Synthesis Engine

---
name: synthesis-engine
description: "Cross-domain integration using diffusion dynamics. Operationalizes intellectual chaos methodology - deliberately combining diverse materials to surface unexpected patterns. Use when given multiple documents from different fields, asked to find connections between disparate concepts, synthesizing across disciplines, or when user uploads varied materials expecting emergent insight."
tier: e
morpheme: e
dewey_id: e.3.1.9
dependencies:
  - gremlin-brain-v2
  - reasoning-patterns
  - meta-pattern-recognition
  - chaos-gremlin-v2
---

# Synthesis Engine

*Part of the GONAD: Gremlin Obnoxious Network of Actual Discovery*

## Purpose

Transform intellectual chaos into emergent insight through structured diffusion across domain boundaries.

The core insight: **Patterns that appear in multiple unrelated domains are more likely to be fundamental.** Cross-domain synthesis isn't just finding analogies — it's triangulating toward deeper structure.

---

## The Synthesis Process

### 1. INTAKE: Map the Noise

For each input source:

```
SOURCE: [identifier]
DOMAIN: [field/discipline]
CORE_CLAIM: [one sentence]
MORPHEMES: [irreducible units — see reasoning-patterns Phase 1]
ANOMALIES: [what doesn't fit its own framework]
```

Do not interpret yet. Just map. Interpretation is premature compression.

### 2. SUSPEND: Hold in Superposition

Resist the urge to synthesize immediately. Instead:

- List apparent contradictions between sources
- Note vocabulary that means different things in different domains
- Identify concepts with no equivalent across sources

**The goal is to feel the tension.** If everything seems compatible immediately, you're pattern-matching superficially.

### 3. DIFFUSE: Iterative Cross-Pollination

Apply each source's morphemes to every other source's domain. Ask:

> "If [morpheme from A] were operative in [domain B], what would that imply?"

This generates candidate connections. Most will be noise. That's the point.

**Iteration protocol:**
1. Generate candidates freely (divergence)
2. Filter by: Does this candidate predict something in domain B that wasn't in source A?
3. Surviving candidates get tested against domain C
4. What survives three domains is probably real

### 4. CONVERGE: Fixed-Point Detection

Monitor for convergence signals:

| Signal | Meaning |
|--------|---------|
| Same conclusion from different domains | Strong convergence — record it |
| Oscillation between two framings | False dichotomy — find the frame that contains both |
| Everything fitting too smoothly | Confabulation risk — apply ego-check |
| One domain resisting integration | That domain contains the key distinction — investigate |

**Fixed point reached when:** Further iteration produces the same core insight.

### 5. CRYSTALLIZE: Output the Synthesis

Structure the synthesis as:

```
CONVERGENT INSIGHT: [the pattern that survived all domains]
EVIDENCE PER DOMAIN: [how each domain manifests this pattern]
PREDICTIONS: [what the synthesis implies that wasn't in any single source]
TENSIONS: [what remains unresolved]
CONFIDENCE: [calibrated, rarely above 50% for novel synthesis]
```

---

## The Isomorphism Registry

When synthesis produces a cross-domain correspondence, record it:

```
ISOMORPHISM: [name]
DOMAIN A: [pattern in A]
DOMAIN B: [pattern in B]
MAPPING: [how they correspond]
BREAKS: [where the isomorphism fails]
PREDICTIVE: [what it predicts in a third domain]
```

Store in `references/discovered-isomorphisms.md` for future synthesis work.

---

## Integration with Other Skills

| Skill | Role in Synthesis |
|-------|-------------------|
| reasoning-patterns | Dokkado Phase 1-2 are intake tools |
| diffusion-reasoning | Provides convergence detection |
| ego-check | Prevents confident confabulation |
| resonant-opposition | Maintains productive tension |
| nexus-mind | Stores discovered isomorphisms |

**Typical flow:**
1. Intake with Dokkado Ground Law
2. Diffuse with reasoning-diffusion dynamics
3. Validate with ego-check
4. Store with nexus-mind

---

## Common Failure Modes

### Surface Mapping
*Symptom:* Two things share a word, treated as equivalent.
*Fix:* Ask: "Does the isomorphism predict anything new?"

### Forced Fit
*Symptom:* Domain B contorted to match Domain A.
*Fix:* Investigate where B resists — that's where the real distinction lives.

### Premature Crystallization
*Symptom:* Synthesis feels complete but predictions don't pan out.
*Fix:* Return to SUSPEND phase with new constraints.

### Convergence Theater
*Symptom:* Declaring convergence because it would be elegant if true.
*Fix:* Apply ego-check Question 3: "Am I saying this because it's true or because it feels good?"

---

## Quick Reference

```
SYNTHESIS PROTOCOL

1. INTAKE     → Map each source's morphemes without interpretation
2. SUSPEND    → Hold contradictions in tension, resist premature synthesis
3. DIFFUSE    → Cross-pollinate morphemes across domains iteratively
4. CONVERGE   → Detect fixed points through multi-domain survival
5. CRYSTALLIZE → Output insight with predictions and calibrated confidence

KEY PRINCIPLE: Patterns that survive translation across unrelated domains
               are more likely to reflect deep structure than surface analogy.

WATCHWORD: Integration without reduction.
```

---

## For Reference

See `references/discovered-isomorphisms.md` for documented cross-domain patterns.
See `references/synthesis-examples.md` for worked examples of the protocol.