- Episodic memory: https://genos.work/en/concepts/episodic-memory — Records of contextualized events, actions, observations, and outcomes.
- Semantic memory: https://genos.work/en/concepts/semantic-memory — Indexed facts and relationships retrieved to supply context to a mission.
- Vector memory: https://genos.work/en/concepts/vector-memory — A similarity-search representation used to retrieve related items from an embedding index.
- Procedural memory: https://genos.work/en/concepts/procedural-memory — Reusable instructions or action sequences represented as runtime knowledge.
- Synaptic plasticity: https://genos.work/en/concepts/synaptic-plasticity — A bounded update of selected associations or weights based on observed signals.
- STDP: https://genos.work/en/concepts/stdp — Spike-timing-dependent plasticity as a computationally bounded timing-based update rule.
- Fossilization: https://genos.work/en/concepts/memory-fossilization — Terminal archival of a lineage or experience while preserving its historical record.
- Memory retrieval: https://genos.work/en/concepts/memory-retrieval — A ranked selection of stored items that can inform, but cannot prove, a current decision.
- Morphogenesis: https://genos.work/en/concepts/morphogenesis — A runtime for composing bounded executable graphs from topology and control operators.
- Epistemic meristem: https://genos.work/en/concepts/epistemic-meristem — Tracks verified coverage gaps and revalidates them before proposing growth.
- Unblocking spiral: https://genos.work/en/concepts/unblocking-spiral — Changes method and scope through a bounded, evidence-gated response to a repeated blockage.
- Counterexample cambium: https://genos.work/en/concepts/counterexample-cambium — Stores counterexamples and gates their reuse in later work.
- Aperiodic chronotaxis: https://genos.work/en/concepts/chronotaxis — Persists observation schedules and receipts to expose missed temporal windows.
- Contracted statistical risk: https://genos.work/en/concepts/risk-ledger — Accounts for a bounded promotion-risk budget before opt-in statistical promotion.
- Dynamic organizations: https://genos.work/en/concepts/dynamic-organizations — Nineteen declared organization patterns with deterministic step algorithms and varying maturity.
- Worker kinds: https://genos.work/en/concepts/worker-kinds — A catalog of runtime worker types selected against role and method capability requirements.
- Resident daemons: https://genos.work/en/concepts/resident-daemons — Long-lived project services that observe state and may propose or carry out bounded work under policy.
- Agent relationships: https://genos.work/en/concepts/agent-relationships — Declared and measured links that shape coordination, routing, or communication between agents.
- Relational physiology: https://genos.work/en/concepts/relational-physiology — A bounded kernel for relation constraints, provenance independence, authority and admitted routing across 29 relation types.
- Communication ecology: https://genos.work/en/concepts/communication-ecology — Signals and channels that let agents coordinate through bounded, typed communication patterns.
- Zero-text signal plane: https://genos.work/en/concepts/signal-plane — Persisted signals and deliveries, inbox/ACK, leased handler retries and bounded cognitive escalation.
- G-CIR: https://genos.work/en/concepts/g-cir — A residual cognitive contract that keeps versioned obligations, refusals and per-invocation visibility explicit.
- Mission continuity: https://genos.work/en/concepts/mission-continuity — Durable mission ownership, completion checks, dormancy, wake-up and guarded regeneration across process restarts.
- AgentGit: https://genos.work/en/concepts/agent-git — Versioned agent state and lineage operations such as commit, branch, diff, replay, and merge.
- Counterfactual workspaces: https://genos.work/en/concepts/counterfactual-workspaces — Isolated workspaces used to compare alternative changes before any result is promoted.
- Trinity: https://genos.work/en/concepts/trinity — Three candidate worlds are compared through a declared evidence barrier before a candidate is retained or promoted.
- A-Team: https://genos.work/en/concepts/a-team — Specialists coordinate work through dependencies, explicit handoffs, and integration of their deliverables.
- Biocenosis: https://genos.work/en/concepts/biocenose — A community evaluates proposals through evidence, explicit ballots, and policy-bound aggregation.
- Holobiont: https://genos.work/en/concepts/holobionte — A host composes specialist symbionts under capability, resource, lifecycle, and execution contracts.
- Syncytium: https://genos.work/en/concepts/syncytium — Members coordinate around shared, versioned state with merge operations and invariant checks.
- Rhizome: https://genos.work/en/concepts/rhizome — A persisted capability graph with callable bounded multi-hop routing, local traces, and evidence-gated growth services.
- Metapopulation: https://genos.work/en/concepts/metapopulation — Semi-independent demes exchange typed propagules over directed corridors, with receiver validation, verified migration cycles, and gated recolonization.
- Biome: https://genos.work/en/concepts/biome — Eleven variants share a bounded transactional ecology loop with persistent niches, measured resource use, authorized adapters and independent verification gates.
- Mutation: https://genos.work/en/concepts/mutation — Rust operations change genome data models; MCP biomimicry mutation handlers update scenario registries and do not change an active agent’s AgentDNA.
- Selection: https://genos.work/en/concepts/evolution-selection — A policy for comparing candidates and retaining outcomes under explicit criteria.
- Plasmids: https://genos.work/en/concepts/plasmids — Transferable capability packages that can be acquired, checked, and applied under runtime rules.
- Horizontal transfer: https://genos.work/en/concepts/horizontal-transfer — A governed transfer of capabilities or artifacts between agent lineages.
- Swarm intelligence: https://genos.work/en/concepts/swarm-intelligence — Local coordination and optimization patterns implemented as bounded organization steps.
- Stigmergy: https://genos.work/en/concepts/stigmergy — Coordination through traces deposited in a shared environment and read by later work.
- Niches: https://genos.work/en/concepts/niches — Declared regions of a search or work environment associated with resources and specialized capabilities.
- Populations: https://genos.work/en/concepts/populations — Groups of agents or candidate solutions whose composition can change under explicit policies.
- Cryptobiosis & dormancy: https://genos.work/en/concepts/cryptobiosis — A bounded suspended state that preserves continuity until an explicit resume condition is met.
- Natural creative ecology: https://genos.work/en/concepts/natural-creative-ecology — A model of open-ended search through recombination, exploration, plasticity, and selection.
- Sensorium: https://genos.work/en/concepts/sensorium — Runtime structures for observations and perception signals, with boundaries between simulated and real inputs.
- Metabolism: https://genos.work/en/concepts/metabolism — Resource accounting and allocation rules that constrain the work an agent can perform.
- Homeostasis & allostasis: https://genos.work/en/concepts/homeostasis — Target regulation and anticipatory adjustment under changing conditions and declared bounds.
- Resilience & recovery: https://genos.work/en/concepts/resilience — Recovery planning and bounded continuation after a failure, loss, or interruption.
- Development: https://genos.work/en/concepts/development — A staged process for changing an agent or project while retaining lineage, constraints, and evidence.
- Symbionts: https://genos.work/en/concepts/symbionts — Specialist services composed by a host under explicit capability and execution contracts.
- Web foraging: https://genos.work/en/concepts/web-foraging — A resource-selection calculation for web exploration; it is not by itself a continuous browser controller.
- Foveation: https://genos.work/en/concepts/foveation — A process for selecting a region of an observation for higher-resolution inspection.
- Animal senses: https://genos.work/en/concepts/animal-senses — A catalog of sensory mechanisms used as constrained design analogies for runtime perception.
- Animal control primitives: https://genos.work/en/concepts/animal-control-primitives — Bounded control patterns inspired by animal behavior and translated into software operations.
- Adaptive epistemic immune system: https://genos.work/en/concepts/adaptive-epistemic-immunity — A threat-response model for checking claims and evidence while preserving authority and provenance boundaries.
- Immune system: https://genos.work/en/concepts/immune-system — A set of threat detection and containment mechanisms; no absolute immunity is implied.
- Nosology: https://genos.work/en/concepts/nosology — A classification of runtime pathologies used to describe failure patterns and possible responses.
- Computational pathologies: https://genos.work/en/concepts/pathologies — A model of degraded or unsafe runtime states, not a medical diagnosis of people.
- Autoimmune failure: https://genos.work/en/concepts/autoimmune — A pathology category for safeguards that overreact to legitimate runtime behavior.
- Degenerative failure: https://genos.work/en/concepts/degenerative — A pathology category for progressive loss of runtime capability or state quality.
- Infectious failure: https://genos.work/en/concepts/infectious — A pathology category for unwanted propagation across agents, workspaces, or state boundaries.
- Genetic failure: https://genos.work/en/concepts/genetic-pathology — A pathology category for harmful inherited or encoded agent traits.
- Cancer-like failure: https://genos.work/en/concepts/cancerous — A pathology category for uncontrolled growth or replication in a runtime model.
- Metabolic failure: https://genos.work/en/concepts/metabolic-pathology — A pathology category for unsustainable resource use or allocation.
- Cardiovascular failure: https://genos.work/en/concepts/cardiovascular — A pathology category for disrupted flows or communication in a runtime model.
- Psychiatric failure: https://genos.work/en/concepts/psychiatric — A pathology category for modeled behavioral instability; it is not a claim about human mental illness.
- Environmental failure: https://genos.work/en/concepts/environmental-pathology — A pathology category for harmful interactions between a runtime and its operating environment.
- Lean verification: https://genos.work/en/concepts/lean-verification — A verification path that submits formal proof artifacts to a Lean kernel.
- ProofArtifact: https://genos.work/en/concepts/proof-artifact — A structured artifact that binds a formal claim, proof material, and verification result.
- Deterministic verification: https://genos.work/en/concepts/deterministic-verification — A repeatable check whose result depends on declared inputs and a bounded verifier.
- Mathematical organism: https://genos.work/en/concepts/mathematical-organism — A formal model of an agent system expressed through explicit mathematical structures and limits.
- Verification registry: https://genos.work/en/concepts/verification-registry — A registry of available verification contracts and their declared evidence requirements.
- SMT solver layer: https://genos.work/en/concepts/smt-solver — A solver integration surface whose maturity and runtime status must be checked against its implementation record.
- Deterministic procedures: https://genos.work/en/concepts/deterministic-procedures — Bounded procedures whose declared inputs and rules make their results reproducible.
- Snapshots: https://genos.work/en/concepts/snapshots — Versioned captures of runtime or agent state that can support restoration and comparison.
- Virtual filesystem: https://genos.work/en/concepts/vfs — A bounded filesystem interface used to scope access to project and runtime files.
- SQLite persistence: https://genos.work/en/concepts/sqlite — The local relational storage layer used by runtime services that persist sessions and operational state.
- Persistence: https://genos.work/en/concepts/persistence — Storage and rehydration mechanisms that preserve selected runtime state across calls or restarts.
- Shared state: https://genos.work/en/concepts/shared-state — Versioned data that multiple members can read or update under an explicit consistency policy.
- Model routing: https://genos.work/en/concepts/model-routing — Selection and dispatch rules for available inference providers and execution profiles.
- MCP interfaces: https://genos.work/en/concepts/mcp — Tool interfaces exposed through the Model Context Protocol with declared contracts and permissions.
- Sandbox: https://genos.work/en/concepts/sandbox — An execution boundary intended to constrain filesystem, process, or tool effects.
- Observability: https://genos.work/en/concepts/observability — Events and state views used to inspect runtime behavior and its evidence trail.
- Identity & authority: https://genos.work/en/concepts/identity-authority — Identity, role, and permission rules that constrain which actors may perform runtime operations.
- Governance: https://genos.work/en/concepts/governance — Approval, policy, and audit mechanisms that bound execution and promotion decisions.
- Workflows & jobs: https://genos.work/en/concepts/workflows — Persisted task graphs and jobs that coordinate execution, dependencies, and lifecycle events.
- Mind: https://genos.work/en/concepts/mind — Mind is retained as a philosophical reference for discussing cognition; GenOS does not implement a mental substance or claim subjective experience.
- Mental: https://genos.work/en/concepts/mental — Mental denotes the family of states and processes discussed by philosophy of mind, not a single executable GenOS capability.
- Consciousness: https://genos.work/en/concepts/consciousness — Consciousness names subjective experience in its phenomenal sense; the GenOS registry treats it as a reference-only concept rather than a verifiable software property.
- Operational consciousness: https://genos.work/en/concepts/operational-consciousness — Operational consciousness is GenOS vocabulary for internal consistency regulation using dissonance, budget and bounded stop signals; it is not phenomenal consciousness.
- Functional consciousness: https://genos.work/en/concepts/functional-consciousness — Functional consciousness groups measurable access, attention, self-model and metacognitive indicators without treating them as evidence of subjective experience.
- Perception: https://genos.work/en/concepts/perception — Perception is a bounded bridge from canonical observations to memory and control; it does not establish biological sensation or direct access to the world.
- Interoception: https://genos.work/en/concepts/interoception — Interoception reads internal software signals such as budget, dissonance and health for allostatic planning; it is not visceral sensation.
- Imagination: https://genos.work/en/concepts/imagination — Imagination recombines stored fragments into bounded candidate scenarios; generated possibilities remain simulations rather than observations.
- Internal simulation: https://genos.work/en/concepts/internal-simulation — Internal simulation predicts possible outcomes before execution using an explicit trace; its result is advisory and never proof of the real outcome.
- Reflection: https://genos.work/en/concepts/reflection — Reflection monitors reported confidence and errors so a system can revise or abstain; it does not demonstrate phenomenal reflexivity.
- Metacognition: https://genos.work/en/concepts/metacognition — Metacognition measures confidence, discrimination and overconfidence on resolved attribution history; it is a functional evaluation, not meta-consciousness.
- Calibration: https://genos.work/en/concepts/calibration — Calibration compares stated confidence with resolved outcomes and updates confidence conservatively; it never makes evidence more permissive by itself.
- Cognitive dissonance: https://genos.work/en/concepts/cognitive-dissonance — Cognitive dissonance is a software inconsistency signal derived from errors, repetition, drift and health; it is not a clinical or subjective diagnosis.
- Autobiographical memory: https://genos.work/en/concepts/autobiographical-memory — Autobiographical memory stores salient agent episodes, lessons and conditional recall; stored history remains fallible and must be revalidated before use.
- States of consciousness: https://genos.work/en/concepts/consciousness-states — States of consciousness are retained as a reference taxonomy; GenOS may expose functional levels but does not measure phenomenal states.
- Metric–experience boundary: https://genos.work/en/concepts/metric-experience-boundary — The metric–experience boundary prevents functional scores from being reported as evidence of subjective experience.
- Falsifiable hypothesis: https://genos.work/en/concepts/falsifiable-hypothesis — A falsifiable hypothesis is an explicit proposition paired with observations that could contradict it.
- Hypothesis ledger: https://genos.work/en/concepts/hypothesis-ledger — The hypothesis ledger preserves hypotheses, falsifications and revisions as append-only history instead of silently replacing prior states.
- Prediction: https://genos.work/en/concepts/prediction — A prediction states an observable expected outcome from a model or hypothesis before that outcome is resolved.
- Counterexample: https://genos.work/en/concepts/counterexample — A counterexample is an observation satisfying a claim’s preconditions while contradicting its predicted consequence.
- Direct evidence: https://genos.work/en/concepts/direct-evidence — Direct evidence is a receipt, test output or artifact explicitly associated with a claim; association alone does not establish business truth.
- Independent evidence: https://genos.work/en/concepts/independent-evidence — Independent evidence comes from a producer sufficiently separate from the claimant under a declared independence policy.
- Evidence reception: https://genos.work/en/concepts/evidence-reception — Evidence reception validates and normalizes an incoming evidence report before attaching it to a worker or claim dossier.
- Grounding: https://genos.work/en/concepts/grounding — Grounding links a symbolic claim to observations or artifacts that connect it to an operational context.
- Epistemic state: https://genos.work/en/concepts/epistemic-state — An epistemic state records evidence quality, derived confidence, phase and unresolved epistemic debt for a claim or belief.
- Formal result: https://genos.work/en/concepts/formal-result — A formal result is produced by a deterministic procedure with a bound receipt and declared verification semantics.
- Unverified result: https://genos.work/en/concepts/unverified-result — An unverified result is archived as provisional with explicit epistemic debt and cannot support promotion.
- Promotion eligibility: https://genos.work/en/concepts/promotion-eligibility — Promotion eligibility is an explicit decision that is true only when every policy obligation for a candidate is satisfied.
- Belief revision: https://genos.work/en/concepts/belief-revision — Belief revision updates a versioned claim when new relevant evidence arrives while retaining the prior state and contrary evidence.
- Epistemic dissonance: https://genos.work/en/concepts/epistemic-dissonance — Epistemic dissonance records conflict between beliefs and evidence or among incompatible beliefs without silently resolving it.
- Bias: https://genos.work/en/concepts/bias — Bias denotes systematic steering or calibration error detected under a declared probe; a detected pattern requires scoped interpretation.
- Hallucination indicator: https://genos.work/en/concepts/hallucination — The hallucination concept marks observable software conditions such as unsupported claims, untested code proposals, placeholders or empty values; it is not a medical diagnosis.
- Hallucination monitoring: https://genos.work/en/concepts/hallucination-monitoring — Hallucination monitoring continuously records declared unsupported-claim indicators and their context without converting counts into truth judgments.
- Weighted consensus: https://genos.work/en/concepts/weighted-consensus — Weighted consensus aggregates attributed agent contributions while preserving dependence and uncertainty; agreement is not automatic truth.
- Quorum with abstention: https://genos.work/en/concepts/quorum-abstention — Quorum with abstention refuses a collective decision when integrity, confidence or uncertainty conditions do not support one.
- Evidence barrier: https://genos.work/en/concepts/evidence-barrier — An evidence barrier checks that every cited claim and artifact belongs to the attributed source dossier before a collective result can advance.
- Promotion gate: https://genos.work/en/concepts/promotion-gate — A promotion gate applies the candidate’s declared replay, independent-verification, approval and preservation policies before any integration.
- Agent ecosystem: https://genos.work/en/concepts/agent-ecosystem — An agent ecosystem relates agents, niches and biomes through explicit capability and resource constraints; it is a software ecology analogy.
- Genomic stability: https://genos.work/en/concepts/genomic-stability — Genomic stability tracks mutation rates, anomalies and Stable, Compromised or Critical thresholds in versioned software genomes.

## Topologies (8)
- Trinity [partial]: https://genos.work/en/topologies/trinity — Three-world baseline with twelve controlled variants.
- A-Team [partial]: https://genos.work/en/topologies/a-team — Specialists, handoffs and integration.
- Biocenosis [partial]: https://genos.work/en/topologies/biocenose — Community evaluation and explicit ballots.
- Holobiont [partial]: https://genos.work/en/topologies/holobionte — A host, symbionts and lifecycle contracts.
- Syncytium [partial]: https://genos.work/en/topologies/syncytium — Persisted shared-state sessions and invariant-aware operations.
- Rhizome [implemented]: https://genos.work/en/topologies/rhizome — Persisted capability graphs, bounded routing and verified growth.
- Metapopulation [partial]: https://genos.work/en/topologies/metapopulation — Regional populations, verified migration and recovery.
- Biome [partial]: https://genos.work/en/topologies/biome — Persisted niches, measured resources and bounded transactional cycles.

## Product claims (28)
- Versioned workspace state [partial; integration]: Snapshot, fork, diff and replay operations are available for supported state. Open: Coverage and recovery vary by path; external effects and processes are not automatically restored.
- Eight topology modes [partial; integration]: Eight modes have composition and coordination services. Open: Each mode has distinct integration and end-to-end gaps; registration does not establish feature parity.
- Trinity variants [partial; integration]: Twelve runners share comparison, isolated execution, bounded research and persisted resume guards with targeted runtime tests. Open: The pre-correction R3 campaign ended 12/12 ESCALATE and 0 merge; no corrected-path real-mission campaign is published.
- A-Team specialist assignment [partial; integration]: Structured work graphs, independent scheduling, versioned handoffs and program-runner multi-team sub-runs connect to proof-gated completion. Open: Text-only formation remains lexical; general expertise and generic dispatch outside the qualified program path remain unproven.
- Biocenosis live missions [partial; integration]: Twelve executable variants share persistent deliberation, transactional round receipts, resume and proof-gated finalization. Open: Models, independent oracles and action adapters remain explicit dependencies. Runtime tests do not establish causal superiority; the dated October 4 campaign reported no verified promotion.
- Syncytium shared state [partial; integration]: Causal replay, retry fingerprints, persistence rollback and committed-state validation are connected. Open: Local tests do not establish universal convergence, distributed consensus or general worker correctness.
- Rhizome graph routing [implemented; end-to-end]: The provider-backed path resolves, executes, verifies and admits a capability before adding it to persisted routing; the 36-suite Rhizome profile includes an SQLite growth-and-routing E2E. Open: Hosts must still register real providers and verifiers; composition alone does not start them or prove mission quality.
- Holobiont lifecycle [partial; integration]: The shared mission runtime connects capability admission, independent verification, immune gates, quotas and atomic contribution/memory updates; the 55-script suite includes robust host-veto cases. Open: Physical isolation and measurements remain adapter responsibilities; no general or longitudinal gain is established.
- Biome resource cycles [partial; integration]: Eleven variants share bounded transactional cycles, authorized adapters, measured consumption and independent verification, with SQLite restart tests. Open: A global goal requires its own verifier; comparative ecological benefit and real provider quality remain unqualified.
- Metapopulation recovery [partial; integration]: The autonomous regional runtime persists reviewed migration, recolonization, local evidence and recovery; 14 topology suites and a minimal external adapter qualification are recorded. Open: Global qualification and the quality of host-provided evaluators, receivers and solvers remain open.
- Nineteen dynamic organizations [partial; integration]: All nineteen declared organizations now have deterministic runOrganizationStep coverage and executable browser counterparts on this site. Open: A guidance policy is not autonomous dispatch, and no general improvement in mission outcomes follows from algorithm coverage alone.
- Morphogenesis topology plugins [partial; integration]: Registered plugins can execute a validated graph on the explicit opt-in path. Open: Execution receipts are unverified traces; graph completion does not prove a mission outcome or apply a transition.
- Qualified native workers [partial; benchmark]: The inventory covers 19 WorkerKind values, 25 methods and 24 benchmark cases; all 19 types have C2/C3 evidence on one bounded task, with independent oracles and persisted mission receipts. Open: This does not establish correctness on arbitrary missions, complete qualification of every method, or broad parity with other agent systems.
- MCP interfaces [partial; integration]: JS stdio and backend dispatch enforce leases and direct calls. Open: Rust parity and individual tool nominal/refusal coverage vary by tool.
- Continuous web perception [experimental; unit]: Web observation and foveation handlers exist. Open: A complete observation–action–verification mission loop is not published.
- G-CIR obligation compilation [partial; integration]: The Node Omega gateway executes six guarded operations with persisted session visibility, MMU resolution and verification receipts. Open: General free-form mission compilation, complete Rust production bindings and empirical PGO qualification remain open.
- Relational physiology [partial; integration]: A kernel and targeted routing are integrated. Open: General mission integration is pending; relation constraints cannot grant new authority.
- Signal Plane delivery [partial; integration]: Persisted deliveries, an inbox and ACK states, cognitive escalation and leased handler retries are connected. Open: Receiver registration is not wired at server startup; delivery is at least once and external actions require idempotency.
- Mission continuity [partial; integration]: Completion, feedback, dormancy and regeneration have connected paths, including a two-process SQLite WAL authority test. Open: A multi-instance deployment has not been qualified.
- Epistemic meristem [partial; integration]: Transactional experiment waves connect verified coverage, Trinity hypotheses and Biome niches under explicit contracts. Open: Host adapters must ensure real isolation and verification; empirical advantage is not qualified.
- Unblocking spiral [partial; integration]: Bounded counterfactual search connects authorized interventions, verified history and snapshot restoration. Open: Host restoration and general mission outcomes remain domain-specific and need comparative evidence.
- Counterexample cambium [partial; integration]: Declarative procedures support isolated compression replay, transactional consolidation, scoped recall and transitive invalidation. Open: Known-case preservation does not establish general agent-memory safety or coverage of future failures.
- Aperiodic chronotaxis [partial; integration]: Resident ticks dispatch read-only probes and persist observed coverage, failures and missed windows. Open: Creating a schedule does not start a daemon; field detection and coverage remain unqualified.
- Contracted infinity [partial; integration]: Preregistered evaluation units, inherited risk grants, provenance checks and signed receipts guard protected topology promotions. Open: Valid individual statistical tests and field evidence remain required; these gates do not replace deterministic safety checks.
- Natural creative ecology [partial; integration]: Native numerical procedures, POET snapshots, cultural reuse and persisted phenotype run with SQLite reopen and six executed ablation arms. Open: Open-ended creativity, arbitrary user tasks and all 25 conceptual mechanisms are not demonstrated; search costs are not equalized across arms.
- GQWF v2 durable workspaces [partial; integration]: Versioned roots, immutable bases, private overlays, signed inventory, explicit materialization and candidate ingestion are connected to opted-in supervised missions. Open: No head is published automatically; general promotion still requires a path-specific verifier and real runtime evidence.
- Bounded philosophy audits [partial; integration]: All 420 registry entries map to an explicit bounded audit over eleven shared primitives; missing or violated observations create verification work instead of success. Open: These audits do not validate philosophical theories, external truth, subjective consciousness or general mission benefit.
- Endogenous Thought Test v1 [partial; integration]: A deterministic, read-only chain detects an internal signal, formulates and triages a falsifiable question, queues it through Ontogenesis and persists a bounded unverified proposal. Open: The concept document specifies independent classification, but the verifier service and test are absent from the pinned published tree. Learning and the reserved comparative campaign are also not implemented; ETT is not a test of phenomenal consciousness.

## Rules
- A simulation is not a run. A run is not a benchmark. A benchmark is not a general proof.
- Implementation, integration, and evidence are three separate axes.
- FR pages without FULL status are SUMMARY pages linking to the full EN page.