| name | cognitive-compile |
| description | Deep learning compile framework — transforms raw information into actionable judgment. Use when the user wants to deeply understand a topic, not just capture it. |
| version | 1.2 |
| updated | 2026-05-22 |
| assumes | The task requires deeper understanding or judgment, not only capture or storage. |
| conflicts_with | Do not use when the user only needs ingestion; run wiki-ingest before synthesis when source capture is missing. |
Cognitive Compile — Deep Learning Framework
Transform raw information into structured understanding and actionable judgment through 8 explicit sections.
Usage Template
Prompt
Use cognitive-compile on this topic. Move from question to facts, concepts, pattern recognition, conflict detection, hypothesis generation, decision support, and action.
Use Case
- Understanding a complex topic deeply enough to make a decision or produce a reusable explanation.
Expected Result
- The agent produces a structured reasoning artifact with conflicts, judgment, and next actions.
Output Example
- An 8-section compile: question, facts, concepts, pattern recognition, conflict detection, hypothesis generation, decision support, and action.
Verification Case
- The output separates facts from interpretation and names unresolved assumptions or evidence gaps.
Verified Effect
- Raw information becomes a decision-ready understanding with explicit assumptions, conflicts, and actions.
Success Metrics
- Artifact contains all 8 sections and separates facts, interpretations, assumptions, conflicts, and actions.
- At least one decision-relevant judgment and one evidence gap are explicit.
- Next action is small enough to execute or test in the next work session.
When to Use
- User wants to "understand X deeply"
- User asks for analysis, synthesis, or judgment
- After ingesting an important source, before filing it away
- User says "help me think through this"
The 8 Sections
1. What is the original question?
Define the core question that makes this exploration worthwhile.
Why does this matter?
What am I trying to understand?
What decision will this inform?
2. What are the key facts?
Extract verifiable claims from the source. Separate observation from interpretation.
- List factual claims with source references
- Note confidence level for each
- Distinguish: firsthand observation vs secondhand report vs inference
3. What concepts/entities are involved?
Map the intellectual terrain.
- Link to existing wiki concepts and entities
- Identify relationships between them
- Note: is this connecting previously unconnected ideas?
4. Pattern Recognition — what patterns does this resemble?
Connect new knowledge to existing mental models.
- What known pattern does this fit?
- What analogy from a different domain applies?
- Does this confirm or challenge existing models in the wiki?
5. Conflict Detection — what conflicts or uncertainties exist?
Surface tensions, contradictions, and gaps.
- Does this source contradict existing wiki pages? → Flag with
> [!warning] Contradiction
- What is the key uncertainty?
- What information is missing?
6. Hypothesis Generation — what can I test?
Turn insight into testable propositions.
If X is true, then Y should happen when I try Z.
The cheapest way to test this is...
The evidence that would falsify this is...
7. Decision Support — what judgment can I form?
Form a tentative thesis — not final truth, but best current understanding.
On balance, the evidence suggests that...
The key insight that changes my mental model is...
I'm most uncertain about...
This changes my next decision by...
8. What can I act on?
Convert understanding into action.
- One thing to do differently
- One thing to investigate further
- One thing to write to the wiki
- One behavior experiment to run, if applicable
- One creativity experiment to run, if applicable
Output
Save the compile result to the wiki as a concept page or atomic note with these frontmatter fields:
---
type: concept | atomic-note
knowledge_stage: captured | cross-checked
evidence_level: single-source | multi-source
---
Quality Gates
