The Decade-Old Feud That's Deciding AI's Future
TL;DR
The OpenAI–Anthropic split started as a deeply personal fight, not just a philosophical one — the Wall Street Journal traces the feud back to 2016–2017, when Dario Amodei saw Elon Musk’s layoffs as cruel and reportedly called Greg Brockman’s idea of selling AGI to governments including China and Russia “treason.”
Sam Altman’s rise didn’t calm things down; it made the internal power struggle worse — after taking over in 2018, Altman allegedly made conflicting promises to Dario Amodei, Greg Brockman, and Ilya Sutskever, while tensions peaked around who would control the GPT research that became OpenAI’s core advantage.
The GPT era was also the breaking point — as Alec Radford’s transformer-based work turned into GPT-2 and GPT-3, Dario blocked Brockman from the language model effort, and Daniela Amodei reportedly offered to step down rather than let him onto the project.
Anthropic’s founding was framed as a values split with real structure behind it — when Dario, Daniela, and nearly a dozen employees left in late 2020, Dario wrote that the ideal AI company should be “75% public good and 25% good for the market,” a direct contrast to OpenAI’s drift toward a more commercial model.
This feud now matters because both companies sit in the tiny group of labs deciding AI’s direction — the hosts argue the real U.S. frontier today is Google DeepMind, OpenAI, and Anthropic, with Meta and xAI as tier-two players, meaning two of the most important labs are effectively “at war” while racing toward IPOs and government contracts.
Politics is now part of the competitive map, not a side issue — Brockman’s reported $25 million pro-Trump super PAC donation, Anthropic’s battles over government access, and OpenAI’s push for contracts all suggest these labs aren’t just building models, they’re positioning for power in Washington as much as in enterprise.
The Breakdown
A Wall Street Journal scoop turns old rumors into a real origin story
The episode starts with a fresh Wall Street Journal investigation that stitches together a decade of bad blood between OpenAI and Anthropic. The hosts frame it bluntly: this isn’t just a disagreement over AI safety or philosophy — it’s a feud, rooted in personal wounds, status, and power, with consequences that now reach into enterprise sales, IPO timing, and government contracts.
Dario’s early OpenAI years were already full of red flags
Going back to 2016, Dario Amodei joins OpenAI and quickly sees two things that deeply unsettle him: Elon Musk ordering layoffs in a way he considered cruel, and Greg Brockman floating the idea of selling AGI to nuclear powers on the UN Security Council. The hosts linger on the most explosive detail — during a 2017 ethics-and-policy discussion about government involvement, Brockman reportedly entertained selling AGI to governments including China and Russia, and Dario saw that as “treason.”
The nonprofit dream starts cracking after Musk exits
Once Musk leaves in 2018 and Sam Altman takes over, the company moves further toward the for-profit path that would later define it. The hosts tie this moment to the larger Musk–Altman split that’s still heading to trial, but they emphasize that inside OpenAI, the more immediate issue was who would actually control the future of the lab.
GPT becomes the battlefield, and Brockman is the recurring flashpoint
Alec Radford’s early work on transformers and generative pre-trained transformers becomes the hinge point for everything. As OpenAI realizes language models may be the real breakthrough, Brockman wants in — but Dario, then research director, reportedly blocks him, and Daniela Amodei, co-leading the project with Radford, even offers to step down rather than let Brockman join. The hosts present Brockman as the name that keeps showing up “over and over and over again” whenever things blow up.
Recognition, access, and leadership all become part of the rupture
The personal dimension sharpens when Altman and Brockman go meet former President Barack Obama and Dario is left out, despite his growing role in scaling laws and the GPT effort. That exclusion, plus conflicting leadership promises from Altman, helps push Dario toward an ultimatum: either report directly to the board or leave. By late 2020, Dario, Daniela, and nearly a dozen others do leave and form Anthropic, carrying with them a memo that argued an AI company should be 75% public good and 25% market-driven.
The old feud is now colliding with today’s business reality
The hosts connect the history to current pressure: OpenAI recently shut down its Sora video app after it was reportedly burning $1 million a day and falling to just under 500,000 users, while Anthropic is gaining ground in enterprise. They say company leaders they talk to are “very very” commonly moving toward Anthropic, and even OpenAI applications head Fidji Simo reportedly described Anthropic’s enterprise momentum as a wake-up call.
Why this matters: a handful of labs now control the board
The episode zooms out into a map of frontier AI. In their view, the top U.S. labs right now are Google DeepMind, OpenAI, and Anthropic, with Meta and xAI as second-tier wild cards; the traits that define a true frontier lab are funding, data centers, energy, Nvidia chips, and top-tier models. That makes the human drama feel less like gossip and more like governance by personality, because these are the people deciding the economy, geopolitics, and the path to things like agents, reasoning, recursive self-improvement, and world models.
Politics is the last layer, and maybe the messiest one
The hosts finish by arguing that political alignment could become a strategic liability for AI labs. Google is portrayed as trying to stay neutral, while OpenAI now carries the baggage of Brockman’s reported $25 million pro-Trump super PAC donation, Anthropic is described as left-of-center and currently at odds with the administration, and Meta and xAI are seen as fully aligned with Trump. Their takeaway is simple: if power in Washington flips, the contract landscape could flip too — and that makes understanding who these builders are, and what they actually want, impossible to ignore.