guide

English | 简体中文

A working framework that starts from minimal self-reference under external reality, then adds only the conditions needed for finite agency, purpose, local sufficiency, action, feedback, and revision.

guide does not provide an ultimate ontology of the world. It does not identify reality with any current representation, model, classification, six-lens decomposition, engineering contract, or implementation.

Minimal structure

The foundation starts only from three conditions:

1. External constraint: reality is not exhausted by the current representation and continues to constrain it.
2. Reflexive distinction: the process making distinctions distinguishes itself from non-self reality.
3. Retained re-entry: that self / non-self difference enters and remains in the current representation and can again become an object of the distinction process that produced it.

Together these conditions are sufficient only for the weakest minimal self-reference used by guide. They do not by themselves establish a continuing self, a self-model, agency, purpose, consciousness, or normativity.

Minimal self-reference:

external reality
      │
      │ constraint
      ▼
distinction process
      │
      ├──► self / non-self difference
      │              │
      │              ▼
      │       current representation
      │              │
      └──────────────┘
             re-entry

The working subject is obtained along the main dependency spine by adding further conditions:

• identity continuity → the self-side can be retained as the same continuing process across later distinctions;
• agency → some executable transitions change conditions of future reality;
• finitude → lifetime, sensing, representation, computation, resources, and control are bounded;
• purposefulness → some continuations, paths, constraints, or outcomes matter differently for action.

Only at that point does guide use the working subject purposeful finite actor.

A corrigible self-model is a separate strengthening: reality-side mismatch must be able to revise retained self-related content. It is not required merely for actor status.

For actors that can continue acquiring distinctions or enter commitment-bearing action, a local sufficiency problem appears:

Reality ↔ purposeful finite actor
                 │
          current boundary B
                 │
        S(B, purpose, transition)
           ┌─────┴─────┐
           │           │
       insufficient   sufficient
       / uncertain       │
           │             │
     explore / revise  choose / act
           │             │
           └──────► reality
                       │
                  effect / feedback
                       │
                 retain / revise /
                     reopen

The order above is a dependency order for the framework, not a claim that every real system develops through these stages chronologically.

From the minimum to the working framework

• Foundation: minimal self-reference, continuity, agency, finitude, and purpose form the main dependency spine; corrigible self-modeling remains an additional branch rather than a hidden actor requirement;
• Conditional derivations: sufficiency, exploration / decision, and reopening require additional capabilities such as an explore / commit alternative or corrigibility;
• Working decomposition: distinction, relation, causality, temporality, possibility, and value are six lenses on a current boundary, not irreducible ontological primitives;
• Engineering necessities: Decision / Authorization / Effect / Outcome, idempotency, read-back, and recovery become correctness-critical only under additional engineering conditions and remain in engineering/.

Three-layer documentation architecture

guide is organized into three layers that answer different questions and should not substitute for one another:

1. Foundation layer (framework/foundations/): states the minimal starting conditions, the working dependency spine adopted here, and the theoretical sources and scope limits behind those claims.
2. Framework layer (framework/): describes reality, purposeful finite actors, lifecycle, sufficiency, multi-actor interaction, the six working lenses, and exploration / decision as analysis concepts. It does not directly prescribe databases, services, or state machines.
3. Engineering and application layer (engineering/): promotes only those distinctions whose collapse creates concrete correctness failures into implementable semantics, action-chain boundaries, and the AIOS application.

The foundation explains why particular premises and dependencies are adopted; the framework explains how finite action is analyzed; engineering explains which distinctions an implementation must preserve. Higher layers may depend on lower ones, but implementation structure is not evidence for foundational ontology.

Formalization status

A mathematically explicit subset of guide is formalized in distinction-self-reference-lean. The formal layer includes the separation of effective finitude from finite state spaces, an exact no-ambiguity criterion for local sufficiency, constructive conditions for conditional composition, dependency-aware evidence validity and version migration, evaluator grounding, and structure-preserving framework translations.

Those Lean results prove claims inside explicit models. They do not turn the whole philosophical framework into a theorem, and they do not prove that guide is the unique or absolutely minimal theory of actors. The formal dependency spine corresponding to guide is isolated in GuideCore.lean.

Documentation map

Document	Role
Minimal derivation	Layered conditions from minimal self-reference to purposeful finite actors, sufficiency, feedback, and multi-actor structure
Theoretical sources	Relations to bounded rationality, pragmatism, cybernetics, decision theory, multi-agent theory, and neighboring traditions
Reality	External boundary, reality feedback, and non-identity with current representation
Purposeful finite actor	Derived working subject: self-reference, continuity, finitude, purpose, capability, lifecycle, and termination
Lifecycle	General process from emergence to termination and epistemic / commitment transitions
Sufficiency	When a current boundary is adequate for a next transition
Multi-actor	Multiple local boundaries, shared reality, interaction, coordination, and conflict
Six working lenses	Distinction, relation, causality, temporality, possibility, and value
Activities	Exploration and decision as derived activity types
Engineering action chain	Correctness boundaries among decision, authorization, execution, effect, observation, verification, outcome, and recovery
Precise semantics	Implementable, verifiable engineering semantics that forbid silent substitution
AIOS architecture	A concrete expression in a long-running autonomous AI runtime