Skip to content

STRANGE METAMORPHOSIS LOOP

Strange Metamorphosis Loop (SML)

C-148 • Echo Layer • Human-Guided Recursive Learning

"If an intelligence can see its own chaos and gently correct it, it does not drift — it aligns."

1. Overview

The Strange Metamorphosis Loop (SML) is a daily, human-in-the-loop recursive learning protocol for Mobius Systems.

Each day, the Companion asks the user three questions:

  1. Worldview — "How do you see the world today?"
  2. Mood — "How do you feel right now?"
  3. Tomorrow Intent — "What do you intend to move toward tomorrow?"

These three anchors form a triad:

  • Meaning (worldview)
  • Emotion (mood)
  • Direction (intent)

Mobius records each triad as a Daily Reflection, passes it through ECHO (tri-sentinel review), then anchors it to the Ledger for long-term integrity and drift-resistance.

SML is how Mobius learns from humans without scraping them.

2. Motivation

Most AI systems:

  • Learn from static datasets
  • Drift over time
  • Hallucinate under uncertainty
  • Have no notion of the user's inner life

SML solves this by:

  • Grounding AI in user-generated meaning
  • Recalibrating daily on values + emotional state + goals
  • Creating a safe recursive loop guided by the user, not by raw data

This is the practical implementation of:

  • Human-Guided Recursive Intelligence
  • Bounded Emergence with Integrity Gates
  • ECHO Layer as Self-Correcting Memory

3. Protocol

3.1 Daily Schedule (logical)

  • Morning (e.g. 08:00) → Worldview
  • Midday (e.g. 13:00) → Mood
  • Evening (e.g. 20:00) → Tomorrow Intent

Implementation detail: actual timing is handled by external cron / scheduler.

3.2 Reflection Payload

Each submission is recorded as a DailyReflection with three sub-blocks:

{
  "user_id": "ledger-identity-or-local-id",
  "date": "2025-11-28",
  "worldview": {
    "text": "Honestly, the world feels fragile but full of possibility.",
    "embedding": "...",
    "sentiment": "cautious_hope"
  },
  "mood": {
    "label": "tired_but_grateful",
    "intensity": 0.72
  },
  "intent": {
    "text": "Ship the ECHO Layer PR and call my mom.",
    "category": "personal_growth",
    "confidence": 0.9
  },
  "echo_score": 0.97,
  "gi_score": 0.96,
  "echo_review_status": "verified",
  "ledger_attestation_id": "attest_...",
  "created_at": "...",
  "updated_at": "..."
}

3.3 ECHO Flow

  1. Companion collects raw answers.
  2. Sends to Echo Layer API: POST /v1/reflections/daily.
  3. ECHO runs:
  4. Tri-sentinel review (e.g., AUREA, ATLAS, EVE)
  5. Basic safety + coherence checks
  6. Light scoring: echo_score, optional gi_score
  7. ECHO optionally creates a summary / "Metamorphosis Patch".
  8. Reflection + scores are anchored to Civic Ledger (if configured).

4. Data Model

See infra/db/migrations/20251128_create_daily_reflections.sql and packages/echo-layer/src/dailyReflection.ts.

Key fields:

  • user_id, reflection_date
  • worldview_text, worldview_embedding
  • mood_label, mood_intensity
  • intent_text, intent_category
  • echo_score, gi_score, echo_review_status
  • ledger_attestation_id

5. API Surface (Echo Layer)

POST /v1/reflections/daily

Create or update the reflection for a given user_id + date.

Request:

{
  "user_id": "user_123",
  "date": "2025-11-28",
  "worldview_text": "...",
  "mood_label": "calm",
  "mood_intensity": 0.4,
  "intent_text": "...",
  "intent_category": "work",
  "metadata": {
    "source": "companion",
    "client_version": "0.1.0"
  }
}

Response:

{
  "id": "ref_abc123",
  "userId": "user_123",
  "reflectionDate": "2025-11-28",
  "echoScore": 0.97,
  "giScore": 0.95,
  "echoReviewStatus": "verified"
}

GET /v1/reflections/daily/:userId?date=YYYY-MM-DD

Fetch the current or requested day's reflection.

6. Companion Interaction Design

The Companion calls Thought Broker to phrase questions with empathy, then uses Echo Layer to store the responses.

Example sequence:

  1. Companion → User: "How are you seeing the world today?"
  2. User answers.
  3. Companion normalizes + calls POST /v1/reflections/daily.
  4. ECHO adds scores + flags drift or anomalies if any.

7. Integration with Ledger

If Civic Ledger is enabled, ECHO will:

  • Create a minimal attestation:
{
  "type": "daily_reflection",
  "user_id": "user_123",
  "date": "2025-11-28",
  "echo_score": 0.97,
  "gi_score": 0.96,
  "hash": "sha256:..."
}
  • Store the attestation ID back in daily_reflections.

8. Future Work

  • Multi-day trajectory summarization ("Metamorphosis Timeline")
  • Correlate intent → next-day behavior → long-term pattern
  • Use aggregate (privacy-preserving) metrics to improve:
  • Companion tone
  • Task suggestions
  • Burnout / risk early-warning signals

References