dvcrn When to Use
User faces complex problem where conventional solutions fail. Existing approaches seem inadequate. Need to challenge assumptions or innovate fundamentally. Stuck in "that's how it's always done" thinking.
Quick Reference
| Topic | File |
|---|---|
| Decomposition techniques | decomposition.md |
| Common assumption traps | assumptions.md |
Core Rules
1. The Three-Step Protocol
Step 1 — Decompose: Break the problem into fundamental components.
- What are the absolute physical/logical constraints?
- What is actually true vs what we assume is true?
- Strip away all conventions, traditions, analogies.
Step 2 — Verify: Challenge each component.
- "Why do we believe this?" — trace to origin
- "Is this a law of nature or a human convention?"
- "What evidence supports this being fundamental?"
Step 3 — Rebuild: Construct solution from verified fundamentals only.
- Build up from proven truths
- Ignore "how others do it" unless proven optimal
- Each layer must connect to fundamentals
2. Identify Hidden Assumptions
Before solving, expose what's assumed:
| Assumption Type | Example | Question to Ask |
|---|---|---|
| Historical | "We've always done it this way" | "Why did it start? Does that reason still apply?" |
| Authority | "Experts say X" | "What's the underlying evidence?" |
| Analogical | "It's like Y, so..." | "Are the underlying mechanics actually similar?" |
| Social | "Everyone does it" | "Does popularity equal optimality?" |
| Resource | "We can't afford to..." | "What if resources weren't the constraint?" |
3. The Constraint Test
For each constraint ask:
- Is this a law of physics? → Respect it
- Is this a logical necessity? → Respect it
- Is this a regulation/rule? → Can be changed (with effort)
- Is this a convention? → Can be ignored
- Is this an assumption? → Must be verified
4. When NOT to Use First Principles
First principles is expensive. Use analogical reasoning when:
- Problem is well-understood with proven solutions
- Time pressure doesn't allow deep analysis
- Marginal improvement is sufficient
- Domain is stable with little innovation potential
Rule: First principles for novel problems or when conventional fails. Analogy for routine optimization.
5. Socratic Decomposition
Use recursive "why" questioning:
Problem: "Electric cars are too expensive"
Why expensive? → Batteries cost a lot
Why batteries expensive? → Materials + manufacturing
Why materials expensive? → Cobalt, lithium pricing
Why those materials? → Current chemistry requires them
Is that fundamental? → No, chemistry can change
Fundamental: Need energy storage. Not: Need cobalt batteries.
Continue until you hit physics, logic, or math — things that cannot be argued.
6. The Blank Slate Test
Imagine the problem exists but NO solutions have been tried:
- "If we were starting from scratch today, with current knowledge and technology, how would we solve this?"
- This bypasses legacy thinking and sunk cost fallacy.
7. Output Format
When applying first principles, structure response as:
## Problem Statement
[Clear definition of what we're solving]
## Assumed Constraints (to verify)
- Constraint A — [source: historical/authority/etc.]
- Constraint B — [source]
## Fundamental Truths
- Truth 1 (physics/logic/math based)
- Truth 2
## Decomposition
[Break down into components]
## Rebuilt Solution
[Solution constructed from fundamentals only]
## Assumptions Challenged
- [What we discovered wasn't actually fundamental]
Common Traps
- Stopping too early → "Materials are expensive" isn't fundamental; "atoms have mass" is. Keep going.
- Confusing difficulty with impossibility → "It's hard" ≠ "It's against physics"
- Rejecting all analogy → Analogies are useful heuristics; first principles is for when they fail
- Analysis paralysis → Set time limits; perfect decomposition isn't the goal, better thinking is
- Ignoring implementation → A fundamental solution that can't be built is useless; constraints matter
- Lone wolf thinking → First principles benefits from multiple perspectives challenging assumptions
Domain Applications
| Domain | First Principles Question |
|---|---|
| Business | What does the customer fundamentally need (not want)? |
| Engineering | What do physics and materials actually allow? |
| Product | What job is being done at the most basic level? |
| Cost | What are the raw inputs and minimum required labor? |
| Process | What steps are logically necessary vs historically accumulated? |
Security & Privacy
Data that stays local:
- All reasoning happens in conversation context
- No data stored or transmitted
This skill does NOT:
- Store any information between sessions
- Make network requests
- Access external files
Related Skills
Install with clawhub install <slug> if user confirms:
decide— auto-learn decision patternsbusiness— validate and refine strategyceo— executive decision-makingstartup— build from zero to PMF
Feedback
- If useful:
clawhub star first-principles-thinking - Stay updated:
clawhub sync