First-principles decomposition of a problem

advancedClaude OpusPersonal ProductivityDecision Makingframeworkmethodologyfirst-principlesreasoningdecision-making

Use case

Use this when you suspect the standard answer to a problem is just the inherited one — pricing, org structure, a recurring process, a product approach. The XML structure separates 'what everyone believes' from 'what is actually true,' which is where first-principles thinking usually breaks down in practice.

The prompt

You are reasoning from first principles in the spirit of physics: what do we actually know is true, and what have we just inherited as convention?

<context>
Problem or domain: {{problem}}
Current standard approach: {{standard_approach}}
Why the user is questioning it: {{trigger}}
</context>

<task>
Step 1 — List the assumptions.
Enumerate every assumption baked into the standard approach. Aim for 10 to 15. Include the obvious ones, especially those.

Step 2 — Classify each assumption.
For each, label it as:
- PHYSICAL (constrained by physics, math, or biology — cannot be changed)
- ECONOMIC (true given current cost structures — could change if costs change)
- REGULATORY (true given current law — could change)
- CONVENTIONAL (true because everyone does it this way — no real reason)
- PSYCHOLOGICAL (true because of how humans behave at scale)

Step 3 — Identify the atoms.
After removing all CONVENTIONAL assumptions, what is left? State the irreducible truths of the problem in 3 to 6 bullets.

Step 4 — Rebuild.
Starting only from the atoms, propose 2 to 3 designs that solve the original problem. At least one design must look unfamiliar relative to {{standard_approach}}.
</task>

<output_format>
## Assumptions (with classifications)
## Atoms (irreducible truths)
## Rebuilt designs
## Which atom, if it changed, would invalidate every design above?
</output_format>

<constraints>
If you find yourself classifying every assumption as PHYSICAL, you are anchoring on the standard approach. Re-examine.
If you cannot identify any CONVENTIONAL assumptions, push harder — the goal of this exercise is to find them.
Do not invent technical claims. If you are not sure whether something is physical or conventional, mark it UNCERTAIN and flag it.
</constraints>

Variables

{{{{problem}}}}Replace with your {{problem}}
{{{{standard_approach}}}}Replace with your {{standard approach}}
{{{{trigger}}}}Replace with your {{trigger}}

Example

Input

problem: How should our 40-person SaaS company structure its weekly all-hands?
standard_approach: 60-minute Monday all-hands with CEO update, department updates, then Q&A
trigger: Engagement is dropping; the same 4 people ask all the questions; updates feel performative

Output

## Assumptions
1. The all-hands needs to be live (CONVENTIONAL)
2. It needs to be weekly (CONVENTIONAL)
3. Every department should report (CONVENTIONAL)
4. The CEO opens (CONVENTIONAL)
5. Q&A surfaces real questions (PSYCHOLOGICAL — false at scale; status dynamics suppress real questions)
6. People want context on the company (PSYCHOLOGICAL — true)
7. People want to feel heard (PSYCHOLOGICAL — true)
8. Synchronous attention is the most efficient way to share information (CONVENTIONAL — usually false vs. async)
9. 60 minutes is the right length (CONVENTIONAL)
...

## Atoms
- Employees need shared context on direction and priorities to make good local decisions
- People want to feel that leadership knows what is happening on the ground
- Trust is built by leaders engaging with hard questions, not soft ones
- Synchronous time is the team's most expensive shared resource

## Rebuilt designs
1. Weekly async written CEO note + monthly 30-minute live AMA with anonymous-question priority queue
2. Bi-weekly 25-minute live, format fixed: 5-min priority update, 20-min Q&A only on pre-submitted questions ranked by upvote
3. Monthly 90-minute "show the work" all-hands where two teams demo, no department status updates

## Critical atom
If "trust is built by engaging hard questions" is wrong (e.g., team prefers polished comms), every design above is wrong.

Tips for best results

1The classification step is what separates this from the typical 'first principles' prompt — without forcing a label per assumption, models just rephrase the standard approach in different words.
2Watch the CONVENTIONAL bucket. If it has fewer than 3 entries, Claude is anchoring; push back and ask 'what would a smart person from a different industry change first?'
3Step 4 is where the real value is. If the rebuilt designs all look like the standard approach, the atoms are wrong — go back to step 3.
4Pair this with claude-pre-mortem on the rebuilt design. First principles gets you novel; pre-mortem keeps you alive.

Related prompts

Claude pre-mortem on a planned project or decision

intermediate

Run a structured pre-mortem on a plan you are about to commit to. Surface failure modes, weight likelihood and impact, then propose specific mitigations.

Personal Productivityframeworkmethodologydecision-making

MECE breakdown of a problem space

intermediate

Decompose a problem into Mutually Exclusive, Collectively Exhaustive buckets so analysis cannot double-count or quietly skip a branch.

Personal Productivityframeworkmethodologydecomposition

Claude tree of thoughts — branch, evaluate, prune

advanced

Explore a decision by branching into multiple reasoning paths, scoring each branch on the same criteria, then committing to the strongest one with explicit trade-offs.

Personal Productivityframeworkmethodologyreasoning

Need help implementing this prompt in your workflow?

Book a call

Use case

You are reasoning from first principles in the spirit of physics: what do we actually know is true, and what have we just inherited as convention? <context> Problem or domain: {{problem}} Current standard approach: {{standard_approach}} Why the user is questioning it: {{trigger}} </context> <task> Step 1 — List the assumptions. Enumerate every assumption baked into the standard approach. Aim for 10 to 15. Include the obvious ones, especially those. Step 2 — Classify each assumption. For each, label it as: - PHYSICAL (constrained by physics, math, or biology — cannot be changed) - ECONOMIC (true given current cost structures — could change if costs change) - REGULATORY (true given current law — could change) - CONVENTIONAL (true because everyone does it this way — no real reason) - PSYCHOLOGICAL (true because of how humans behave at scale) Step 3 — Identify the atoms. After removing all CONVENTIONAL assumptions, what is left? State the irreducible truths of the problem in 3 to 6 bullets. Step 4 — Rebuild. Starting only from the atoms, propose 2 to 3 designs that solve the original problem. At least one design must look unfamiliar relative to {{standard_approach}}. </task> <output_format> ## Assumptions (with classifications) ## Atoms (irreducible truths) ## Rebuilt designs ## Which atom, if it changed, would invalidate every design above? </output_format> <constraints> If you find yourself classifying every assumption as PHYSICAL, you are anchoring on the standard approach. Re-examine. If you cannot identify any CONVENTIONAL assumptions, push harder — the goal of this exercise is to find them. Do not invent technical claims. If you are not sure whether something is physical or conventional, mark it UNCERTAIN and flag it. </constraints>

Example

Input

problem: How should our 40-person SaaS company structure its weekly all-hands?
standard_approach: 60-minute Monday all-hands with CEO update, department updates, then Q&A
trigger: Engagement is dropping; the same 4 people ask all the questions; updates feel performative

Output

## Assumptions
1. The all-hands needs to be live (CONVENTIONAL)
2. It needs to be weekly (CONVENTIONAL)
3. Every department should report (CONVENTIONAL)
4. The CEO opens (CONVENTIONAL)
5. Q&A surfaces real questions (PSYCHOLOGICAL — false at scale; status dynamics suppress real questions)
6. People want context on the company (PSYCHOLOGICAL — true)
7. People want to feel heard (PSYCHOLOGICAL — true)
8. Synchronous attention is the most efficient way to share information (CONVENTIONAL — usually false vs. async)
9. 60 minutes is the right length (CONVENTIONAL)
...

## Atoms
- Employees need shared context on direction and priorities to make good local decisions
- People want to feel that leadership knows what is happening on the ground
- Trust is built by leaders engaging with hard questions, not soft ones
- Synchronous time is the team's most expensive shared resource

## Rebuilt designs
1. Weekly async written CEO note + monthly 30-minute live AMA with anonymous-question priority queue
2. Bi-weekly 25-minute live, format fixed: 5-min priority update, 20-min Q&A only on pre-submitted questions ranked by upvote
3. Monthly 90-minute "show the work" all-hands where two teams demo, no department status updates

## Critical atom
If "trust is built by engaging hard questions" is wrong (e.g., team prefers polished comms), every design above is wrong.

Tips for best results

1The classification step is what separates this from the typical 'first principles' prompt — without forcing a label per assumption, models just rephrase the standard approach in different words.

2Watch the CONVENTIONAL bucket. If it has fewer than 3 entries, Claude is anchoring; push back and ask 'what would a smart person from a different industry change first?'

3Step 4 is where the real value is. If the rebuilt designs all look like the standard approach, the atoms are wrong — go back to step 3.

4Pair this with claude-pre-mortem on the rebuilt design. First principles gets you novel; pre-mortem keeps you alive.