Elon Musk
What Elon Musk has actually said about building things (the five-step "algorithm", deletion, first principles, iteration speed, urgency, hiring and mission), checked against Walter Isaacson's 2023 biography, his own talks and interviews, and the record of what happened next. The reasoning method itself has its own sheet, first principles; the practical build pipeline is in idea to MVP. Compare with Steve Jobs and Peter Thiel.
Who he is and why listen
| Company | Role | What it shows |
|---|---|---|
| Zip2 (1995), X.com / PayPal (1999–2002) | co-founder | the first exits that funded everything else |
| SpaceX (2002) | founder, CEO, chief engineer | reusable orbital rockets; the clearest case of his engineering method working |
| Tesla (investor and chairman 2004, CEO 2008) | CEO, "product architect" | mass-market EVs; also the clearest case of the method failing first (Model 3 automation) |
| Twitter / X (bought October 2022) | owner, then executive chair and CTO | what the same style does to a software and advertising business |
Why listen: few people have built two hardware companies from nothing to category leaders, and both leaned on the same few habits: ruthless deletion of requirements, cost reasoned from materials, and very fast test cycles. Why be careful: he also has a long record of missed deadlines, labor and safety problems, and public behavior that cost his companies money. Take the engineering method; judge the rest on the evidence.
Most of the famous Musk process quotes come from two places: Isaacson's Elon Musk (2023), which reproduces the algorithm as Musk dictated it to his teams, and Musk's July 2021 Starbase tour with Tim Dodd (Everyday Astronaut), where he walked through it on camera. Wording differs slightly between the two; both are given below.
The algorithm
Musk told Isaacson he became "a broken record" on a five-step process for any design or production line, and he insisted it be run in order. The step names below are Musk's (via Isaacson); the rest of each row is summary.
| Principle | What it means | Source | Apply it by… |
|---|---|---|---|
| 1. Question every requirement | every requirement must come with the name of the person who made it, never "legal" or "safety"; requirements from smart people are the most dangerous because nobody challenges them, including Musk's own | Isaacson 2023; Starbase tour 2021 ("make the requirements less dumb") | writing an owner's name next to every line of your spec and asking each owner "what breaks if we drop this?" |
| 2. Delete any part or process you can | delete aggressively; if you don't end up adding back at least 10% of what you cut, you didn't cut enough | Isaacson 2023; Starbase tour 2021 | cutting features, form fields, approval steps and integrations until something you need breaks, then restoring only that |
| 3. Simplify and optimize | only now improve what survived; optimizing a thing that should not exist is the classic smart-engineer mistake | Isaacson 2023; Starbase tour 2021 | refactoring and polishing only code paths and screens that survived step 2 |
| 4. Accelerate cycle time | every process can be sped up, but only after steps 1–3 | Isaacson 2023; Starbase tour 2021 | shortening deploy, review, and customer-feedback loops for the flows that remain |
| 5. Automate | last; automating a process that should have been deleted locks the mistake in | Isaacson 2023; Starbase tour 2021 | doing the task by hand (a spreadsheet, a manual email) until it's stable, then scripting it |
Two lines worth having verbatim. From Isaacson, on step 1: requirements must be traced to "the name of the real person who made that requirement". From the Starbase tour (as transcribed by Startup Archive): "Your requirements are definitely dumb". Musk's point is not that requirements are useless; it is that every requirement starts out wrong to some degree and has to earn its place.
The mistake of running it backwards
Musk's own worked example of doing it wrong is the Tesla Model 3 ramp (2017–18). In the version Isaacson reproduces, he says his big mistake in Nevada and Fremont was trying to automate every step first, and that he spent a lot of time accelerating processes that should have been deleted. That is steps 5 → 4 → 3 run in reverse.
- July 2017: at the first Model 3 handover Musk warned of "at least six months of production hell" (NBC News).
- April 2018: after the line missed its targets, he tweeted "Yes, excessive automation at Tesla was a mistake. To be precise, my mistake. Humans are underrated." Tesla had guided to 5,000 Model 3s a week by the end of 2017; Musk announced the first 5,000-car week on 1 July 2018.
- On the Starbase tour he gave a smaller example of step 2: fibreglass mats on the Model 3 battery pack that the battery team and the noise-and-vibration team each said the other needed. Nobody could name an owner, and the mats were deleted (his account; also retold by Isaacson).
The lesson for a startup: automation, scale and optimization are multipliers. Multiply a wrong process and you get more wrong, faster. Most early-stage "we need to automate onboarding" or "we need a data pipeline" problems are step-2 problems in disguise.
The corollaries
Isaacson lists "corollaries" Musk added to the algorithm. Quoted wording is Musk's as reproduced in the book.
| Corollary | What it means | Apply it by… |
|---|---|---|
| technical managers must be hands-on | software managers should spend at least 20% of their time coding; solar-roof managers do installs | founders keeping a direct hand in the product, support and sales until well past product–market fit |
| comradery is dangerous | friendship makes people reluctant to challenge each other's work | separating "I like you" from "this design is wrong" in reviews |
| "It's OK to be wrong. Just don't be confident and wrong." | admit mistakes early; overconfidence is the failure | stating confidence levels on decisions ("70% sure") and revisiting them |
| never ask your troops to do what you won't | leadership by example | doing the worst job yourself once before delegating it |
| skip-level meetings | when there is a problem, meet the people below your managers | talking to the engineer or support agent closest to the problem |
| hire for attitude | skills can be taught; attitude changes "require a brain transplant" | weighting trial work and references over credentials |
| "A maniacal sense of urgency is our operating principle." | speed is a default, not a phase | see urgency below |
| physics is the only hard rule | the only rules are the laws of physics; everything else is a recommendation | separating true constraints (law, math, physics) from habits |
The best part is no part
Musk's deletion slogan predates the 2021 tour. In the Q&A at SpaceX's Starship Update in Boca Chica on 28 September 2019 he said: "the best part is no part. The best process is no process. It weighs nothing, costs nothing, can't go wrong." In the same answer he said the thing he is most impressed by in SpaceX design reviews is the question "What did you undesign?", and gave his schedule mantra: "If a schedule is long, it's wrong. If it's tight, it's right." (Wording per Wikiquote's transcript of the video; Tim Dodd tweeted the "best part" line as a Musk quote in December 2019.)
On the Starbase tour he tied it to his claim that manufacturing, not design, is the hard problem: he said designing the production system for Raptor was 10–100× more effort than designing the engine, and the Everyday Astronaut write-up records him describing Starship features removed under the algorithm (a stage-separation pusher system, dedicated hot-gas thrusters on the booster).
| Where it applies in software | Deleting it looks like |
|---|---|
| features | cut the settings page, the second pricing tier, the admin dashboard nobody uses |
| process | drop the approval step, the weekly status report, the design sign-off for trivial changes |
| infrastructure | one database, one deploy target, no microservices until one service hurts |
| sign-up | email only; no profile, no tour, no card until they've seen value |
The 10% add-back rule is the useful calibration: if nothing you deleted ever came back, you were too cautious. Scoping the MVP is the same idea applied before you build.
Requirements, physics and first principles
First principles, briefly
Musk's standard explanation (TED 2013, with Chris Anderson): first principles means "boil things down to their fundamental truths and reason up from there, as opposed to reasoning by analogy". His best-known example is from a September 2012 interview with Kevin Rose: battery packs were said to cost about $600/kWh and always would, but the raw materials priced on the London Metal Exchange came to about $80/kWh, so the gap was a manufacturing problem, not a law of nature. BloombergNEF's December 2025 survey put the average lithium-ion pack at $108/kWh ($99/kWh for battery-electric-vehicle packs); most of that fall came from the whole industry, especially Chinese LFP makers, not from Tesla alone. The full method, the idiot index math and worked examples are on first principles; they are not repeated here.
The idiot index
Isaacson describes Musk asking for the ratio of a part's finished cost to the cost of its raw materials. A high ratio means the design or the manufacturing is wasteful, so that is where to look first (paraphrased from Isaacson). The software analogue is the ratio of what you pay (vendor price, engineer-weeks) to what the thing does: a $2,000/month tool used to send one weekly email has a high idiot index.
Physics as the only hard constraint
Isaacson's corollary puts it bluntly: the only rules are the ones dictated by the laws of physics, and everything else is a recommendation. Read charitably, it's a way of sorting constraints:
| Kind of constraint | Examples | Musk's stance |
|---|---|---|
| physics and math | energy density, orbital mechanics, latency across an ocean | hard; design around them |
| law and safety regulation | FAA launch licenses, vehicle safety standards | treated as negotiable in practice, which is where his companies repeatedly got into trouble (see critiques) |
| industry convention | "launches cost $X", "cars are sold through dealers" | a recommendation; this is where the upside lives |
| internal habit | "we always do a security review for that" | ask who made the rule (step 1) |
For a startup the useful part is the bottom two rows. The second row is where copying Musk gets people fined.
Iteration speed and testing hardware
Musk's claim is that speed of iteration beats quality of any single iteration, when failure is cheap. He is explicit that this depends on the stakes (Starbase tour, per the Everyday Astronaut write-up):
| Vehicle | Failure tolerance | Development style |
|---|---|---|
| Crew Dragon | can never fail | huge margins, exhaustive testing |
| Falcon 9 | can lose a landing, not an ascent | in between |
| Starship (uncrewed) | blowing up is data | build many prototypes, push each until it breaks |
His contrast was the Space Shuttle, whose design he said "froze" because every flight carried crew, so change was high-risk and low-reward. On Starship: "Starship does not have anyone on board so we can blow things up. It's really helpful."
What happened: the first full Starship stack launched on 20 April 2023 and exploded about four minutes after liftoff, and the launch threw debris and sand over the surrounding area. On 13 October 2024, flight 5, the Super Heavy booster was caught by the launch tower's arms, a design chosen to delete landing legs (step 2 again). NASA's crewed programs, by contrast, are built around not failing at all; that is the right choice with people on board and a slow way to learn.
Transfer to software: your "uncrewed prototype" is a feature flag, a staging environment, or a small cohort. Ship small, instrument it, expect some to fail. Don't apply Starship rules to payments, auth, or anything that loses customer data; that's your Crew Dragon. See launch and iterate.
Vertical integration and the factory
Musk's recurring theme is that the factory is harder than the product. On the Starbase tour he said a factory is underrated and design is overrated. SpaceX builds its engines, structures and much of its avionics in-house rather than buying from the traditional aerospace supply chain, and Tesla makes its own battery packs, motors and much of its software. The motive is cost and cycle time: an outside supplier's quote, lead time and change process sit between you and every iteration. (Precise in-house percentages circulate widely; I could not verify one, so none is given here.)
| Integrate when | Buy when |
|---|---|
| the part is core to the product's cost or performance | it's a commodity with many good suppliers |
| supplier margins or lead times block iteration | you'd be building for fewer than a handful of units |
| you need to change it weekly | the vendor is better than you'll be for years (payments, email delivery, cloud) |
A seed-stage software company should integrate almost nothing: buy auth, payments, hosting and email, and spend engineering time only where you differ. That's the algorithm applied honestly: vertical integration is itself a requirement to question.
People: hiring, managers and communication
| Principle | What it means | Source | Apply it by… |
|---|---|---|---|
| hire for attitude and ability, not credentials | skills can be taught, attitude can't | Isaacson 2023 (corollaries) | using paid trial projects and asking for the hardest thing the candidate has built |
| evidence of exceptional ability | Musk has often said he looks for a track record of exceptional work rather than degrees (paraphrased; the exact phrase circulates without a primary source I could check) | Musk interviews, various | asking "what's the most impressive thing you've done, and what was your part?" |
| managers stay hands-on | managers who can't do the work can't judge it | Isaacson 2023 | founders reviewing real code, calls and tickets |
| communication goes direct | anyone should talk to anyone to get the job done, not route through the chain of command | Musk email to Tesla staff, April 2018, as leaked and widely reported (paraphrased; I could not retrieve the original text) | letting any engineer message any customer, investor or founder directly |
| skip-levels | go below managers to find the truth | Isaacson 2023 | talking to the newest support hire when metrics look odd |
The April 2018 email (sent during the Model 3 crisis) is usually summarized as: cut large and frequent meetings, walk out of meetings where you're not adding value, drop jargon and acronyms, and communicate by the shortest path rather than the org chart. The spirit is sound for any company over about 15 people; a four-person startup has no chain of command to bypass.
Urgency, surges and "Elon time"
Isaacson's corollary: "A maniacal sense of urgency is our operating principle." On the Starbase tour Musk explained why, for SpaceX: "if we operate with extreme urgency we have a chance of making life multi-planetary. It's still just a chance, not for sure. If we don't act with extreme urgency, that chance is probably 0." Isaacson describes his habit of surges: sudden all-hands pushes to break a bottleneck, such as the 2018 Model 3 ramp, when an extra assembly line was built in a tent at Fremont.
The flip side is deadlines that are wrong by years, widely mocked as "Elon time". Examples, from Wikipedia's sourced summaries:
| When | Claim | What happened |
|---|---|---|
| December 2015 | "complete autonomy" by 2018 | not delivered |
| end of 2016 | Tesla to demonstrate full autonomy by end of 2017 | not delivered |
| 2017 | Model 3 at 5,000 a week by end of 2017 | first 5,000-car week announced 1 July 2018 |
| April 2019 (Autonomy Day) | one million robotaxis on the road in 2020; no driver oversight by Q2 2020 | Robotaxi launched 22 June 2025 in Austin for invited riders, with safety monitors |
| early 2021 | Level 5 autonomy by end of 2021 | not delivered; the product is now sold as "Full Self-Driving (Supervised)" |
| 2023 | full autonomy "later this year" | not delivered that year; small driverless fleets ran in several Texas and Florida cities by September 2026 |
Blunt read: impossible deadlines do pull work forward (a tight schedule forces step 2), but only inside the team. When the same deadlines are told to customers and investors they become broken promises, and Tesla has faced lawsuits and regulatory scrutiny over how FSD was marketed. Set aggressive internal targets; make external commitments you can keep.
Mission and master plans
Musk reasons backwards from a mission: SpaceX exists to make life multiplanetary, Tesla to accelerate the move to sustainable energy. The mission is used to justify priorities, and to justify urgency.
The template is his blog post "The Secret Tesla Motors Master Plan (just between you and me)" (2 August 2006). In it he framed Tesla's purpose as helping move from a "mine-and-burn hydrocarbon economy" toward a solar electric one, and laid out the sequence (paraphrased): build an expensive, low-volume sports car; use that money to build a cheaper, higher-volume car; use that money to build an affordable mass-market car; and alongside, offer zero-emission power generation. That is roughly what happened (Roadster, Model S, Model 3). Sequels followed: Master Plan, Part Deux (July 2016: solar, more vehicle types, autonomy, car sharing), Part 3 (2023) and Part 4 (2025).
| Principle | What it means | Source | Apply it by… |
|---|---|---|---|
| enter at the high end | early adopters pay for the learning curve | Master Plan (2006) | starting with the customer segment that will pay most for a rough product |
| publish the plan | a written sequence is a commitment and a recruiting tool | Master Plan (2006) | writing a one-page "if this works, then…" plan for your team |
| mission justifies focus | anything off-mission is deleted | Starbase tour 2021; Master Plans | testing each project against the one-sentence mission |
The caution: a mission is also a way to excuse anything. See critiques.
Feedback and being wrong
| Principle | What it means | Source | Apply it by… |
|---|---|---|---|
| seek negative feedback | "really pay attention to negative feedback, and solicit it, particularly from friends" | TED 2013 (Chris Anderson interview) | asking five users "what would make you stop using this?" and friends "what's the weakest part?" |
| don't be confident and wrong | admitting a mistake is fine; clinging to it isn't | Isaacson 2023 (corollaries) | holding post-mortems on your own decisions, not just the team's |
| everyone is wrong some of the time | including the person who wrote the requirement | Starbase tour 2021 (paraphrased in the Everyday Astronaut write-up) | re-opening old decisions when the evidence changes |
Musk's TED 2013 advice pairs the feedback line with his first-principles line; both are about not trusting received wisdom, including your own. He applies it unevenly: in 2023 he conceded his earlier self-driving predictions had been too optimistic, then made another one.
Critiques and limits
Kept to the documented record.
| Issue | What's documented |
|---|---|
| workplace safety at Tesla | Reveal (Center for Investigative Reporting) reported in 2018 that Tesla under-counted Fremont injuries; Tesla called it an ideologically motivated attack; Cal/OSHA later confirmed under-reporting for 2018. Tesla has acknowledged its injury rate exceeded the industry average in 2013–16. |
| workplace safety at SpaceX | Reuters in 2023 documented at least 600 previously unreported workplace injuries since 2014, including amputations and one death. |
| labor law | in 2021 the NLRB upheld a ruling that Tesla had illegally fought a union drive at Fremont and ordered a fired organizer reinstated; in 2023 a jury awarded $3.175m to Owen Diaz over racial harassment at Fremont (a retrial after a much larger 2021 award was cut) |
| Twitter / X | about half of staff were laid off from 4 November 2022; on 16 November Musk told the rest to commit to being "extremely hardcore" (long hours at high intensity) or leave; he later told the BBC headcount went from about 8,000 to under 1,500. The service kept running, but advertisers left: Fidelity marked its stake down 79% by October 2024 before the business was folded into xAI in March 2025. |
| burn-out culture | the urgency and surge style depends on people working extreme hours; Isaacson and many former employees describe it as exhausting and high-turnover |
| erratic public behavior | his August 2018 "funding secured" tweet about taking Tesla private led to an SEC fraud charge; he and Tesla each paid $20m and he stepped down as chairman (SEC, 29 September 2018) |
| regulation as "a recommendation" | FSD marketing, the SEC case and repeated clashes with regulators follow directly from treating non-physics rules as optional |
| survivorship | the method is celebrated because SpaceX and Tesla survived; both nearly went bankrupt in 2008, and the same habits produced the Model 3 near-disaster and the Twitter advertiser collapse |
Blunt conclusion: the engineering method (question, delete, simplify, then speed up, then automate) is excellent and costs nothing to copy. The management style (berating staff, which Isaacson calls "demon mode", extreme hours, mass firings, deadlines as marketing) is not a method, and there is little evidence it is why his companies succeeded rather than something they survived.
What transfers to a small startup
| Transfers | Doesn't transfer |
|---|---|
| the algorithm, in order, applied to your spec, your backlog and your processes | vertical integration: buy everything that isn't your product |
| naming an owner for every requirement | "physics is the only rule": laws, privacy and consumer protection apply to you |
| hardware-rich testing → feature flags, cohorts, cheap experiments | blowing things up in production where customer data or money is involved |
| tight internal deadlines that force deletion | promising those deadlines to customers and investors |
| hands-on founders, direct communication | "extremely hardcore" ultimatums; a four-person team that burns out has no bench |
| cost from first principles (what should this cost?) | billion-dollar capital intensity; his runway is not yours |
| a written mission and master plan | using the mission to excuse bad behavior |
For the step-by-step build process see building the MVP; for how YC frames the same speed and focus, see Y Combinator.
Takeaways checklist
[ ] Every requirement in the spec has a person's name on it
[ ] I asked each owner what breaks if we drop it
[ ] I deleted until something broke; ~10% came back
[ ] I only optimized what survived deletion
[ ] I sped up loops (deploy, review, feedback) after that
[ ] I automated last, and only stable, repeated work
[ ] Hard constraints (physics, law, math) listed apart
from conventions and habits
[ ] Risky, reversible work ships fast behind flags;
payments, auth and data get "Crew Dragon" rules
[ ] Internal deadlines are tight; external promises are
ones we can keep
[ ] I buy commodity parts; I build only what makes us
different
[ ] I asked users and friends for negative feedback
this week
[ ] Our mission fits in one sentence and I used it to
say no to something
[ ] Nobody on the team is being run into the groundReferences
- Walter Isaacson, Elon Musk (Simon & Schuster, 2023): the five-step algorithm, its corollaries, the idiot index, surges and the Model 3 ramp
- Farnam Street, "The Busy Person's Algorithm for Solving Problems" (opens in a new tab): reproduces the algorithm and corollaries as printed in Isaacson
- Everyday Astronaut, "Starbase Tour and Interview with Elon Musk" (Trevor Sesnic, 11 August 2021) (opens in a new tab): write-up of the 30 July 2021 tour, including the five steps, Dragon vs Starship testing and the urgency quote
- Everyday Astronaut, "Starbase Tour with Elon Musk [PART 1 // Summer 2021]" (YouTube) (opens in a new tab): the primary video
- Startup Archive, "Elon Musk explains his 5-step algorithm for running companies" (opens in a new tab): transcribed clip ("Your requirements are definitely dumb")
- SpaceX, "Starship Update" (28 September 2019, YouTube) (opens in a new tab): Q&A source of "the best part is no part" and "if a schedule is long, it's wrong"
- Wikiquote, "Elon Musk" (opens in a new tab): transcript of the 2019 Starship Update quote with timestamp
- Tim Dodd on X (December 2019) (opens in a new tab): early citation of "the best part is no part"
- TED, "Elon Musk: The mind behind Tesla, SpaceX, SolarCity" (2013), transcript (opens in a new tab): first principles vs analogy; "pay attention to negative feedback"
- Kevin Rose, "Elon Musk and Kevin Rose" (Foundation, 2012, YouTube) (opens in a new tab): the battery-cost example
- Startup Archive, "Elon Musk explains first principles thinking" (opens in a new tab): transcript of the battery passage
- Chris Anderson, "Elon Musk's Mission to Mars", Wired (October 2012) (opens in a new tab): rocket materials cost vs price
- BloombergNEF, "Lithium-Ion Battery Pack Prices Fall to $108 Per Kilowatt-Hour" (December 2025) (opens in a new tab): current pack prices
- Elon Musk on X, 13 April 2018 (opens in a new tab): "excessive automation at Tesla was a mistake"
- TechCrunch, "Elon Musk says 'humans are underrated'" (13 April 2018) (opens in a new tab): context of the automation tweet
- NBC News, "Elon Musk Warns of 'Manufacturing Hell' for Model 3" (July 2017) (opens in a new tab): the production-hell warning
- Elon Musk, "The Secret Tesla Motors Master Plan (just between you and me)" (Tesla blog, 2 August 2006) (opens in a new tab): the original plan
- Elon Musk, "Master Plan, Part Deux" (Tesla blog, 20 July 2016) (opens in a new tab): the sequel
- SEC press release 2018-226 (29 September 2018) (opens in a new tab): settlement over the "funding secured" tweet
- Wikipedia, "Tesla Autopilot" (opens in a new tab) and "Tesla Robotaxi" (opens in a new tab): sourced lists of autonomy predictions and outcomes
- Wikipedia, "Tesla Model 3" (opens in a new tab): production targets and the July 2018 milestone
- Wikipedia, "SpaceX Starship" (opens in a new tab): flight test 1 (April 2023) and the first booster catch (October 2024)
- Wikipedia, "Criticism of Tesla, Inc." (opens in a new tab): Reveal investigation, Cal/OSHA, NLRB and Diaz rulings, with citations
- Wikipedia, "SpaceX" (opens in a new tab): Reuters 2023 injury investigation, with citation
- Wikipedia, "Twitter under Elon Musk" (opens in a new tab): layoffs, the "extremely hardcore" email, headcount and valuation markdowns