After reading more and reflecting over the past week or so, I think the most obvious mistakes in my original narrative are:
- Sam probably didn’t convene the board to oust Helen, since he wouldn’t have wanted things to be so overt. Instead, his side could have at any time leaked Helen’s report to the press, manufacturing a crisis leading to Helen’s dismissal. This explains the time pressure that Helen, Tasha and Adam were under. For more on this, I recommend Gwern’s commentary ([1](https://www.lesswrong.com/posts/KXHMCH7wCxrvKsJyn/openai-facts-from-a-weekend?commentId=E7rEuPdmsSKSYWnPz), [2](https://www.lesswrong.com/posts/KXHMCH7wCxrvKsJyn/openai-facts-from-a-weekend?commentId=3cj6qhSRt4HoBLpC7), [3](https://www.lesswrong.com/posts/ANStTHjj6it8ysaRJ/why-did-openai-employees-sign?commentId=K2KTqZBHNN6a582yp)).
- As Gwern explains, Ilya was in a Slack channel with Sam where there was discussion of getting rid of Helen because of her EA connections, and he also witnessed (perhaps via Helen) Sam’s attempt to oust Helen on completely different pretenses. So the situation was probably very transparent to Ilya (maybe Sam didn’t even realize he was in the Slack channel? that would have been quite the error), which helps explain why he was so steadfast in his initial explanations to employees. This also suggests that the board did in fact have pretty hard evidence against Sam, but not evidence that they could share without fear of repercussions for sharing private conversations.
- My original narrative paints Helen, Tasha and Adam as having planned everything out in advance, whereas in practice I am sure there were responding to a rapidly-evolving situation like everyone else. I think it’s possible that they did ultimately want Sam gone as CEO, not just gone from the board. That being said, they must have expected Sam to fight back, and if they did want Sam completely gone then they must have known that they might have to settle for less. Regardless, “Helen, Tasha and Adam won” is perhaps not the best summary.

- Nevertheless, I still think it’s reasonable to entertain the possibility that they did ultimately want to keep Sam as CEO. “Why would the not just remove Sam and Greg from the board, without firing Sam as CEO?” you might ask. But what would have happened if they did that? Sam would still have fought back, painted the situation as unreasonable, perhaps even leaked Helen’s report to the press, and he would surely have ended up with concessions. People would have said “why didn’t you fire him as CEO if what he did was so bad?”. I think Helen, Tasha and Adam knew that Sam had all the soft power; their only realistic way to fight back was to make the best use of their hard power that they could, by firing Sam and putting him in a tougher negotiating position. They ultimately reinstated him as CEO, but in doing so they got concessions from Sam that prevented him from further resisting his lost board seat.
- Everything is not over yet: the board managed to get Sam to agree to an internal investigation over his actions, so he might still face further repercussions for them, possibly even up to being removed as CEO again. Of course, if this happens, all the drama will behind closed doors, since the new, independent board won’t be under the same imminent pressure to act. “Why wouldn’t Sam just go to Microsoft again?” you might ask. But if the board is able to present its hard evidence to the exec team, it might be able to persuade them. With them on board, I think the employee situation looks very different, and far fewer people threaten leave for Microsoft. Of course, this would all just be a hypothetical entertained in private negotiations. And after everything is said and done, my guess is that Sam will still have enough people on his side, and enough other negotiating leverage, to avoid losing his position again. Whatever goes down, if there are any externally-facing changes, these will surely be presented to the public as a “done deal”, with enough concessions to Sam that he still comes out looking like he was happy about everything.
- There’s a lot of speculation that this had anything to do with Q*. I really think that’s very unlikely. OpenAI makes internal breakthroughs all the time, and likes to hype them up. Don’t get me wrong, I’m sure Q* is a great piece of research, but these sorts of breakthroughs are always incremental. Ilya would have been aware of the research the whole time and would have if anything been supporting it.