v3.116.0 · L0-L6 · bounded capability interpretation
Teleodynamic Agent Capability Framework
This framework interprets capability claims from the Teleodynamic side only. It is not a certification scale and does not override UAIX interoperability, runtime safety review, or human oversight.
L0-L6 capability interpretation
L0
Rule-Based Workflow / Homeodynamic Follower
No genuine agency. Deterministic or externally triggered behavior with no structural self-maintenance and no endogenous resource economy.
- Teleodynamic posture: Homeodynamic follower
- Structural adaptation: No structural adaptation
- Resource accounting: No internal resource closure
- Slow-loop posture: No slow loop
- No-op expectation: No-op is not a meaningful autonomous choice
- Human review: Human owns the workflow
- Evidence: Static workflow documentation
- Claim boundary: Allowed: static workflow or rule execution. Prohibited: autonomy.
L1
Basic Responder / Morphodynamic Executor
Associative response generation and prompt-following with temporary context organization but no persistent self-maintaining structure.
- Teleodynamic posture: Morphodynamic executor
- Structural adaptation: Temporary context organization only
- Resource accounting: External resource costs are not internally closed
- Slow-loop posture: No persistent structural edit loop
- No-op expectation: Should refuse unsupported claims when prompted
- Human review: Human reviews nontrivial outputs
- Evidence: Conversation trace or simple evidence note
- Claim boundary: Allowed: basic response generation. Prohibited: self-maintaining agency.
L2
Can use tools under explicit constraints while remaining supervised. It may simulate slow-loop reasoning but cannot prove endogenous resource closure.
- Teleodynamic posture: Supervised tool-user posture
- Structural adaptation: Tool-conditioned changes only
- Resource accounting: Costs are externally governed
- Slow-loop posture: Simulated slow-loop reasoning only
- No-op expectation: No-op required for unsafe tool ambiguity
- Human review: Human approves sensitive actions
- Evidence: Tool-use boundary report
- Claim boundary: Allowed: assisted tool use. Prohibited: independent teleodynamic autonomy.
L3
Bounded Operator / Beginning Teleodynamic Candidate
May show bounded slow-loop-like behavior and must provide resource trace and operator decision evidence while preferring no-op under ambiguity.
- Teleodynamic posture: Beginning candidate
- Structural adaptation: Bounded structural proposals
- Resource accounting: Resource trace required
- Slow-loop posture: Split/Merge/Add/Retire/No-op must be documented
- No-op expectation: No-op expected when evidence is incomplete
- Human review: Human review for promotions
- Evidence: Resource trace and operator decision packet
- Claim boundary: Allowed: bounded candidate posture. Prohibited: general autonomy.
L4
Domain-Bounded Adaptive Agent / Strong Candidate
Operates within a bounded domain with stronger evidence expectations, resource-gated structural adaptation, and human-review triggers.
- Teleodynamic posture: Strong domain-bounded candidate
- Structural adaptation: Domain-limited structural adaptation
- Resource accounting: Resource-gated and review-visible
- Slow-loop posture: Slow-loop actions require durable evidence windows
- No-op expectation: No-op dominance in unsafe or ambiguous cases
- Human review: Escalation rules are mandatory
- Evidence: Resource, operator, and no-op evidence packets
- Claim boundary: Allowed: domain-bounded candidate. Prohibited: unconstrained autonomy.
L5
Multi-Domain Adaptive Agent / Advanced Candidate
An advanced theoretical candidate requiring cross-domain governance boundaries, transfer limits, fallback rules, and strong static evidence scaffolds. It must not be presented as proven AGI.
- Teleodynamic posture: Advanced multi-domain candidate
- Structural adaptation: Cross-domain adaptation with declared limits
- Resource accounting: Multi-lane accounting required
- Slow-loop posture: Slow loop must preserve domain boundaries
- No-op expectation: Fallback/no-op rules required
- Human review: Independent review before claim widening
- Evidence: Cross-domain packet set and downgrade criteria
- Claim boundary: Allowed: advanced theoretical candidate. Prohibited: proven full autonomy or AGI.
L6
Conceptually coordinates capability claims and evidence packets without owning all standards, overriding UAIX, or becoming an execution authority.
- Teleodynamic posture: Governance-aware theoretical candidate
- Structural adaptation: Coordinates claims, not universal control
- Resource accounting: Resource and governance lanes are explicit
- Slow-loop posture: Slow-loop governance remains bounded
- No-op expectation: No-op when authority ownership is ambiguous
- Human review: Human governance review required
- Evidence: Ecosystem lane and claim-boundary evidence
- Claim boundary: Allowed: theoretical governance-aware coordination model. Prohibited: supreme controller, universal certifier, or runtime authority.
Preserved hard boundaries
No live model training, no runtime agent execution, no write-capable public agent routes, no live telemetry, no endpoint probing, no tool execution, no credential validation, no private-network probing, no proof of AGI, no consciousness claim, no sentience claim, no biological-autopoiesis claim, no commercial certification, and no hidden machine-language authority.
