Musk Wiki

First principles

NextFree-speech absolutism

First principles

Most people treat decarbonizing the planet as a question of values or political will. Master Plan Part 3 (2023) treats it as an engineering problem with a number attached. State the whole energy system in physical units, then argue from those units that a fully sustainable version is not only possible but cheaper. The 2006 and 2016 plans asserted a mission. This one tries to prove the mission is achievable, and hands the public the math to check.

The document is signed “The Tesla Team,” not Elon Musk, so the wording belongs to the company rather than to him personally. But the move is the institutional expression of the first-principles habit Musk is known for: quantify the whole system from the bottom up, then act on whatever the arithmetic says.

The shape of the argument

  • State the total problem in physical units: ~180 PWh/year of demand, ~30 TW of generation, ~240 TWh of storage, ~$10 trillion of investment.
  • Compare against the existing system and argue the sustainable path needs less material extraction, not more — a deliberately counter-intuitive claim built from the numbers.
  • Publish the assumptions and invite scrutiny instead of asserting authority.

Evidence

The thesis as a feasibility claim:

“The overarching goal of this plan is to demonstrate that a fully sustainable global energy system is achievable and would require less investment and material extraction than the current unsustainable energy system.” ↗

Showing the work and inviting challenge:

“This paper outlines the assumptions, sources and calculations behind that proposal. Input and conversation are welcome.” ↗

Reframing the conclusion from moral necessity to economic logic:

“The transition to a sustainable energy economy is not only necessary but also economically advantageous.” ↗

Only physics is a real rule

Walter Isaacson’s 2023 biography catches the same habit in a more personal, almost combative key. Isaacson has Elon Musk working from one principle: the only rules that genuinely bind are the laws of physics, and everything else is a recommendation, open to challenge rather than obeyed by default. (The exact phrasing Isaacson reports surfaces publicly only on quote-aggregator sites, so the biography is paraphrased rather than block-quoted; see Isaacson biography (2023).)

The byte-verifiable, first-person version comes from the 2023 Lex Fridman conversation:

“Like physics is the law, everything else is a recommendation. I’ve seen plenty of people break the laws made by man, but none break the laws made by physics.” ↗

In the same exchange he roots it in method: a conclusion has to be tested against reality, which he treats as the final arbiter. That is the temperamental twin of what Part 3 does on paper. The plan rebuilds a problem from physical units; this credo says treat any inherited rule as something to question, not to follow on trust. It is also the ground his truth-seeking philosophy stands on, and his test for whether an AI can be trusted.

Two weeks after #400, at the November 2023 DealBook Summit, Sorkin asks whether being told he’s wrong has become a red flag for him. Musk reaches for the same credo and names physics as the check that never blinks:

“Physics is unforgiving. Physics is unforgiving. I have these various little sayings that I’ve come up with, that physics is the law and everything else is a recommendation.” ↗

“In the sense that you can break any law made by humans, but try breaking a law made by physics.” ↗

What DealBook adds is the epistemic payoff he draws straight from it. Physical reality, not opinion, decides whether he’s right, because being wrong has consequences you can watch happen:

“So if you are wrong and persist in being wrong, the rockets will blow up and the cars will fail.” ↗

It’s the same credo as the 2021 “never met anyone who could break physics” and the #400 versions, turned here into an argument about confidence. He trusts his own judgment over a chorus of doubters not out of ego but because the arbiter has kept settling the question in his favor, and would punish him with blown-up rockets if it did not.

Physics as the test for what is real (2019)

In the 2019 Lex Fridman conversation the same instinct becomes an epistemology, a rule for deciding what counts as real. Asked whether we could build an AI we could love, Musk skips the sentimental answer and reasons from physics. Reality is whatever survives every test you can throw at it. So if no test tells a thing apart from the real thing, the difference doesn’t exist.

“from a physics standpoint, essentially, if it loves you in a way that you can’t tell whether it’s real or not, it is real.” ↗

“if there’s no test that you can apply that would make it, allow you to tell the difference, then there is no difference.” ↗

This is the physics-is-the-law credo in test-first form. Not “physics constrains what you can build” but “an empirical test is the only arbiter of what is real.” He turns the same rule on the simulation on the spot: if no test separates base reality from a simulation, then “from a physics perspective, it might as well be the same thing” (more on Simulation hypothesis). It’s the earliest first-person statement in the wiki of the verificationist streak under both his metaphysics and his demand that an AI be judged against reality, not authority.

The follow-up #49 conversation (November 2019, Tier-1 verified) names the principle behind the move outright. Asked whether consciousness permeates all matter, he answers from method rather than intuition (I believe in scientific method) and reasons that something is true only to the degree you can test it, otherwise you’re “just talking about preferences or untestable beliefs.” That’s the explicit version of the test-first rule #18 applies without spelling it out (block-quoted on Consciousness and death and the source page).

The algorithm, in his own words

Scope note: the fixed five-step building method below (question requirements → delete → simplify → accelerate → automate, “the best part is no part,” “the factory is the product”) now has its own page, The engineering algorithm, whose primary source is the 2021 Everyday Astronaut Starbase tour — its fullest public statement. This section keeps the #438 “mantra” version, because it’s the point where first-principles reasoning turns into a procedure. The algorithm sits on top of the reasoning habit described here.

The most concrete first-person version of that habit is in the 2024 Lex Fridman conversation (#438): a fixed, ordered procedure he says he runs as a mantra. The first step, and the one he leans on hardest, is to attack the question before the solution.

“I have this very basic first principles algorithm that I run kind of as a mantra, which is to first question the requirements, make the requirements less dumb.” ↗

The rationale is pure first-principles: however smart the person who wrote the requirements, they’re still dumb to some degree, and if you skip this you risk solving the wrong problem perfectly (paraphrased). Step two is to delete the part or process outright before improving it, with a number attached: if you aren’t later forced to add back at least 10% of what you cut, you didn’t cut enough. That ordering produces his sharpest line on engineering:

“the most common mistake of smart engineers is to optimize a thing that should not exist” ↗

Only after question-then-delete do the rest follow: optimize and simplify, then accelerate, then automate. He insists on that order because accelerating something that shouldn’t exist is, in his word, absurd (paraphrased). He even diagnoses why the delete step is the hard one through his limbic–cortex model. People overcorrect from the remembered sting of a past deletion, so holding the discipline is a deliberate cortical override to a limbic instinct.

Reason up from the axiomatic base (Lex Fridman #252, 2021)

The 2021 Lex Fridman conversation (#252) gives the method its plainest spoken form — the same move Master Plan 3 makes on paper, now described as a general procedure for “really any walk of life”:

“let’s boil something down to the most fundamental principles, the things that we are most confident are true at a foundational level, and that sets your axiomatic base, and then you reason up from there.” ↗

The constraint underneath it all is physics, the one rule that won’t bend — the 2021 echo of the “physics is the law, everything else is a recommendation” credo:

“I’ve met a lot of people that can break the law, but I have never met anyone who could break physics.” ↗

Then he walks through two physics tools he says he uses constantly. The first is thinking in the limit: scale a thing to an extreme (what if our volume were a million units a year?) and if it’s still expensive, the cause isn’t volume, it’s the design. The second he calls the “platonic ideal” or “magic wand” move. Price a product against the raw-material value of its atoms, the lowest cost there can be, then picture the perfect arrangement of those atoms and work backward to how you’d get them into that shape. The point is to start from the physics rather than from the tools and parts you already happen to know. Both are paraphrased here; each runs across a long, winding exchange. What stands out in the 2021 version is how plainly it states the axiom-first core, boil down then reason up, more plainly than the 2024 question-delete algorithm does.

Money is a proxy for effort (Battery Day 2020)

Battery Day 2020 has a small, characteristic move on the question of scaling battery production. Pressed on it as a problem of cost, Musk refuses the framing and reduces money to what it really stands for — people and machines doing work:

“Yeah, and I think we should view this as more than just a question of money. Money is sort of an ethereal thing, but it’s really the amount of effort.” ↗

You can’t simply spend your way to terawatt-hour scale (“it’s not that easy”). The binding constraint is organized effort, so the real optimization is making a fixed amount of effort yield the most batteries. It’s the 2021 “axiomatic base” move aimed at economics: strip the conventional unit, dollars, down to the physical thing under it, effort, and optimize that.

The same event restates the discipline behind the method. Closing the cathode-materials Q&A, a teammate paraphrases his standing rule that every existing spec and method is wrong; Musk agrees and compresses it to a sentence:

“Exactly. We’re wrong, just the question of how wrong. Trying to be less wrong.” ↗

It’s the 2020 version of what he tells the 2017 World Government Summit: “always take the position that you are to some degree wrong, and your goal is to be less wrong over time”. Three years on, pointed at engineering specs. Every belief is provisional, and the only goal is to be less wrong than before.

The earliest contrarian read — 1995, and space economics (Stanford eCorner, 2003)

The 2003 Stanford eCorner talk is the earliest dated source in this collection, and the first-principles instinct is already running a decade before he had a name for it. Not yet the polished “framework for thinking” of TED2013, but a contrarian, read-the-situation take on two problems.

The first is the 1995 internet bet. His proof that it was the right call is that the people paid to spot the next big thing had missed it — to him that was the opportunity, not a warning:

“In fact, most of the venture capitalists that I talked to hadn’t even heard of the internet which sounds bizarre on Sand Hill Road.” ↗

From the situation as it actually stood, no internet company worth joining, he reasons to the only move left instead of inheriting a path by analogy:

“The only way to get involved in the internet in '95, that I could think of, was to start a company.” ↗

And he derives the business model from one hard constraint, the need for fast revenue, rather than from what other companies were doing (the transcript’s “I though” is a preserved 2003 typo):

“So I though we got to make something that’s going to return money very, very quickly.” ↗

The second problem is space, and it’s the cleaner specimen. In the Space Mining vs. Human Space Flight clip, at the very start of SpaceX, he rejects two then-fashionable space ideas on no engineering grounds at all. He simply won’t let a project exist unless the numbers close:

“the economics don’t make sense. They just can’t close the economic case.” ↗

“Closing the economic case” is the same bottom-up move the 2023 Master Plan Part 3 makes for energy and the 2013 reusable-rocket argument makes for launch cost, here used twenty years earlier to decide what not to chase. The constructive flip side, where the economics do close (a self-sustaining off-Earth civilization), is on Mars colonization.

Knowledge as a “semantic tree” — fundamentals before details (Reddit AMA, 2015)

The January 2015 Reddit AMA catches the habit in a form it rarely takes: not a reasoning method (TED2013) or a building algorithm (#438), but a rule for how to learn. Asked how he acquires knowledge, he reaches for the metaphor he’s now best known for on the subject:

“View knowledge as sort of a semantic tree – make sure you understand the fundamental principles before diving into details.” ↗

It’s the same instinct as the 2013 TED “boil things down to their fundamental truths and reason up from there”, aimed at acquiring knowledge rather than designing things. Knowledge has a structure (trunk → branches → leaves), and detail only sticks if it has a secured foundation to hang on. Written down and self-paced, it packs the whole habit into one sentence — eighteen months before the 2016 material and six years before the 2021 “axiomatic base” restatement. It’s the earliest “how to acquire knowledge” statement in the wiki, and a clean sign that by early 2015 the method was, for him, already a conscious pedagogy.

The canonical public formulation (TED2013)

The 2013 TED conversation is the earliest long-form interview in this collection, and it carries the line most often quoted as Musk’s definition of the method. He says it in his own voice, eight years before the 2021 “axiomatic base” version and more than a decade before the 2024 “algorithm”. Asked at the close how one person innovates across cars, rockets and solar, he names the framework outright:

“Well, I do think there’s a good framework for thinking. It is physics. You know, the sort of first principles reasoning.” ↗

Then the canonical phrasing of the method itself:

“boil things down to their fundamental truths and reason up from there, as opposed to reasoning by analogy.” ↗

And he scopes when it’s needed. Ordinary life runs on analogy, copying others with slight variations; the physics approach is for doing something new:

“But when you want to do something new, you have to apply the physics approach.” ↗

In the same breath he adds a second piece of method that almost none of the later sources repeat. He doesn’t just tolerate criticism, he goes after it:

“and then also to really pay attention to negative feedback, and solicit it, particularly from friends.” ↗

Here, in its plainest and earliest public form, is what the 2021 “reason up from the axiomatic base” statement and the 2024 question-then-delete algorithm would later formalize. The “solicit negative feedback” note is the complement to all of it: the truth-test (physics is the law) needs an input, and he says he hunts down the disconfirming kind. It runs on the same reasoning as his other TED2013 moves: solar as “indirect fusion”, reusability argued from every other mode of transport, each a problem rebuilt from a physical base instead of inherited by analogy.

“Physics is the best framework,” and “to some degree wrong” (World Government Summit 2017)

Two months before TED2017, at the February 2017 World Government Summit, he gives the same habit as advice to the young, and pairs the physics framework with the discipline that has to sit under it. The framework, named outright:

“I’d recommend studying the thinking process around physics.” ↗

“the way of thinking in physics, it’s the best framework” ↗

The discipline he says it teaches: hold your own beliefs as provisional, and aim only to be less wrong over time.

“always take the position that you are to some degree wrong, and your goal is to be less wrong over time.” ↗

The failure mode he names is wishful thinking, ignoring the truth in favor of what you’d like to be true:

“One of the biggest mistakes people generally make and I’m guilty of it too is wishful thinking.” ↗

“You ignore the real truth, because of what you want to be true.” ↗

And the corrective-feedback rule, almost word for word the 2013 TED “solicit negative feedback, particularly from friends,” four years on:

“And solicit critical feedback, particularly from friends.” ↗

So it’s the 2013 physics-plus-solicit-criticism pairing again, with the “less wrong over time” and anti-wishful-thinking discipline now spelled out. This is the temperament behind the “physics is the law” credo, offered as the thing he most wants the young to learn. (Earlier in the same interview he names keeping a “corrective feedback loop” alive over time, “even when people won’t … tell you exactly what you want to hear”, as one of his biggest challenges.)

The future as a branching stream of probabilities (TED2017)

The 2017 TED conversation is where he’s clearest about the stance under the whole habit: how he relates to the future at all. He doesn’t treat it as a fixed destiny to be predicted, but as a probability distribution that deliberate action can shift:

“I look at the future from the standpoint of probabilities. It’s like a branching stream of probabilities, and there are actions that we can take that affect those probabilities or that accelerate one thing or slow down another thing.” ↗

That’s what makes his missions hang together. The energy transition is a high-probability branch, so Tesla’s job is just to accelerate it; multi-planetary life is a low-probability one that needs a hard push, “definitely not inevitable.”

The same talk restates the entropy-and-decay argument the 2016 Y Combinator talk grounds in thermodynamics: progress isn’t automatic, and the default is decline.

“People are mistaken when they think that technology just automatically improves. It does not automatically improve.” ↗

“It only improves if a lot of people work very hard to make it better, and actually it will, I think, by itself degrade, actually.” ↗

He reaches for the same historical evidence as 2016, civilizations that lost what they once knew how to do:

“You look at great civilizations like Ancient Egypt, and they were able to make the pyramids, and they forgot how to do that.” ↗

The Egypt and Rome examples and the “technology does not automatically improve” claim come back almost verbatim a year apart, 2016 to 2017. That recurrence marks this as a load-bearing piece of his worldview rather than a one-off line: physical law sets the default, decline, and progress is the exception you have to work for. (In the same talk, the tunneling cost breakdown is the constructive face of the same habit: halve the diameter, since cost scales with cross-sectional area.)

Entropy is not on your side (2016)

The 2016 Y Combinator conversation gives the physical principle under the whole habit, years before the master-plan and Lex versions. His argument is that technological progress isn’t automatic; it’s a fight against decay. People assume technology improves on its own. He insists on the opposite:

“It only gets better if smart people work like crazy to make it better.” ↗

He grounds it in history, Egypt forgetting how to build pyramids, Rome forgetting aqueducts and plumbing, and pulls the rule from thermodynamics itself:

“So I think, sure, let’s bear in mind that entropy is not on your side.” ↗

It’s the later physics-is-the-law credo aimed at civilizations instead of rockets. Physical law, here the second law, is the binding constraint, and decline is the default unless someone works against it. It’s also the urgency engine under the missions, his reason that progress has to be pushed rather than waited for, and a close cousin of the great-filter survival argument.

The feasibility seen by few, and the energy-accounting correction (Investor Day 2023)

Opening the Master Plan 3 presentation at Investor Day 2023 (March 2023), Musk gives a clean case of the gap between first-principles and received wisdom. A fully sustainable Earth is, to him, demonstrably reachable, yet even capable people don’t see it:

“most of the smart people I know actually don’t see this clear path” ↗

That’s the method’s contrarian signature: a conclusion that looks obvious once you do the calculation from fundamentals but stays invisible to anyone reasoning by analogy or consensus. The correction he leans on is an accounting one. People assume an electrified civilization needs the same primary energy as a combustion one. It doesn’t, because combustion throws most of its energy away as heat:

“electrified civilization versus a combustion civilization this is not true” ↗

Strip the problem to physical fundamentals (how much energy actually does useful work) and the apparent scale of the transition collapses, because most “primary energy” in a fossil system was never useful to begin with. (A gasoline car turns under ~20% of the oil’s energy into motion, fully accounted for.) It’s the same mission feasibility the Master Plan 3 white paper quantifies, shown here as a first-principles result that, in his telling, “the smart people” miss precisely because they don’t recompute the baseline.

Reduce the question to the physics (Tesla earnings calls, 2010-2012)

The early Tesla earnings calls carry three of the clearest spoken instances of the method on the record — the same boil-it-down reflex, now aimed at engineering arguments rather than the mission. On the perennial large-cell-vs-18650 debate (Q4 2010), he won’t fight by analogy and collapses the whole question to two physical numbers:

“For those that think it’s that there’s a larger format cell that’s better, I would simply ask, “What is the cost per kilowatt hour, and what is the energy density?” Until some cell supplier comes back to us with a number that is better than the 18650. It’s really as simple as that.” ↗

On crash safety (Q3 2011) he reasons from the physics of deceleration and makes it vivid by analogy, the structural advantage of having no engine block to drive through the passenger:

“it seems like if you jump out of a five-story window, you probably want to jump into an Olympic-sized swimming pool rather than a kiddie pool. Having a long crumple zone means that you can spread out the deceleration of the car over a much longer distance” ↗

And on cost reduction (Q1 2012) he states a rule he treats as true of any technology, the scale-plus-iteration law under the down-market ladder:

“if you scale up production by a factor of 10, your costs will generally drop by half. I mean, that’s a good generalization, I think. … which is continued iteration on the design and economies of scale. That’s generally true for any technology.” ↗

The same instinct gets pointed at an entire industry norm too. Car-buying becomes a problem to redesign from scratch rather than inherit, the earliest seed of the direct-sales philosophy (“most people would rate purchasing a car as their worst retail experience … We want it to be … a delightful experience,” Q1 2011, block-quoted on Tesla Earnings Calls 2010-2012). These are the engineering-domain twins of the energy-accounting and tunneling breakdowns above: reduce, quantify, ignore the consensus framing.

Reason in the limit, and best-case the rival (Tesla earnings calls, 2013-2015)

The 2013–2015 earnings calls are dense with the method, and they add two distinctive moves. The first is reasoning in the limit: take a variable to its extreme to find the optimum. Asked why Tesla uses small commodity cells, he pushes the reject-rate problem to its boundary (Q2 2013):

“So I say in the limit where the battery where the whole battery is 1 cell, the reject rate would be virtually 100%. And then as you make the cell smaller and smaller, the reject rate will reduce.” ↗

The second is to beat a rival technology by granting it its theoretical best case and showing it still loses, so the project is hopeless before it starts. On hydrogen fuel cells (Q2 2014):

“if the best case in our opinion, the best case fuel cell car and obviously the current fuel cell cars are far from best case, cannot beat the current case electric car, well, why even try? That just makes no sense. Success is not one of the possible outcomes.” ↗

He backs it with the physics distinction he repeats for years: “hydrogen is an energy carrier not an energy source.” The same physics-first habit runs all through the era. Cost reduction becomes getting “the molecules in the right shape in a smarter way” rather than stripping value (Q1 2014). The empiricist’s filter is “Don’t send us PowerPoint … Just send us one cell that works” (Q3 2014). There’s a terawatt-hour macro-energy calculation for a fully-electric civilization (Q1 2015), and a stationary-storage buffering argument that, “in principle,” you could “shut down half of the world’s power plants” — decoupled from the renewables story (Q2 2015). These are block-quoted on Tesla Earnings Calls 2013-2015.

The S-curve, moats-are-lame, and reason-by-analogy (Tesla earnings calls, 2016-2018)

The 2016–2018 earnings calls add two distinctive first-principles frameworks. The first is the S-curve production model, born and refined entirely within the era. A new product’s ramp is “extremely difficult” to predict at its “initial portion” (Q1 2017), it can “go backwards because something’s broke” (Q2 2017), and the cognitive-bias kernel: “human intuition tends to be a straight line extrapolation, but we’re really on a very steep exponential” (Q3 2017). It’s the same anti-linear instinct he turns on the skeptics (“which one of you predicted that Tesla would go from 2,500 units delivered to 250,000”; “consider your assumptions… or perhaps pessimistic”).

The second is the famous moats-are-lame, pace-of-innovation belief, competitiveness boiled down to its real variable: “What matters is the pace of innovation. That is the fundamental determinant of competitiveness,” proved by historical analogy (“Amazon vs Walmart… The outcome was obvious a long time ago,” Q1 2018). The era also opens (Q1 2016) with the cleanest statement of the method’s negative half, “it’s always tempting for people to reason by analogy instead of first principles”, and its vivid illustration: you can kill a fly with a thermonuclear weapon or a fly swatter, same result, wildly different difficulty. With it come the cross-domain SpaceX↔Tesla transfer he leans on, and the single-lever insight that merely shortening “the time from factory to the end customer” would let a company “outcompete all other companies.” These are block-quoted on Tesla Earnings Calls 2016-2018.

Derive from physics, not from the market (Tesla earnings calls, 2019-2021)

The 2019–2021 earnings calls show the method across several domains. The cleanest case is the Battery-Day recap, where he names the move outright: instead of benchmarking competitors, “just say from a physics standpoint… what’s the limit of physics? What’s the Platonic ideal of a perfect cell, and how close can we get there?” (Q3 2020). The same derive-from-structure habit produces the vision-only autonomy argument in one line: “Once you solve passive optical, you’ve solved Self-Driving, so why bother with anything else?” (Q3 2020). It also reduces the renewable-energy claim to engineering rather than science: “no new physics is necessary, no new materials necessary. We just need to scale this thing up” (Q1 2021).

Two cognition-level framings recur. The anti-linear instinct comes back as a claim about human wiring: “We didn’t evolve to feel an exponential. We can feel a linear, but we could only understand an exponential at a cognitive level” (Q2 2019). And the cost-vs-value distinction sharpens into a design principle: “any fool can take cost out of a car and make it worse,” the real task is to “make it lighter, and simpler” (Q1 2020). The era also restates his foundational ideas-vs-execution belief: “there’s way too much value placed on the idea… Going to the moon is the hard part by far” (Q4 2021). And it stretches first-principles reasoning into economics to justify Optimus: “the foundation of the economy is labor. Capital equipment is distilled labor” (Q4 2021).

Reason from physics, frame the question right (Tesla earnings, 2022-2026)

The 2022-2026 earnings calls are dense with his signature reasoning moves. There’s the limiting-factor heuristic (“Whatever the limiting factor is, we’ll do. We do not artificially constrain ourselves,” Q3 2022) and the Platonic-ideal design philosophy (“there’s some Platonic ideal of the perfect product, where… you have exactly the right atoms in there in exactly the right position,” Q2 2022). The camera-only conviction comes out as a biological analogy (“humans drive without shooting lasers out of their eyes… humans drive with eyes and a neural net and a brain neural net,” Q4 2024), beside the physics-first-principles design of Optimus (“nothing worked for a humanoid robot at any price. We had to design everything from physics first principles,” Q4 2024). The company-as-organism model shows up (“A company is kind of like a creature growing… 10 cells versus 100 versus 1 million… where humans are around 35 trillion cells,” Q1 2024). And the maxim about framing that anchors the whole AI-company reframe: “If you value Tesla as just an auto company… it’s just the wrong framework. If you ask the wrong question, then the right answer is impossible” (Q1 2024).

Physics-as-arbiter, expected value, and anti-IP (tweets, 2011-2014)

The 2010-2014 tweets carry the reasoning move in several early forms. He states the maxim outright on SpaceX’s tenth anniversary (“10 years ago today, SpaceX was founded. Many battles fought. Physics always won”) and reasons his way to the reusability conviction in its most self-critical form: “All current rocket tech, including ours, sucks. Only when it becomes fully reusable, will it not suck.” The clearest first-principles argument on the page is the climate one. He cuts policy loose from belief entirely (“Belief in climate change isn’t necessary. Even a small probability of a severe outcome justifies a carbon tax”) and reasons to pricing the externality over subsidies (“no need for clean energy subsidies, as externality is priced into mkt behavior”; “Let market decide”). The same instinct shows in his capital-allocation rule (“Paying div is a sign that a company cannot find good ways to spend money”) and his anti-patent, open-source stance, reasoned as progress over IP rents and then acted on: “I really hate patents unless critical to company survival. Will publish Hyperloop as open source” (2013), then “Regarding Tesla patents” (the 2014 patent open-sourcing).

Physics, valuation, and evidence (tweets, 2015-2017)

The 2015-2017 tweets keep the explanatory, evidence-demanding voice running. He reasons from Newton’s third law to debunk a misconception (“In vacuum, there is nothing to “push” against. You must react against ejected mass”), swats down pseudo-archaeology with a sharp epistemic standard (“Stacking stone blocks is not evidence of an advanced civilization”), and frames physics itself in information-theory terms (“Interesting to think of physics as a set of compression algorithms for the universe. That’s basically what formulas are”). The same instinct produces his distilled epistemology: go to the primary source (“Reading the source material is better than reading other people’s opinions about the source material”), put science above any authority including himself (the Atatürk maxim, “If one day, my words are against science, choose science”), and match belief to evidence (“So strange that people often believe things inversely proportionate to the evidence”). It also drives the first-principles valuation argument that past performance is irrelevant (“A stock price represents risk-adjusted future cash flows”) and reusability as the obviously correct norm (“That’s how it is for cars & airplanes and how it should be for rockets”).

The most-quoted aphorisms — physics, manufacturing, money (tweets, 2018-2020)

The 2018-2020 tweets are where the first-principles voice mints its most durable one-liners, repeating them often enough to mark them as core. The signature aphorism shows up three separate times across the window: “Physics is the law, everything else is a recommendation”. Alongside it sit the Rutherford-flavored “All of physics is either impossible or trivial. It is impossible until you understand it, and then it becomes trivial” and the reductive “Run a physics sim long enough & you’ll get intelligence.” The production-difficulty principle becomes a refrain: “Rocket design is relatively easy, making one is hard, making many is extremely hard. Manufacturing is underrated,” “Production is by far the hard part,” and the execution-over-ideas creed reaches its bluntest form in “the idea of going to the moon is trivial, but going to the moon is hard.” Reusability is restated as inevitability (“In the future, it will be as strange to have expendable rockets as it would be to have expendable airplanes today”) and as a moral and economic “tragedy” (the SLS “a billion dollar rocket is blown up every launch”).

The money-as-information-system reframing appears twice in his own notation (“‘Money’ == Series of heterogeneous databases … the value of money–”; “just a (slow, lossy & unsecure) database for labor allocation”). Around it cluster the rest: the reductive definition of a company (“just a group of people gathered together to make products”), the incentive-structure model (“Outcome-based contracting … rewards results & latter rewards waste”; “Incent outcome, not path”), the anti-monopoly economics (“Saying you like ‘moats’ is just a nice way of saying you like oligopolies”), and the epistemic-humility models (“Pace of innovation is all that matters in the long run”; the Dunning-Kruger “we’re almost always dumber than we think we are”).

The 2021-2022 tweets turn the method on money, science, and his own mind. The densest application is the year-long 2021 money-as-information thread: “The thing we call money is just an information system for labor allocation … Whichever has least error & latency will win,” “Money is just data that allows us to avoid the inconvenience of barter … subject to latency & error. The system will evolve to that which minimizes both,” and “Goods & services are the real economy, any form of money is simply the accounting thereof”. That model frames all of his crypto reasoning. He also restates the epistemological bedrock: “Physics is simply the search for truth. Nothing is more rigorous”; “People are able to break any laws made by humans, but none made by physics”. And he gives the clearest 2022 form of his signature method, “the fundamental error is reasoning by analogy, rather than first principles,” paired with the humility maxim “Our view of reality is always wrong, just a question of how wrong.” The same instinct produces his “mental firewall” thesis on ideological self-programming: “Critical & first principles thinking should be a required course in middle school. Who wrote the software running in your head? Are you sure you actually want it there?”. He reprises it year-end as “Run antivirus software in your brain”. It also drives his definition of science as questioning rather than deference: “New Twitter policy is to follow the science, which necessarily includes reasoned questioning of the science”; “Anyone who says that questioning them is questioning science itself cannot be regarded as a scientist”. Characteristic contrarian calls round it out: hydrogen (“Fuel cells should be called fool sells!”), competitive strategy (“Don’t build moats, build tech trees”), and autonomy’s inevitability (“Won’t be long before we view gasoline cars the same way we view steam engines today”).

Intelligence, the “in the limit” heuristics, and execution (tweets, 2023-2026)

The 2023-2026 tweets keep minting durable maxims: intelligence defined as prediction and compression, the recurring “in the limit” heuristics (stupidity→malice, incompetence→fraud), and the execution-over-ideas creed.

“Intelligence seems to be about how tightly you can compress reality and, from that compression, predict the future” ↗

“The best definition of ntelligence is simply the ability to predict the future imo.” ↗

“Stupidity in the limit is indistinguishable from malice” ↗

“The California high-speed rail project is an example of where incompetence in the limit is indistinguishable from fraud” ↗

“With rare exception, ideas really are trivial compared to execution. For example, the idea of going to the Moon is simple, but ACTUALLY going to the Moon is staggeringly difficult.” ↗