In August 2017, 29-year-old OpenAI cofounder Greg Brockman was wrestling with his feelings about the future of the organization — and what role his “fucking famous” cofounder Elon Musk should have in it. 

According to an email written by an OpenAI advisor and produced in Musk’s lawsuit against Brockman, Sam Altman, and OpenAI, Brockman was so caught up in making artificial intelligence safe that he “broke up with this unbelievably awesome woman” in order to “focus all his time” on it. 

Nine years later, the “blood, sweat, and tears” Brockman testified he has poured into OpenAI, maker of ChatGPT, have paid off, as he is $30 billion richer and accompanied on most days in court by his wife, Anna, often in her crisp white blazer. 

For those following the case, the emails, text messages, and testimonies shine a light on how the tech elite live — from their favorite desserts to their late-night texting habits. They also offer a unique illustration of the shifting hierarchy between a handful of young idealists as they navigate egos and allegiances on the path to becoming tech’s next generation of titans. 

Musk is suing OpenAI and his cofounders for allegedly tricking him into donating millions to a nonprofit, only to cut him out before an influx of capital from Microsoft helped make it one of the most valuable companies in history. The jury, which begins deliberations on Monday, must decide whether Musk missed his chance to bring the suit due to statute of limitations and, if he didn’t, whether the OpenAI defendants intentionally duped him, broke their promises, and enriched themselves in the process.

Few regular people had heard the phrase “artificial general intelligence” in 2015, when the 11 founders of OpenAI were strategizing on how to protect the world from the very technology they were racing to build. 

At the time, Musk was already the CEO of Tesla and SpaceX and famous for founding PayPal and being featured on The Simpsons. To his cofounders, Musk’s reputation was as much a draw as his stockpile of cash. 

On the witness stand Monday, OpenAI cofounder Ilya Sutskever acknowledged that the opportunity to work with Musk was enough to tempt him away from the promise of a salary of $6 million a year at Google, which had acquired AI startup DeepMind the year before. 

Over three weeks of trial, Musk’s lawyers rarely missed a chance to emphasize this.

“Could you have picked up the phone and called Satya?” Musk’s lead counsel asked Brockman, referring to Microsoft’s CEO. Brockman said he didn’t know Satya Nadella in the early days. 

The following week, the same lawyer asked Altman, now OpenAI’s CEO, if he had considered himself in the “same league” as Musk, Bill Gates, Pierre Omidyar, and other industry leaders he’d reached out to for fundraising in OpenAI’s early years. 

“No,” Altman replied. “At the time, no, of course not.” 

While they might not have had the biggest tech names on speed dial in 2015, Brockman and Sutskever were by no means just starting out; Brockman had been CTO of payments company Stripe, and Sutskever was a well-respected research scientist. When the two got together for dinner, it was for fine dining at the likes of Woodside’s The Village Pub, which, according to court documents, serves Brockman’s “favorite dessert.” 

As OpenAI’s capabilities advanced, Brockman explored his feelings about his changing role in a private journal that, much to his chagrin, became court evidence. Describing the journal as a “stream of consciousness,” both he and his lawyer tried to frame the document as a testing ground for ideas. But it nonetheless illustrates Brockman’s ambitions. 

“What do I really want?,” he asked himself on Aug. 21, 2017. “I want to be an engineer. But I think now is a crazy shot to be the one in charge and to step up to the challenge.”

OpenAI hit a milestone in the summer of 2017: winning an international video game tournament in Seattle. “There was a big stadium with 20,000 people watching our AI play,” Brockman recalled in his testimony.

The victory was a big deal for AI, and for OpenAI in particular, Altman and Musk agreed. But it also set off a new era of OpenAI’s history that put Brockman and Sutskever in head-to-head negotiations with their idol. It culminated six months later with Musk storming out. 

Musk, who didn’t attend the tournament, emailed the team to invite them to meet up the next day and discuss OpenAI’s future at his “Haunted Mansion (opens in new tab),” a 16,000-square-foot house in the South Bay where he liked to throw parties, according to Business Insider. 

“It’s kind of crazy and weird and we’ll have party carnage,” Musk wrote in an email. “But it might make for a good backdrop.”

When Sutskever, Brockman, and Altman arrived, it was, indeed, “clear there had been a party there the night before,” Brockman said in his testimony. “There was confetti and cups and the whole thing all around.” The team sat on couches, he said, and Musk’s then-girlfriend Amber Heard “served some nice whiskey.” 

The mood that day was celebratory, Brockman said, but within a few weeks, it had deteriorated. As the group tried to hash out a plan for transforming OpenAI the nonprofit into a company that could attract the capital necessary to achieve its AI goals, disagreements cropped up over control; specifically, whether Musk could be trusted to eventually give it up, and if he had the AI chops to head the project.  

To smooth over the difficult feelings, Musk gave Brockman and Sutskever each a Tesla. In return, Sutskever — a hobbyist painter — gave Musk a painting of a Tesla. “At least we’re getting our Teslas,” Sutskever joked to Brockman in a text message sent Aug. 22, 2017. “Will a Model 3 make you be willing to accept massively unfavorable terms?” 

The gift exchange didn’t have the desired effect. According to Brockman’s testimony, he and Sutskever met with Musk again a week later, when they made it clear that they weren’t willing to bend to Musk’s wishes.

“At the end of the meeting, he sat quiet for several minutes just thinking,” Brockman recalled, before rejecting the terms. Then, he said, Musk “stood up, and he kind of stormed around the table” where the painting Sutskever had made for him sat. 

“I thought he was going to hit me. I thought he was going to physically attack me,” Brockman said. Then Musk “grabbed the painting and stormed out.”

Brockman once again turned to his journal. 

“I’m starting to feel more of the fighting spirit,” he wrote a couple weeks after the incident. “Honestly, I just defer to him and say nice things and never push back on anything. I wouldn’t respect me very much in his shoes.”

As Brockman was finding the gumption to stand up to Musk, Shivon Zilis, another key early player at OpenAI, was also struggling to reconcile her relationships. 

Zilis, who got her start in venture capital, joined OpenAI as an adviser in 2016. She was present at the video game tournament in Seattle and the Haunted Mansion meeting and went on to have four children with Musk. In 2020, she joined OpenAI’s board. 

One of Zilis’ jobs throughout this period was to serve as a “bridge” between Altman and Musk, whom she worked for at Tesla and later Neuralink. The role required a soft touch. Text message records with Altman show her trying to predict what might set Musk off and celebrating after a meeting concluded without a blowup. 

In early 2018, with Musk weeks away from announcing his departure from OpenAI’s board, Zilis texted Altman to propose a possible way forward for the company as a subsidiary of Tesla.The friendly exchange that followed, in which Altman asked whether she was planning to attend the SpaceX Falcon Heavy rocket launch in Florida, show a 32-year-old still navigating her proximity to power. 

“I so badly want to but feels too entitled,” Zilis said. 

“Not at all,” Altman replied. “It’s your bday.” 

“The simulation video alone gave me chills ten times over,” Zilis continued. But she remained wary of using her connections to access an event that wasn’t “publicly viewable.”

“No way I’m going,” she told Altman. “Would never in a million years ask!!!” 

A few days later, she texted Musk, “Do you prefer I stay close and friendly to OpenAI to keep info flowing or begin to disassociate? Trust game is about to get tricky so any guidance for how to do right by you is appreciated.” 

 “Close and friendly,” he responded. 

The trust game did, indeed, become tricky. Just 10 days later and a few days ahead of Musk’s departure, court records show, Altman texted Zilis again, this time to grill her about Musk’s effort to poach engineers, and accusing Musk of telling staff he “didn’t think OpenAI could succeed.” 

“This interaction didn’t feel right,” Zilis wrote in response. “Not urgent, but let’s chat briefly next week about some general stuff. I would love to keep helping you as much as I can but important to figure out the correct frameworks of trust for each other.” 

Zilis did continue helping OpenAI for a few more years. Altman didn’t learn until 2022 that Musk was the father of her children. Even then, he decided to keep her on OpenAI’s board, testifying that it was a “close call for me personally” but that he “values her counsel.”

But by early 2023, it had become clear that Musk intended to launch a competing AI business, xAI, and the awkward arrangement was no longer sustainable. 

Zillis was asked to resign from OpenAI’s board in February 2023. She and Musk “had a talk” afterward, Zilis told a friend via text message, and he apologized for having “pruned” her “friend network.”

On day one of trial, Judge Yvonne Gonzalez Rogers advised potential jurors to think of the Musk trial not as a “technical case,” but “just a case about promises and breaches of promises.”

As they strive to determine whether promises made a decade ago were, in fact, broken, Altman and Musk have continued on with their lives. In the second week of trial, Altman threw a party in San Francisco for the latest release of ChatGPT, OpenAI’s flagship product, where he was photographed talking to fans and standing next to a portrait of himself in the style of the Mona Lisa. 

The next day, Musk announced a major deal for SpaceX — now SpaceXAI — to provide computing power to Anthropic, OpenAI’s chief rival. According to a tweet he posted, Musk had used his free time after court to get to know the company’s leaders. 

‘Everyone I met was highly competent and cared a great deal about doing the right thing,” he said. “No one set off my evil detector.”