Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
Back to Podcast Digest
The Musk v. OpenAI Trial Just Got Even Messier
TL;DR
Musk tried to settle right before trial, then sent a threat-like text — OpenAI says Elon Musk texted Greg Brockman two days before trial about settlement, and when Brockman suggested both sides drop their suits, Musk replied, “By the end of this week, you and Sam will be the most hated men in America.”
The trial is exposing a much bigger 2017-2018 power struggle than most people knew — testimony from Greg Brockman and Siobhan Zilis suggests Musk wanted OpenAI merged into Tesla, offered Sam Altman a Tesla board seat, and even explored recruiting Demis Hassabis and Andrej Karpathy into a Tesla AI lab.
Siobhan Zilis’s role came off like a political thriller subplot — the hosts describe her as effectively an intermediary and later a “plant” on OpenAI’s board, with her relationship to Musk undisclosed at the time and only surfacing publicly after a 2022 Business Insider report.
Mira Murati’s deposition revived the 2023 Altman firing drama in brutal detail — Murati testified Altman was “not always” candid, said he undermined her and pitted executives against each other, and new texts showed Altman asking her in real time what was happening as the board moved to remove him.
One line from former board member Helen Toner captured the whole OpenAI coup saga — describing Murati’s role in Altman’s ouster, Toner said Murati was “waiting to see which way the wind would blow and she didn’t realize that she was the wind.”
Microsoft’s internal emails reveal it nearly backed OpenAI out of fear, not conviction — executives worried OpenAI would “storm off to Amazon and talk Microsoft and Azure on the way out,” were spooked by a proposed $300 million Azure-scale ask, and only later shifted after OpenAI’s work moved from Dota 2 toward language models.
The Breakdown
The settlement text that made everything feel darker
The episode opens on the wildest new reveal: OpenAI says Musk texted Greg Brockman just two days before trial to feel out a settlement. When Brockman floated the idea that both sides should just drop their suits, Musk allegedly shot back, “By the end of this week, you and Sam will be the most hated men in America” — a line the judge later ruled inadmissible because it was introduced too late.
Stuart Russell brings AGI risk into the courtroom
Musk’s lone AI expert witness, UC Berkeley’s Stuart Russell, told jurors there’s a real tension between building AGI and doing it safely. He walked through the full menu of risks — misalignment, cybersecurity, discrimination, job loss, and even people getting emotionally attached to AI — giving Musk’s case a broader safety frame beyond just corporate infighting.
Brockman’s testimony turns early OpenAI into a Mars-Tesla power story
Greg Brockman pushed back on Musk’s version of OpenAI history and said Musk wanted majority control, partly tied to funding ambitions around SpaceX and his Mars plans. Brockman also alleged OpenAI staff secretly spent months helping Tesla’s Autopilot effort in 2017 while Musk was still publicly presenting OpenAI as a charity, and Musk’s lawyers countered by hammering Brockman’s now-massive stake in the restructured company, reportedly worth around $30 billion.
The Tesla AI lab plot twist nobody saw coming
What really stunned the hosts was evidence that Tesla had active 2017 plans to build an AI lab to rival Google DeepMind. According to testimony and reporting from Max Zeff, the brainstorms included getting Sam Altman to run Tesla AI, recruiting Demis Hassabis, and even using NeurIPS as a staged launch moment with Altman positioned to effectively announce it onstage.
Siobhan Zilis becomes the trial’s most surreal character
Zilis testified she spent years acting as a go-between for Musk and OpenAI leadership, and said Musk wanted OpenAI folded into Tesla and offered Altman a Tesla board seat. The hosts then go deeper into the backstory: Zilis later joined OpenAI’s board in 2020 without disclosing her relationship with Musk, which they describe as turning her into an “operative” keeping him informed until the relationship became public and she eventually resigned.
Mira Murati reopens the Sam Altman firing wound
Video testimony from former CTO Mira Murati was brutal: she said Altman would tell different people opposite things, lied to her about whether a model needed deployment safety review, and was “not always” truthful. The hosts connect that to the 2023 board coup, including reported texts where Altman, not knowing Murati had helped drive the process, begged her for clarity while she told him things were “directionally very bad.”
“She didn’t realize that she was the wind”
Former board member Helen Toner’s deposition delivered the line of the week. Talking about Murati’s role in Altman’s ouster, Toner said Murati was “waiting to see which way the wind would blow and she didn’t realize that she was the wind,” which the hosts basically treat as the perfect poetic summary of the whole messy power struggle.
Microsoft’s OpenAI bet looked way shakier inside than outside
The last major reveal is from internal Microsoft emails entered into evidence: after OpenAI’s Dota 2 success, Altman asked for something like $300 million in Azure list-price compute, and Microsoft executives balked. Kevin Scott worried OpenAI could “storm off to Amazon,” Jason Zander questioned the economics, and only later — once OpenAI shifted from game-playing toward natural language models — did Microsoft move toward the $1 billion investment that now looks historically inevitable but clearly didn’t feel that way at the time.
Share
Keep Reading
Two Ways to Follow
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.
