Musk Wiki

How Musk Makes Decisions

NextOptimism on a Clock

How Musk Makes Decisions

Four habits, run as one machine. When Elon Musk decides what to build, how to build it, in what order to ship it, and how fast to push, he isn’t pulling one trick. He is running four moves that lock together. He reasons up from physical fundamentals instead of by analogy (First principles). He strips the design that produces down to the leanest thing that still works, by a fixed five-step procedure (The engineering algorithm). He fixes a multi-decade endpoint in public, years before the means to reach it exist, and treats each near-term product as a rung toward it (Secret Master Plan method). And he pushes the whole thing at a deliberately reckless tempo, having already priced in the odds of failure, so a blown-up prototype reads as cheap information rather than defeat (Fear of failure, Work intensity).

What follows traces how the four engage: framing, cutting, sequencing, betting, as four stages of a single loop that hands off from one to the next. Most of the quotes here are Musk himself speaking. The master-plan lines come from Tesla’s published plans rather than off-the-cuff remarks: Musk signed the 2006 and 2016 plans personally, and the 2025 one is bylined “The Tesla Team,” carried on institutionally under his direction.

Summary

Each move answers a different question, and the order they fire in is the procedure:

  1. Frame — what is actually true here, and what is the real question? First principles throws out the inherited version of a problem and rebuilds it from physical fundamentals. Physics is the only binding rule; reality settles every argument.
  2. Cut — what should not exist? The The engineering algorithm takes that framed problem and runs a fixed, ordered procedure: question the requirements, delete, simplify, accelerate, automate. The point is to remove before improving. The best part is no part.
  3. Sequence — in what order do the products reach the goal? The Secret Master Plan method names a multi-decade destination before the means exist, writes it down in public, and casts each near-term product as a deliberate step (the high-end-then-down-market ladder) toward it.
  4. Bet — how hard and how fast do you push, knowing you will often be wrong? Here the risk and speed live (Fear of failure, Work intensity). He runs the first three having already accepted the odds of loss, hating any slack in the line, treating a blown-up prototype as the cheapest possible lesson.

The four are not independent, and that is the whole point. Step 1 of the algorithm, “question the requirements”, is just first-principles reasoning aimed at a spec, so cutting is framing pointed at a design. The master plan is the same boil-it-down-and-reason-up move stretched across decades, so sequencing is framing pointed at a roadmap. And the betting temperament is what makes the other three affordable: accepting the loss in advance is what lets him reframe, delete, and re-sequence without the fear that freezes a normal company. Physics decides what is real, deletion decides what is necessary, the plan decides the order, and pre-accepted loss decides the speed.

ℹ️ One machine, not four tricks. Pulled apart, “first principles,” “the engineering algorithm,” “the master plan,” and “the risk appetite” look like four separate Musk signatures a profile would list as four bullet points. Run together, they are four stages of one loop, each handing off to the next. The reasoning habit hands over a problem stated from fundamentals. The algorithm cuts it to the bone. The plan drops the cut product into a multi-decade order. The betting temperament supplies the nerve and the tempo to ship it before it is safe. What matters here is how those four engage, not the long history of any one of them. The 25-year hardening of the framing move is its own story in From Instinct to Algorithm; in this loop that move is one part of a larger mechanism.

Frame: reason up from physics, not by analogy

The first move in any decision is to throw out the received framing and rebuild the problem from what is physically true. The canonical line is his 2013 TED definition:

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

He is careful about when this expensive mode earns its keep. Ordinary life runs fine on analogy; the physics approach is reserved for doing something new:

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

And one rule settles every framed problem, the one constraint nobody gets to argue with. He put it plainly in 2023:

“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 2021 he framed it as a method for “really any walk of life,” and named the bedrock it rests on the “axiomatic base”: set the foundations you are most sure of, then build up from there.

“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.” ↗

This is the move that decides what is true and what the real question is. What comes out of it, a problem stated from physical fundamentals instead of inherited by convention, is what the next move cuts. How this one habit hardened over 25 years, from a 2003 instinct into a recited mantra, is the story From Instinct to Algorithm tells. Here it is simply the starting move, and the question is how it meshes with the other three.

Cut: delete before you optimize (“the best part is no part”)

With the problem framed, the second move turns it into the leanest design that works, by a fixed and ordered procedure. The first step is just the framing move aimed at the spec: go after the question before the solution, because no requirement is safe. His fullest version is from the 2021 Starbase tour:

“First make your requirements less dumb, your requirements are definitely dumb.” ↗

Only then comes the move he is known for: delete before you improve, and delete hard enough that you sometimes go too far and have to add a part back.

“If you’re not deleting a part or process step, at least 10% of the time, basically if you’re not adding things back in 10% of the time, you’re clearly not deleting enough.” ↗

The order is the whole point, and he says so in the sharpest line in his whole method, a warning against running the later moves first:

“possibly the most common error of a smart engineer is to optimize the thing that should not exist.” ↗

He had already compressed the whole idea into a slogan two years earlier, at the 2019 Starship Update:

“the best part is no part the best process is no process it weighs nothing costs nothing can’t go wrong” ↗

One step here, though, the algorithm cannot decide on its own: what to build toward. At Mars Society 2020 Musk insists the whole design question is empty until the goal is fixed:

“you first have to say what what is the goal” ↗

“once you have what is the goal you can then measure the various designs against that goal” ↗

That is exactly where cutting hands off to sequencing: you can only delete against a goal, and the goal comes from the master plan.

ℹ️ Cutting is framing pointed at a design. Step 1 of the algorithm, “make the requirements less dumb”, is not a different idea from first-principles reasoning. It is first-principles reasoning, aimed at a requirement instead of at a physical problem. The building method is what the reasoning habit turns into on a factory floor: the open-ended “reason up from fundamentals” hardens into a numbered drill — question, delete, simplify, accelerate, automate, in that order. Same instinct, two altitudes. One decides what is true, the other what is necessary. That shared root is why running them backwards (“optimize the thing that should not exist”) is the error he warns about most.

Sequence: declare the endpoint, then ladder up to it in public

The third move decides the order in which decisions reach the market. The pattern is to fix a multi-decade destination before the means to reach it exist, publish it as a commitment, and treat each near-term product as a deliberate step. The original 2006 plan calls itself a living, revisable commitment rather than a fixed script:

“This plan is subject to continued refinement and revision.” ↗

Ten years later he audits it in public, restating the original goal from memory and grading it against reality:

“Basically, we were going to try to accelerate the world’s transition to sustainable energy. That plan is on track.” ↗

The product order the plan runs on is the high-end-then-down-market ladder: enter where the margins are fat, then drive each next model cheaper and higher-volume.

“The strategy of Tesla is to enter at the high end of the market, where customers are prepared to pay a premium, and then drive down market as fast as possible to higher unit volume and lower prices with each successive model.” ↗

By 2025 the staging is treated as a proven recipe, not a hope. The steps themselves are the method:

“All worthwhile journeys are long. And they all begin with a first step.” ↗

This is what keeps decisions coherent across decades. Each framed-and-cut product is not an end in itself but a rung, picked for where it sits on the ladder to a destination he named long before.

ℹ️ Sequencing is framing pointed at a roadmap. The master plan is the same “boil it down and reason up” move, stretched across twenty-five years instead of one design review. Declare the bedrock fact about the whole enterprise (the endpoint: sustainable transport, then whole-planet energy), then reason up through the intermediate products physics and economics actually require to get there. The public commitment is the first move’s let-reality-judge discipline run at company scale instead of problem scale. Publish the reasoning, invite the audit, grade yourself against what you said before. The plan is not a separate strategic instinct bolted onto the engineering. It is the framing move working on the longest horizon the machine has.

Bet: pre-accepted loss, extreme tempo, destruction as data

The first three moves, on their own, would still build a slow and cautious company. The fourth supplies the nerve and the speed. It decides how hard and how fast to push, given that, by his own forecasting model, he will often be wrong. The trick is not fearlessness. It is accepting the loss in advance, so the fear drops out, as he put it in 2016:

“If you just accept the probabilities, then that diminishes fear.” ↗

He routinely rates his own ventures as likely failures and goes all in regardless. The 2022 number:

“When we first started out Tesla, I thought we had — optimistically — a 10% chance of succeeding.” ↗

Accepting the loss up front is what makes destruction cheap, and cheap destruction is what makes the brutal tempo rational. At the 2021 Starbase tour he says the enabling fact straight out. A rocket with nobody aboard is meant to blow up, because every failure buys information:

“Starship does not have anyone on board so we can blow things up.” ↗

And he takes the embarrassment of failing in front of everyone as the price of going fast:

“this is definitely a case where we are washing out laundry in public.” ↗

The tempo that buys is the do-it-or-die-trying stance, where pressing on is the only acceptable mode, whether or not the odds feel good:

“And I don’t care about optimism or pessimism. Fuck that, we’re gonna get it done.” ↗

This is the engine speed of the whole machine. It is why the cutting move can delete aggressively, since a wrong deletion is cheap to add back. It is why the sequencing move can pull a product forward as confidence rises. It is why the framing move can toss out the consensus without flinching. Each time, the downside is already paid for.

ℹ️ Betting is what makes the other three moves cheap to run. A normal company can’t reframe from scratch, delete a load-bearing part, or re-sequence a public roadmap, because each of those risks an expensive, embarrassing failure. So it freezes, and optimizes the thing that should not exist. The betting temperament takes the brake off by accepting the loss in advance. Once a blown-up prototype is the cheapest way to learn (“we can blow things up”), deleting, reframing, and re-sequencing stop being scary and become routine. This is not a “move fast” slogan parked next to the method. It is the power supply the method runs on. Cut it off and the machine seizes. Switch it on and the same reasoning, cutting, and sequencing can run at a reckless tempo without fear.

What the machine reveals

  • A decision is a hand-off, not a single act. Framing produces a problem stated from fundamentals. Cutting produces the leanest design that works. Sequencing drops it onto a multi-decade ladder. Betting supplies the tempo. Each move’s output is the next move’s input, and the signature Musk error, “optimize the thing that should not exist”, is exactly what happens when a later move runs before an earlier one has finished.
  • Three of the four are the same instinct at different scales. Cutting is framing aimed at a spec; sequencing is framing aimed at a roadmap. The reason-up-from-fundamentals habit is the spine, run at the size of a problem, a design, and a company. That shared root is why the parts interlock so tightly: not four philosophies, but one, plus an engine.
  • The fourth move is what makes the first three usable. Pre-accepted loss and the horror of slack are not decoration. They are the condition that makes reframing, deleting, and re-sequencing cheap enough to do at speed. Radical decisions shipped fast come from the combination: physics-grounded framing alone would just be contrarian, and the engineering algorithm alone just tidy, without the temperament that ships them before they are safe.
  • It is built for deciding under uncertainty, not for being right. By his own 50th-percentile forecasting model, he expects about half his dated calls to miss. The machine is built to absorb exactly that: cheap deletions, revisable plans, blown-up prototypes that pay out in information. The aim is not infallible decisions. It is a procedure whose mistakes you can live through.

Connections