First principles

Reason from fundamentals by decomposing a problem to its irreducible level, then stripping the inherited assumptions, so you rebuild from what is actually necessary rather than from convention or analogy.

Flow

graph LR
  nthinkabstractionladdering["Abstraction Laddering"]
  nthinkassumptionreversal["Assumption Reversal"]
  nthinkabstractionladdering --> nthinkassumptionreversal

Each step passes a compressed artifact to the next. The numbered list below is the same chain in text.

The chain

1. think-abstraction-laddering
2. think-assumption-reversal

Job: reason from fundamentals - break a problem or belief down to what is irreducibly true, strip the inherited assumptions and analogies, and rebuild from what is actually necessary.

Use when: a plan or belief is inherited from “how it is usually done”, reasoning is proceeding by analogy or precedent, or you suspect a constraint everyone treats as fixed is actually a convention.

Why a recipe, not a skill: first-principles thinking has no separable mechanism of its own. Its two operative moves are decomposition (ship: abstraction-laddering, or issue-tree for a MECE break) and assumption-stripping (ship: assumption-reversal, or ladder-of-inference-check for inherited inferential leaps). Reasoning back up from the fundamentals is ordinary synthesis. So the honest home for “first principles” is this chain, not a separate catalog entry.

Chain

1. think-abstraction-laddering (skills/think-abstraction-laddering/SKILL.md)
   • Move down the “how / what specifically” axis until you reach the level the problem actually reduces to: the irreducible facts, quantities, or requirements.
   • Carry forward: the irreducible level (the fundamentals), not the whole ladder.
   • Swap in think-issue-tree instead when the break is a MECE decomposition of a question rather than a level shift.
2. think-assumption-reversal (skills/think-assumption-reversal/SKILL.md)
   • Run it on those fundamentals: surface the assumptions baked into the current approach, negate each, and sort which are genuinely necessary from which are inherited convention or analogy.
   • Carry forward: the necessary-vs-convention split.
   • Swap in think-ladder-of-inference-check when the issue is an inherited inferential leap rather than a baked-in premise.

Composite artifact

A fundamentals sheet: the irreducible truths that survive, the conventions stripped away (with why each was only convention), and a reconstruction built up from what is actually necessary, so the next move proceeds from bedrock rather than from precedent.

When NOT to use

When you only need the right altitude on a problem, use think-abstraction-laddering alone. When you only need premises negated to generate options, use think-assumption-reversal alone. First-principles is the chain; do not invoke the ceremony when one move does the work.

Token discipline

Pass only the irreducible level, then the necessary-vs-convention split. Do not re-feed each step’s full output into the next.