Supporters of Marcus Endicott’s Patreon can access weekly or monthly consultations on this topic.
On Friday, December 11, 2015, a blog post appeared under the title "Introducing OpenAI." Its opening sentences have since been read as both scripture and indictment, depending on the reader's mood: the new venture, it declared, was a nonprofit whose goal was to advance digital intelligence "in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return." Freed from the obligations of profit, the founders argued, they could pursue a good outcome for everyone rather than for shareholders. It was a statement of purpose so clean that nearly everything which followed would be measured against it.
The same post contained, a few paragraphs down, the sentence that would haunt the enterprise. Its backers — Sam Altman, Greg Brockman, Elon Musk, Reid Hoffman, Peter Thiel, Amazon Web Services, Infosys, and others — had, it said, "committed $1 billion." The number entered the world as a headline and never quite left, but the operative word was committed, not given. The post itself conceded the founders expected to spend only a tiny fraction in the early years, and the gap between pledge and cash turned out to be vast. Later analysis of the nonprofit's federal tax filings showed that the total actually received through 2021 was roughly $133 million; one named donor, YC Research, contributed nothing at all. Musk, who would one day claim to have poured more than $100 million into the project, left a documented trail of around $38 to $45 million, only some $15 million of it definitively traceable to him. The billion-dollar founding was, from the first day, more promise than bank balance.
OpenAI was, in its first incarnation, less a company than a fragile alliance of strong personalities. Altman and Musk were co-chairs; Brockman, lately the chief technology officer of Stripe, was its CTO; the research would be led by Ilya Sutskever, a student of Geoffrey Hinton and a co-author of the AlexNet result that had detonated the deep-learning era in 2012. Sutskever was the prize the founders most wanted. He was at Google Brain, and both Google and DeepMind fought to keep him, dangling counter-offers that one internal message described in panicked terms — Altman wrote that DeepMind planned to give everyone at the nascent venture "massive counteroffers tomorrow to try to kill it." What finally sealed Sutskever's decision, by Musk's account, was a personal phone call from Musk on the day of the public launch. Musk would later call him "the linchpin for OpenAI being successful," and would describe the recruitment as the thing that ruptured his long friendship with Google's Larry Page; the breaking of that friendship, he said, was over OpenAI, and specifically over recruiting Sutskever. To hold the team against the pull of the giants, the founders raised their offers the very day after launch.
The alliance did not hold. For years the public story of Musk's departure, announced in February 2018, was a tidy one about conflict of interest: Tesla's own work on autonomous driving would compete with OpenAI for the same scarce talent, and indeed Tesla had already hired away the co-founder Andrej Karpathy to run Autopilot. That account was true as far as it went. What it omitted, revealed only later through reporting and then through litigation, was the power struggle underneath. Musk had concluded that OpenAI was losing to Google and proposed a remedy: he would take control himself, or fold the lab into Tesla. The other founders refused. Internal emails from 2017 show Brockman and Sutskever telling Musk, with evident alarm, that he had made plain "absolute control is extremely important" to him, and worrying aloud that he might become an "AGI dictator." Rebuffed, Musk walked away — and the funding he had promised to continue providing simply stopped. His reneging would later be cited as one of the very pressures that pushed OpenAI toward the thing he would one day sue it for becoming.
That thing arrived in March 2019, when OpenAI announced a restructuring into what it called a "capped-profit" company. Building artificial general intelligence, it argued, would require billions for compute and talent that no nonprofit could raise, and no existing legal form struck the right balance; the compromise capped investor returns at one hundred times their stake, with the nonprofit board retaining control and the charter's mission formally taking priority over profit. Altman, freshly departed as president of Y Combinator, became chief executive. Four months later, in July 2019, Microsoft invested $1 billion — this time real, called-down cash rather than a pledge — and became OpenAI's exclusive cloud provider. Further billions followed, culminating in a roughly $10 billion commitment in early 2023. The institution conceived as a counterweight to corporate AI was now bound to one of the largest corporations on earth.
The contradictions came due in November 2023, in a crisis that lasted about five days — not the ninety-six hours of later legend, but a span running from a Friday to the following Tuesday or Wednesday. On Friday, November 17, the board — Sutskever, Adam D'Angelo, Tasha McCauley, and Helen Toner — informed Altman over a video call that he was fired. The public statement said he "was not consistently candid in his communications with the board," and that the board had lost confidence in his ability to lead. Brockman was removed from the board and resigned that evening. Over the weekend, investors led by Microsoft pressed for Altman's reinstatement; the board instead installed the former Twitch chief Emmett Shear as interim CEO, and Microsoft announced it would simply hire Altman and Brockman to run a new in-house lab. Then the staff revolted. By Monday, some 738 of roughly 770 employees had signed a letter threatening to follow Altman to Microsoft unless the board resigned. Sutskever, who had voted to fire Altman, signed it too, and posted publicly that he deeply regretted his participation in the board's actions and would do everything he could to reunite the company. Within days Altman was back, atop a new board chaired by Bret Taylor; Sutskever, Toner, and McCauley were gone.
Why it happened remains genuinely contested, and honesty requires keeping the tiers of certainty distinct. The best-attested account is a breakdown of trust: Toner would later say that Altman had for years made the board's work difficult by withholding information and, in some cases, "outright lying," and Sutskever's subsequent deposition referenced a memo accusing Altman of a pattern of pitting his executives against one another. A dispute over a paper Toner had co-authored, which Altman felt criticized OpenAI, was a documented trigger. The most lurid theory — that a secret mathematical breakthrough nicknamed Q* had spooked the board — belongs in a lower tier; the outlet that first reported it could not verify the claim and walked back its own framing, and the independent review later commissioned by the new board concluded the firing "was not based on concerns regarding product safety or security," nor on finances, but on the collapse of trust between Altman and the directors who removed him.
Sutskever left for good in May 2024 and, that June, launched Safe Superintelligence Inc., describing it as the world's first "straight-shot" lab pursuing a single goal. Musk, meanwhile, took the founding mission to court. He sued in early 2024, withdrew the suit, and refiled in federal court, alleging that Altman and Brockman had induced his donations with promises that OpenAI would remain a nonprofit for the benefit of humanity and had then betrayed them — that they had, in his telling, stolen a charity. The case grew into perhaps the most expensive feud in the industry's history, with a damages expert floating figures as high as $134 billion. And then, on May 18, 2026, after a three-week trial in Oakland, the jury took less than two hours to decide it — and decided it on the narrowest possible ground. Musk, they found, had sued too late; his claims were barred by the statute of limitations. The merits of whether OpenAI had betrayed its founding purpose were never reached. Musk called the outcome a "calendar technicality," insisted the only open question was when the betrayal had occurred, and vowed to appeal to the Ninth Circuit; OpenAI's lead lawyer called the suit a "hypocritical attempt to sabotage a competitor." The deepest irony was structural. The question the whole saga had circled since that first December blog post — whether an institution built to keep artificial intelligence unconstrained by the need for financial return had become precisely what it set out to prevent — was the one question the jury, having ruled it raised too late, declined to answer.
=> The Schism