Decompose to bedrock truth.
First Principles
Break problems to their irreducible truths, then reason up. Cuts through assumption and analogy to reveal what is actually true.
Most reasoning is analogical — we look at how similar problems were solved and copy the pattern. First principles thinking rejects that. It asks: what do we know for certain, from the ground up, without borrowing from how things have been done before?
The method
Start by identifying what you think is true. Then ruthlessly ask: how do I know this? Is this an assumption inherited from convention, or something I can verify? Strip away layers until you reach bedrock — facts that cannot be further reduced. Then rebuild your solution from those facts alone.
Why it matters
Elon Musk used first principles to question why rocket components cost what they did. The industry said 'rockets are expensive' and everyone accepted it. He asked what the raw materials actually cost, found it was a fraction of the price, and built SpaceX. The assumption was wrong. The reasoning chain built on top of it was therefore wrong too.
The hard part
It's cognitively expensive. Analogical reasoning is fast and usually good enough. First principles is slow and demands you sit with uncertainty. The trick is to reserve it for decisions that are expensive to reverse — architectural choices, hiring, strategy — not everyday calls.
In engineering leadership
Applied to technical decisions: don't ask 'how has this database problem been solved?' — ask 'what are my actual consistency, latency, and throughput requirements, and what properties does a storage system need to satisfy them?' You'll reach different answers than cargo-culting the last team's stack.
In practice
Your team says you need microservices because 'that's how modern systems are built.' First principles: what specific scaling, team autonomy, or deployment problems do you actually have today? If the answer is 'none yet', a distributed monolith is probably wrong. You can always extract services later when the seams become obvious.
TL;DR
Strip every assumption until you hit what's actually true, then rebuild. Expensive cognitively — worth it for irreversible decisions.