First-principles thinking

What this subject is

First-principles thinking is reasoning that decomposes a problem to propositions taken as basic — physical laws, material prices, definitions — and rebuilds conclusions from there, instead of adjusting from precedent or analogy. The idea runs from Aristotle's archai and the axiomatic ideal of demonstration, through Descartes' rules of decomposition, into three working crafts: Fermi estimation, constraint analysis against physical limits, and engineering economics built on material cost floors. Its popular revival is tied to Elon Musk's rocket and battery cost decompositions. The mature view, and this package's spine: derivation and analogy are complementary instruments with distinct failure modes, not ranked rivals.

What the package holds

Curated scaffolding your agent loads before it researches, so it starts from vetted ground rather than a cold search.

• Tier 1: 18
• Tier 2: 6
• Tier 3: 8
• Tier 4: 1

The questions and claims below are quoted from the package files.

Where the field disagrees

Each one carries real proponents on more than one side, so your agent cannot quietly pick a winner.

• Is reasoning from first principles superior to reasoning by analogy?

  0 named positions · The dichotomy is largely false. Expert practice alternates derivation and structurally mapped analogy; teach the two as instruments with different failure modes.

• Do the popular Musk narratives attribute to one person's derivation what institutions, teams, suppliers and markets produced?

  0 named positions · Contested historiography. Teach the documented decompositions as real and useful while presenting causal attribution as an open, structurally confounded question.

• Does decomposing a unique case improve forecasts or degrade them?

  0 named positions · Empirically settled in reference-class-rich domains: outside view first. Decomposition earns its keep where no reference class exists — genuinely novel systems — which is exactly where the lens's engineering examples live.

• Do physics cost floors or experience curves predict technology costs?

  0 named positions · Complementary, not rival: floors bound what is possible, curves describe when it arrives. Treating a floor as a schedule is the documented error (ÆON inference from the sources above).

• Can social and organisational systems be redesigned from first principles the way artifacts can?

  0 named positions · Domain-dependent. Derivation compounds where constraints are physical, stable and cheap to test; it degrades where constraints are social, tacit and expensive to test (ÆON inference consistent with the sources above).

Myths the package corrects

Widely held claims with the evidence that settles or bounds them.

• “First-principles thinking is a Silicon Valley invention, popularised into existence by Elon Musk.”

  debunked

  Aristotle defined a first principle as the first basis from which a thing is known and made demonstration from primary, immediate premises the form of scientific knowledge. Euclid built the working archetype; Descartes turned it into an explicit method of decomposition; physics and chemistry have run ab initio (first-principles) calculations for about a century. The lens revived the phrase; it did not coin the concept.

• “Thinking from first principles means distrusting all existing knowledge and deriving everything yourself from scratch.”

  debunked

  The method decomposes to verified fundamentals — physical laws, exchange prices, definitions — every one of them inherited from centuries of collective work. Musk's own worked examples relied on published metal exchange prices and known engineering relations (rose-2012). Aristotle's point cuts deeper: first principles are reached by induction from existing knowledge, not conjured before it; and much operative knowledge is tacit, so a zero-prior rebuild silently discards what cannot be enumerated (polanyi-1966, chesterton-1929).

• “Reasoning by analogy is lazy, second-rate thinking — serious minds derive instead.”

  debunked-as-general-rule

  Cognitive science finds analogy to be core inferential machinery: structure-mapping is rigorous inference over shared relational structure (gentner-1983), working scientists lean on near-domain analogies precisely when problems get hard (dunbar-1995), and on the strong view categorisation itself is analogy-making (hofstadter-sander-2013). Kepler, Maxwell and Darwin built canonical science on explicit analogies. The genuine failure mode is surface mapping — copying appearances instead of structure — not analogy as such.

• “Fermi estimates land within an order of magnitude because overestimates and underestimates always cancel out.”

  boundary-correction

  Cancellation is probabilistic, not guaranteed: it operates only when factor errors are roughly independent and unbiased in log space — hence the craft of geometric means and plausible ranges (mahajan-2014, weinstein-adam-2008). Correlated assumptions — a wish pushing every factor the same direction — compound instead of cancelling, which is exactly the planning-fallacy signature documented for decomposed forecasts (kahneman-tversky-1979).

• “A physics or materials cost floor tells you what the product will soon cost: batteries decomposed to 80 dollars per kWh of materials meant cheap packs were imminent.”

  boundary-correction

  The floor states what physics and markets permit, not when engineering arrives. Battery pack costs fell about 14 percent per year across the entire industry (nykvist-nilsson-2015), and three decades of lithium-ion data fit an experience curve near 20 percent learning per doubling of cumulative production (ziegler-trancik-2021) — the closing of the gap was a slow, industry-wide learning process of the kind Wright described in 1936, not a step change following a derivation.