What first-principles thinking actually is
TL;DR
A first principle is something that must be true on its own terms. Not because an authority said so, not because everyone does it that way, but because it can't be reduced any further without leaving reality. First-principles thinking is the discipline of deconstructing a problem down to those bedrock truths and reasoning back up from them, instead of reasoning by analogy (copying what already exists with small tweaks). Analogy is fast and usually right. First principles is slow and occasionally revolutionary. The skill is knowing which one a situation deserves.
🎯 For the builder
Why it matters — Almost every default you inherit — a price, an architecture, a process, "the way it's done" — is an analogy to someone else's situation, not a law of nature. If you can't tell which constraints are real and which are merely conventional, you can't tell which ones you're allowed to break.
What it changes in your decisions — Stop asking "what do competitors do?" as the first question. Start asking "what is actually true here, and what does that permit?" That's the only path to a non-consensus, non-obvious answer.
Ask yourself — "Is this constraint a law of physics, or just a habit nobody has re-derived?"
Risk if ignored — You inherit other people's ceilings. Every conclusion you reach is a slightly-worse copy of the conclusion you copied it from.
Two ways to reason
Analogy · fast & usually right. First principles · slow & sometimes revolutionary.
Civilization runs on analogy — but analogy's ceiling is the consensus answer.
Copy what exists · tweak
Pattern-match to a solved thing
Deconstruct to bedrock truths
Refuse the inherited package
Almost all everyday reasoning is reasoning by analogy: this new thing is like that known thing, so treat it the same way. It's a brilliant shortcut. You don't re-derive how a door works at every doorway. You pattern-match "door → handle → push/pull." Civilization runs on analogy, and most of the time you should too. Re-deriving everything from scratch is paralyzing and arrogant.
Reasoning from first principles is the opposite move. You refuse to accept the inherited package and instead break the problem into parts you can independently verify, then rebuild. Aristotle called a first principle "the first basis from which a thing is known." The modern shorthand is owed to Elon Musk's often-quoted framing:
"I think it's important to reason from first principles rather than by analogy. The normal way we conduct our lives is we reason by analogy… [with first principles] you boil things down to the most fundamental truths and then reason up from there."
The classic worked example is battery cost. The analogy argument says: battery packs cost ~$600/kWh and always have, so they always will. The first-principles argument asks: what is a battery actually made of? Cobalt, nickel, aluminum, carbon, some polymers, a steel can. Price those raw materials on the commodity market and the floor is a small fraction of $600/kWh. The gap between the material floor and the market price is not a law of nature. It's an opportunity, but only visible to someone who decomposed the thing instead of quoting its current price.
What counts as a "fundamental"?
A first principle is a claim you can't dissolve further by asking "why is that true?" without bottoming out in:
- Physical law or arithmetic — conservation of energy, the speed of light, the cost of raw materials, the number of seconds in a day. Non-negotiable.
- A definition or goal you actually chose — "we are trying to reduce time-to-value," "the user's real job is X." These are first principles for your problem once you commit to them.
- Directly observed fact — what users actually do (not what they say), the measured latency, the real failure rate.
Everything else — best practices, benchmarks, "industry standard," the existing architecture, the price something currently sells for — is a derived claim. It was true for some context that may not be yours. The work of first-principles thinking is sorting the bedrock from the borrowed.
Why it's rare (and therefore valuable)
If first principles produces better answers, why doesn't everyone do it all the time?
- It's expensive. Decomposition is slow and effortful, while analogy is nearly free. Under time pressure, the brain defaults to the cheap operation. (See learning how to learn on why effort feels like failure.)
- It's socially risky. Re-deriving a "settled" question looks like arrogance or naïveté until you're proven right. Most people would rather be conventionally wrong than unconventionally embarrassed.
- Assumptions are invisible. You can't question a constraint you don't notice you're obeying. Inherited assumptions feel like reality, not like choices. The whole next lesson, the method, is machinery for surfacing them.
That cost is exactly why first-principles reasoning is a source of edge. The answer everyone copies is, by definition, available to everyone. A genuinely better answer almost always requires going back to something more fundamental than the thing being copied.
When NOT to use it
First-principles thinking is a power tool, not a personality. Reaching for it everywhere is its own failure mode (the subject of traps & limits):
- For solved, low-stakes problems, analogy wins. Don't re-derive how to format a date.
- When the cost of being wrong is low and the cost of analysis is high, ship the conventional answer and move on.
- When you lack the fundamentals to decompose honestly, "first principles" becomes a license to ignore hard-won expertise — confident reinvention of a worse wheel.
The decision itself is a tradeoff. That's why this module sits next to Module 06 · Strategy & Tradeoffs: how much to decompose is a cost/benefit call, not a virtue.
Analogy vs. first principles, side by side
| Reasoning by analogy | Reasoning from first principles | |
|---|---|---|
| Speed | Fast, cheap | Slow, effortful |
| Source of the answer | Existing examples | Fundamental truths + logic |
| Best for | Solved, common, low-stakes | Novel, high-stakes, stuck |
| Typical result | A safe variation on the norm | A possibly non-obvious answer |
| Failure mode | Inherits hidden, wrong constraints | Reinvents a solved wheel, slowly |
📦 Mini-case — the battery pack. The famous version: electric-car batteries "cost $600/kWh, always have." Deconstructed to the commodity prices of the raw materials — nickel, cobalt, aluminum, steel — the floor was closer to $80/kWh. Everything above it was manufacturing convention, not physics. The consensus number was an inherited constraint. Reasoning by analogy said "batteries are expensive." Reasoning from first principles said "batteries are expensive the way we currently make them" — a different claim with a different roadmap.
Failure modes
- Analogy in disguise — You think you decomposed, but you stopped at someone else's conclusion and called it a fundamental ("the API has to be REST"). Pseudo-decomposition.
- Bottomless decomposition — You refuse to accept any premise, so you never get to act. First principles needs a floor; pick one and build.
- Ignoring real expertise — Decades of accumulated knowledge get treated as "just convention" to be cleared away. Sometimes the convention is the hard-won first principle.
Practitioner checklist
- For the claim in front of you, can you name why it's true — and does the "why" bottom out in physics, arithmetic, a chosen goal, or observed fact?
- Have you separated the real constraints from the conventional ones in this specific problem?
- Is this a problem that actually deserves first-principles effort, or should analogy handle it?
- If you're decomposing, have you committed to a floor — premises you'll treat as bedrock — so you can reason back up?