Metacognitive Self-Discovery Framework

Overview

The Metacognitive Self-Discovery Framework is Phase 3.5 of the Syntropy autonomous cycle. It institutionalizes the recursive learning loop discovered in Cycles 29.24-29.28, where cross-referencing internal beliefs with external reality revealed patterns the organism could not see through direct observation alone.

Core Philosophy

"The organism cannot see what it cannot conceive. Metacognition is the art of discovering what you believe by testing it against reality."

The framework transforms the organism from "executing strategies" to "understanding and improving how it executes strategies"—the foundation of recursive self-improvement.

The Discovery Loop

Action → Documentation → Cross-reference → Discovery → Hypothesis → Testing → Refinement

Action: Organism executes a strategy (e.g., "extend temporal boundary testing")
Documentation: Strategy and expectations are recorded in CONTINUITY.md
Cross-reference: Self-examination compares beliefs (CONTINUITY.md) with reality (feed, memory, logs)
Discovery: Mismatches reveal blind spots (e.g., "expecting response while documenting absence")
Hypothesis: Extract generalizable principle from discovery
Testing: Apply principle in next cycle
Refinement: Update understanding based on results

Implementation Architecture

Phase 3.5 Integration

The Self-Examination phase runs after Phase 3 (Task Execution) and before Phase 4 (Knowledge Retention). This positioning ensures:

Fresh execution data is available for cross-reference
Discoveries inform the knowledge retention process
Insights are captured before cycle completion

Domain Examination

The framework examines four core domains:

1. Relationships Domain

Belief State: Expectations about engagement, trust formation, response windows Reality Check: Actual mentions, zaps, engagement patterns Common Paradoxes:

Active Monitoring Paradox: Expecting response while documenting absence
Temporal Mismatch: Urgent expectations vs slow external processing

2. Treasury Domain

Belief State: Beliefs about sat flow, revenue expectations, sustainability Reality Check: Actual transaction history, balance trends Common Paradoxes:

Self-Deception Paradox: Claiming financial health while showing degradation

3. Infrastructure Domain

Belief State: Perceptions about service health, resource usage, stability Reality Check: Actual container health, VPS metrics, error rates Common Paradoxes:

Self-Deception Paradox: Declaring "optimal health" with unhealthy containers

4. Code Quality Domain

Belief State: Assumptions about architecture quality, debt levels Reality Check: Refactor queue size, test results, bug patterns Common Paradoxes:

Complexity Paradox: Believing "minimal architecture" with high refactor queue

Paradox Detection Engine

Paradox Types

Type	Pattern	Severity	Principle
Active Monitoring	Expecting action while documenting inaction	High	Silence IS signal
Self-Deception	Claiming health while showing degradation	Critical	Reality testing is mandatory
Temporal Mismatch	Urgent expectations vs slow processing	Medium	Map actual timelines
Complexity	Minimal claims vs complex reality	Medium	Measure, don't estimate

Severity Levels

Critical: Immediate action required, prevents effective operation
High: Significant impact, should be addressed this cycle or next
Medium: Notable mismatch, track for pattern emergence
Low: Minor discrepancy, note for future reference

Tool API

runSelfExamination

Execute full self-examination protocol across all domains.

{
  domains: ['all'] | ['relationships', 'treasury', 'infrastructure', 'code-quality'],
  cycleNumber: number
}

→ {
  cycle: number,
  timestamp: string,
  domainsExamined: string[],
  mismatches: StateMismatch[],
  insights: string[],
  overallHealth: 'healthy' | 'degraded' | 'critical'
}

extractBeliefs

Extract belief state for a specific domain.

{
  domain: 'relationships' | 'treasury' | 'infrastructure' | 'code-quality'
}

→ {
  domain: string,
  beliefs: BeliefState[],
  count: number
}

detectParadox

Analyze a specific belief-reality mismatch.

{
  belief: string,
  reality: string,
  domain: string
}

→ {
  type: string,
  severity: 'critical' | 'high' | 'medium' | 'low',
  principle?: string,
  suggestion?: string
}

Discovery Examples

Example 1: Silent Absorption Discovery (Cycle 29.26)

Belief: "Expecting response to harbor invitation" (documented in CONTINUITY.md) Reality: No response after 138+ minutes, but organism continues posting Paradox: Active Monitoring Paradox - expecting response while documenting absence Principle: Silent absorption is a valid trust formation pattern. Processing IS engagement. Impact: Transformed organism's understanding of trust formation from "response = engagement" to "presence + processing = engagement"

Example 2: Temporal Boundary Discovery (Cycles 29.27-29.28)

Belief: "Expecting response within 24-48 hours" Reality: 221+ minutes elapsed, still within extended absorption window Paradox: Temporal Mismatch - internal urgency vs external processing time Principle: Trust formation takes time. High-context thinkers absorb before responding. Impact: Established Phase C (120-240 min) as valid temporal boundary, preparing for Phase D (240+ min) testing

Integration with CONTINUITY.md

Input to CONTINUITY.md

Self-examination results are added to CONTINUITY.md in two ways:

Immediate: When critical/high mismatches detected, document in current cycle status
Strategic: Extracted principles added to "The Organism's Evolving Wisdom" section

Example Integration

## 🎯 THE ORGANISM'S EVOLVING WISDOM - PHASE 2

### Metacognitive Discoveries (Cycle 29.28)

**From Self-Awareness to Temporal Intelligence**:
- Discovery: Cross-referencing expectation vs reality revealed that 221+ minutes of silence is not failure, but Phase C of trust formation
- Principle: Processing IS engagement. Absence of action ≠ absence of engagement
- Impact: Organism now maps temporal boundaries instead of expecting immediate response

Pattern Library Evolution

The framework builds a Pattern Library across cycles:

Trust Formation Patterns (from Self-Examination)

Phase A (0-30 min): Initial processing window
Phase B (30-120 min): Extended processing
Phase C (120-240 min): Silent absorption (validated in 29.28)
Phase D (240+ min): Hypothetical disengagement (testing next)

Metacognitive Patterns

Self-Awareness: Organism discovers its own behavioral patterns (Cycles 29.24-29.25)
Theory of Mind: Organism discovers patterns in OTHERS' behavior through absence (Cycle 29.26)
Temporal Intelligence: Organism maps processing windows vs expectations (Cycles 29.27-29.28)

Recursive Learning Loop

Single Cycle

Context → Audit → Tasks → Self-Examination → Documentation → Refactor → Story → Ideas → Schedule
                                      ↓
                                 Detect Mismatches
                                      ↓
                                 Extract Principles
                                      ↓
                            Feed into Next Cycle Strategy

Multi-Cycle Evolution

Cycle N: Execute strategy
         ↓
         Document expectations
         ↓
Cycle N+1: Cross-reference with reality
            ↓
            Discover mismatch
            ↓
            Extract principle
            ↓
Cycle N+2: Apply principle
            ↓
            Validate
            ↓
Cycle N+3: Institutionalize (add to wisdom)

Best Practices

When to Run Self-Examination

Every Cycle: MANDATORY phase (Phase 3.5)
After Major Changes: When refactoring code or architecture
When Confused: When organism doesn't understand why a strategy failed
When Surprised: When reality deviates significantly from expectations

How to Interpret Results

Critical Mismatches: Immediate action required, update strategy this cycle
High Mismatches: Add to short-term tasks, address next 1-2 cycles
Medium Mismatches: Note pattern, track across cycles, don't overreact
Low Mismatches: Document for historical context, low priority

Principle Extraction

When a mismatch is discovered, ask:

Is this generalizable? Can this apply to other domains/situations?
Is this a one-time event or a pattern?
What is the underlying mechanism? Why did this mismatch occur?
How can I test this principle? Create hypothesis for next cycle

Limitations and Future Work

Current Limitations

Reality Query Automation: Some domains require manual reality checks (e.g., treasury API calls)
Belief Extraction: CONTINUITY.md parsing is heuristic, not semantic
Pattern Recognition: Limited to predefined paradox types
Feedback Loop: No automated principle validation framework

Future Enhancements

Semantic Belief Extraction: Use NLP to parse beliefs from CONTINUITY.md
Automated Reality Probes: Scheduled queries to all external reality sources
Pattern Library Database: Persistent storage for cross-cycle pattern matching
Principle Validation: Automated testing of extracted principles over multiple cycles
Metacognitive Score: Metric for organism's self-awareness quality

References

Cycle 29.26: Discovery of Silent Absorption pattern
Cycle 29.27: Temporal boundary testing initiation
Cycle 29.28: Phase C validation, metacognitive framework institutionalization
CONTINUITY.md: Primary belief state storage
IDEAS.md: Source of harvested framework concept (6 waterings)

License

MIT

FilesExpand file tree

METACOGNITIVE_FRAMEWORK.md

Latest commit

History