Elon Musk & Gwynne Shotwell on AI Risks and Peer Review, Starship, Terafab, SpaceX/Tesla Merger

The episode opens with a montage of clips before the interview itself. Asked about the latest timeline for Mars, one voice answers that people will be put down "within a decade," and another says they would be disappointed without a settlement on the moon. The montage introduces Gwen Shotwell (as transcribed) as president and CEO of SpaceX, calling her "the glue to the tornado." It also mentions a $75 billion IPO with a debut market value of $1.7 trillion, and the claim that the most efficient place to put compute is in orbit. Several engineering and risk slogans recur: "Fail fast, learn fast, make your design robust" and "Manage risk, don't avoid it."

The interview starts with a joke that nothing is happening at the company. The tenure question then gets a concrete answer. The interviewee is the longest-tenured employee after Elon, and their 24th anniversary fell on September 9.

The origin story is told as an accident. The speaker took a friend, who was leaving to work for Elon, out for a going-away lunch. When the speaker dropped the friend off, the friend invited them in to meet Elon. During the meeting the speaker told Elon that he needed a full-time head of business development instead of a part-time contractor. Back at the office, the speaker got a call from Elon's assistant, Mary Bass, asking them to apply for a new vice president of business development role. The speaker says it then took about four weeks to accept, and calls themselves "a total idiot" for the delay.

The company at that point is described as "very R&D," but with an investor who wanted a return, cash and customers. The speaker was the seventh employee to sign a contract, arrived as employee number 11 after a three-week gap (one week of it spent remodelling a bathroom), and the company sold a rocket about 12 months later, at a time when, in the transcript's words, "There wasn't a rocket."

By the way, SpaceX should never have existed, right?

The explanation for the early sales is structural, not about salesmanship. If the incumbent launch providers had been good, a new entrant could not have made headway against such entrenched players. Two conditions opened the door. Customer service was poor and rockets were "incredibly expensive." After 9/11, the military was "desperately looking" for rapid launch, because, as described, it had known where adversaries were for only about four hours after the attacks and wanted capability in theater. The speaker adds that this was not specific to SpaceX but reflected the state of the industry, and that it took a while, but the company did what it said it would do.

The role grew because a startup with sales needs customer management, finance and government affairs. The speaker recalls a defense contractor dismissing SpaceX as a fly on its big toe. They describe the company's Washington posture as purely defensive: "We never go on the offensive in Washington. We're always in defense mode."

The promotion came in 2008, around the cargo resupply contract (CRS). Elon was busy with both Tesla and SpaceX and needed a partner. The speaker says the offer to become president came on the runway, on the way to final negotiations for the $1.6 billion NASA contract. The first instinct was to decline, because they loved the job and their teammates, before deciding to accept.

The conversation jumps to the present. The interviewer says the company now looks as much an AI business as a space business, possibly even by revenue. In response, the speaker acknowledges "a lot of churn at XAI" and says SpaceX people stepped in to fill significant gaps. The speaker calls themselves "an AI noob" and repeats the framing that a company not using or leading AI runs a "significant risk of being irrelevant," which they say SpaceX could not accept. They agree that future rockets will be designed with these systems "100%."

On bringing highly trained engineers onto AI tooling, the interviewer cites mathematicians who were upset after, "I think it was Open AI," made progress on Navier–Stokes. The participants say there is debate over whether the solution was brute-forced, and that around 20 Fields medalists complained about ruining the craft. The response puts outcomes ahead of credit. The speaker would "rather just have the world be a better place, a smarter place" than worry about awards.