Elon Musk’s Latest Interview: 10 Questions Reveal Why First Principles Thinking Is a Superpower
Recently, at the YC AI Startup School, Y Combinator CEO Garry Tan hosted a special conversation with Elon Musk. What made this interview unique was the audience—young founders aged between 18 and 25, many already making waves in the AI space.
In this talk, Musk was strikingly candid. He shared down-to-earth, highly detailed stories—like drilling holes in the floor to run an internet cable for Zip2 in 1995, trying to buy intercontinental missiles in Russia in 2001, and how both SpaceX and Tesla nearly went bankrupt in 2008.
He also explained how he used "first principles thinking" in a “brute-force” way to drastically cut rocket costs, and how he built a data center with 100,000 chips in just six months. Musk went on to share his predictions and hopes for AI and the future of humanity.
This conversation broke the myth of the “genius founder,” instead revealing Musk’s disruptive mindset and pragmatic execution. His insights offer valuable takeaways for anyone on an entrepreneurial journey.
Just Wanted to Build Something Useful
Garry Tan: Today’s founders are getting younger and younger. Can you tell us what it was like when you were 18 or 19, learning to code and thinking about your first startup, Zip2? Let’s talk about your origin story.
Elon Musk: Back in 1995, I had two choices: either pursue a PhD in materials science at Stanford, focusing on electric vehicle battery tech, or dive into the then-unknown world of the internet.
I talked to a professor and asked, “Can I defer for one quarter?” I figured the internet thing would probably fail, and I’d just go back to school. He said, “This might be the last time we ever speak.” He was right.
In 1995, I developed one of the earliest online maps, navigation, and business directories—basically alone. I couldn’t afford a server or a T1 line. Our first office was on Sherman Ave in Palo Alto, and there was an internet service provider right downstairs. So I drilled a hole in the floor and ran a cable down to their network.
I didn’t set out to start a company. I even applied to work at Netscape but never heard back. I hung around their lobby trying to socialize—but I was too shy. So I thought, I’ll just write software myself. I just wanted to do something useful on the internet. Couldn’t get into a company? Then I’d start one.
First Startup Lesson: Let Technology Speak for Itself
Musk: At Zip2, we poured our hearts into developing incredible technology, but from my point of view, it never truly got to shine. Our investors and customers included The New York Times and Hearst, who also sat on our board. They saw everything through the lens of traditional media, pushing us to do things that made sense to them but clashed with our tech.
My first big startup lesson was this: don’t let traditional media companies hold too many board seats or shares.
Zip2 was a success—we sold it for $300 million, which was a huge amount back then. These days, there are unicorns everywhere. In AI, valuations can hit $10 billion or more in less than a year. Some will succeed, but many valuations are pretty shocking.
Going Direct-to-Consumer
Musk: After selling Zip2, I got $20 million. At the time, I was living with four roommates and had only $10,000 in the bank. Then suddenly, I received this huge check.
I poured almost everything into X.com and left all my chips on the table.
At Zip2, we were restricted by our clients—it felt like having clipped wings. I wondered what could happen if we had no limitations and went straight to consumers. That’s how PayPal was born. X.com and Confinity merged, and we had a team full of geniuses. The PayPal mafia probably went on to create more startups than any other group in the 21st century.
SpaceX: Something Only a Startup Could Do
Musk: One day on the Long Island Expressway, my college roommate Adeo Ressi asked me what I’d do after PayPal. I said maybe something space-related, even though I thought commercial space wasn’t realistic.
I was curious about sending humans to Mars. But when I searched NASA’s site, I found nothing about a Mars plan.
At first, I thought of sending a tiny greenhouse to Mars. In 2001 and 2002, I did something wild: I went to Russia and asked their military leadership, “Can I buy a few missiles?” They were decommissioning them due to arms reduction, so I thought—maybe we could buy a couple, remove the warheads, add an upper stage, and launch them into space. But the prices kept rising, and I realized: it’s not that humanity lacks the will—it’s that the cost of existing technology makes it impossible, even for NASA.
So in 2002, I decided to start SpaceX. The chance of success was under 10%—maybe even 1%. But if a startup didn’t try, no one would. Big aerospace companies only serve government contracts, and governments prefer traditional projects. This was a job only a startup could do. A tiny chance is better than no chance.
Around the same time, we were starting Tesla. In summer 2008, SpaceX’s third launch failed, Tesla’s funding collapsed, and both companies were on the verge of bankruptcy. The fourth launch succeeded—but we still needed a NASA contract to survive.
Then on December 22, just before Christmas, NASA called and said, “We’ve decided to award you the ISS resupply contract.” I literally said, “I love you.” On December 24 at 6 PM—the final hour of our funding deadline—Tesla secured its investment. 2008 was brutal.
Many Fail Because Their Ego-to-Skill Ratio Exceeds sin(1)
Garry Tan: From Zip2 and PayPal to now, you’ve consistently attracted amazing talent. If you could go back and give advice to your younger self, what would you say?
Musk: Try to be useful. I think of usefulness like “effective work” in physics. That is, if your product succeeds, how much value does it deliver, and to how many people? That’s how I define “effective work.”
Whether you’re a CEO or any role in a startup, you must be useful and kill your ego. A lot of people fail because their ego-to-ability ratio exceeds sin(1). If that ratio gets too high, it cuts off your connection to reality. In AI terms, it breaks your reinforcement learning loop. You need to maintain a strong feedback loop. That means taking responsibility and staying humble.
That’s also why I prefer the word “engineering” to “research.” I don’t want X.ai to be called a lab—I want it to be a company. The simpler and less self-important your language is, the closer you are to success. You have to stay grounded in reality.