Special Episode: A Full Breakdown of the OpenAI Saga and What It Means for AI Founders

In light of one of the biggest news stories in AI, we’ve put together a special episode to discuss the ramifications of Sam Altmans’ firing from OpenAI. Regardless of what happens between now and when you’re listening to this, the implications of the events that happened over the past few days are certainly worth unpacking.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Jacob Effron?

  This is Unsupervised Learning, Redpoint AI Podcast. In light of one of the biggest news stories in AI we've seen, we have a special episode here to discuss the ramifications of Sam Altman's firing from OpenAI. Regardless of what happens between now and five minutes from now when you're listening to this, the implications are clearly going to reverberate through the ecosystem. So we brought in Alex Conrad, a journalist at Forbes who covers venture capital, and Jason Warner, the former CTO at GitHub, partner at Redpoint, and now the founder of CodeGEM startup Poolside. And specifically, we go over what the implications are for AI startups, who came away as the biggest winners, what's going on behind the scenes and more. I think folks are really going to enjoy this.

• 0:38Jacob Effron?

  Well, I think to start, Alex, you've obviously been living this nonstop since Friday at, what, at 12:30 PT. Maybe it would just be helpful to kind of level set on what happened? What has happened in the last, you know, in the last five days? Walk us through the unfolding here.

• 0:55Speaker 01

  Good thing this is a 10-hour podcast, right?

• 0:57Jacob Effron?

  Exactly, exactly. The mini-series will eventually come out, but for now, we need, you know, just the base facts.

• 1:04Speaker 01

  Yeah, I mean, the broad strokes is that right around market close on Friday, the Board of Directors of OpenAI announced they were firing Sam Altman, who I'm sure the audience has heard of, you know, arguably the most famous CEO in the AI world. And no offense, Jason, maybe you'll be next year. Hey, listen, if I'm number two to Sam, I'm doing something right. I'm fine. I'm good. And then Greg Brockman, the co-founder and president, announced he was resigning. Over the weekend, there has been lots of turmoil. There was discussion of Microsoft investors, employees pushing for Sam to be returned at OpenAI. Presumably, that would then force the board to depart.

• 1:46Speaker 01

  Sam's co-founder, Ilya Sutskever, the chief scientist, was part of the board that had removed Sam. Then Ilya had a change of heart and tweeted that he no longer supported the board's actions. So we spent Sunday wondering if Sam and Greg would return. They did not return, and Satya Nadella, the CEO of Microsoft, announced that he was going to be hiring them for a new AI research unit. But then it looked like maybe Sam and Greg were still going back. Ninety-five percent of employees at OpenAI signed a petition saying they would resign if the board didn't. And Satya went on TV to basically hedge that, hey, maybe Sam and Greg aren't going to be employees, but they'll be working with Microsoft anyway, even at OpenAI.

• 2:30Speaker 01

  And so for the last couple of days, we've sort of been in this strange limbo of will they or won't they? The world's least funny sitcom happening right now.

• 2:39Jacob Effron?

  It's funny. I'm struck by, as you actually just told it from start to finish, like, it feels like even more ridiculous than a ChatGPT-generated, like, you know, fantasy story of what might have happened here. I mean, what a tale of twists and turns. What has life been like for you these last five days? What happens in a reporter's life that, like, went, you know, 3:30 Eastern Time for you, when I'm sure you had nice Thanksgiving plans all lined up, and then something like this happens?

• 3:02Speaker 01

  Yeah, you know, I had all these dreams of, like, going to the gym a lot this week and catching up on what had been a pretty crazy, you know, few weeks before that. And then on Friday, I'm walking out of a meeting and I see the news, and I thought it was a joke. And it wasn't a joke. And so it was like, here we go. And I basically just, like, strapped in and didn't leave my desk much since Friday at, like, 4:00 PM. So it's crazy. Basically, you are just messaging everyone you know. Every VC you know is texting you to be like, what's going on? Which is not really that helpful. Like, maybe I just know the wrong people, but I was like, you're supposed to be telling me what's going on?

• 3:44Speaker 01

  Which is funny. And just lots of group chats, crazy. I mean, I think it's very similar to VCs in the space, like, their phones were blowing up. But the difference being that people like Jason don't feel professionally bad at their jobs every time one of a million other people share something on the story constantly. So I'm following it like everyone else, but also with the sense of, like, I should be somehow magically breaking every story here. And I've broken a few, and I'm proud of that, and so is my team at Forbes. But there are a million pieces of this that we just see along with everyone else, and we're like, what? What the heck? And and we're just, you know, speechless.

• 4:23Jacob Effron?

  Well, I mean, I think, you know, one thing I want to make, you know, sure we hit on is just kind of what this all means for, you know, we have, obviously, we've talked about on this podcast, a ton of businesses that are building AI, AI applications, a ton of startups that are competing with OpenAI. It feels like this was a pretty, you know, monumental event in how both builders and startups are going to be thinking about the landscape going forward. I think Retool came out with this report of where they'd asked a ton of enterprises what they use and like over 80% were like, you know, hey, we we use OpenAI, that's the main model we use. And I think kind of foots with what generally I felt what we see is a lot of folks were just, they were starting with OpenAI, right? They were like, let's see if this use case is possible. I'll take the the state of the art, the best model that's out there in in GPT four and see if I can make this work. And maybe down the line, I'll figure something out around, you know, open source models or other players.

• 5:08Jacob Effron?

  But it seemed like that was a later thing. And really, from at least what we have been seeing, you know, almost everybody was getting started on on the OpenAI side. All these enterprises had rolled out these features, right? And they're they're sitting this weekend like, what is going to happen to this feature that is live with all these customers? Alex, I think you wrote a story about this, right? Like, would love to hear some of the conversations you had.

• 5:28Speaker 01

  We reached out immediately to a bunch of customers and, you know, I I was soliciting founders to message me kind of how they were thinking about using this. And big surprise, we kind of saw the full spectrum of replies. You know, one founder told us at Forbes, like, you know, building on top of GPT is dead, like, you know, it's over. Someone else, I think, was kind of more thoughtful to to Jason's point, shout out. I had a long Twitter exchange with a founder named Tomas Puig, who has a startup called Alembic. I just met met him on Twitter, so hopefully he's he's a, you know, okay guy. But he was sharing some really interesting economics here. He just broke down the cost to basically shift off of GPT.

• 6:12Speaker 01

  And he was saying it would basically double his cost to switch to Llama, one of the big open source alternatives. You know, I think eight dollars, basically eight dollars an hour for for server requests, which at scale becomes really a lot of money. And I I think, you know, then you saw a lot of people dunking and being like, of course, everybody's using multiple things. Like, imagine a company building on only one of these models, which, you know, if you're like really tech savvy and maybe you're like a cutting edge startups like that.

• 6:45Speaker 02

  That doesn't happen. I like, no matter what people say here, that's the hybrid cloud story of this space. Everyone touts hybrid cloud. That's not a thing. Hybrid cloud is not.

• 6:54Speaker 01

  Jason, have you seen this too, though? I've seen a bunch of people be like, imagine putting all your eggs in one basket with OpenAI. And it's like, come on, a lot of people do that, you know?

• 7:01Speaker 02

  That's literally the Twitter. That's a Twitter persona who just never built anything at scale of substance, of real maturity. Like armchair quarterbacking somebody else's startup decisions. And anyway, I go to a rabbit hole on this one, but no, no one's doing that. But that person was right, you know, OpenAI is a quick, easy way to get up and running on stuff. And it's the state of the art model. Nothing else is as good as it. And you get it for cheaper than running your own versions of these other things. And yes, at some point, you might make a different decision to self-host or to build your own model, which is really where people should be going in the future in this way. But in general, no, like they're going to use OpenAI.

• 7:41Speaker 02

  And so on the flip side, if you're a startup and you're wrapping OpenAI, this is, this is a little bit of a, oh shit, what do we do now? What happens in a scenario like this? And so for instance, of the several scenarios that are possible, you know, they've got to have some sort of contingency plan in case. You know, Sam and Greg and, you know, 770 out of however, number of people go over to Microsoft at some point, they got to have some sort of contingency plan in place. And the answer to that is not going to be just to go use XYZ off the shelf because it's not, it's just not that simple.

• 8:12Speaker 01

  One, one thing I heard from a couple of founders, and I'm curious if you guys think it carries any water is they said, I'm not worried right now about using OpenAI because it's backstopped by Microsoft. But if everybody really did move over to Microsoft, would that backstopping be a real thing? Or at what point would we see major degradation? Because I think ChatGPT has already been down for some users, you know, today and yesterday, at least it was for me.

• 8:39Speaker 02

  My view here, if I was CTO in any one of these organizations, I would be sitting there and I would be coming up with contingency plans. If you're actually running a business, not if you're in the, hey, we're building the business, we're pre-launched before we got to customers phase. But if you've got a running business, you need to be coming up with other plans here. You can't be worried, you can't be relying on the fact that it'll stay up because they'll figure this out or Microsoft will backstop it. Even if the only thing that you're doing is using the Microsoft hosted version of these things as your contingency. You've got to be coming up with something. If you're not, you're not doing your job.

• 9:11Jacob Effron?

  And I think a lot of these companies are asking, there's like the short term where it's like, will this go down like tomorrow? If people don't go back, you know, or aren't working on maintaining these things. But then there's like the longer term, like it felt like OpenAI was just so ahead in terms of the developer experience they were providing, the reliability they were providing. It felt safe to just kind of like hang your hat on what they were doing. And so I do think there's like, you know, now that halo is a little challenged, right? There is definitely going to have to be some reassuring, even if everybody comes back. Like, are we still all full speed ahead on something that you could kind of just like, you know, that you really were able to trust?

• 9:41Speaker 01

  Can we also point out, like, less than, what, it was a week and a half after OpenAI had its debut developer day? Which seemed like such a flex and it seemed like, you know, they were on that progression to handholding customers and like helping them build custom GPTs. And like knowing how to manage that experience really well. And it seemed like the community, from what I could tell, seemed very favorably reacting to that dev day. And then it's like a week and a half later, should I use this at all? The whiplash is crazy here.

• 10:16Jacob Effron?

  I mean, everything OpenAI was doing was a flex. It's wild. Like, they were just on such a hot streak.

• 10:22Speaker 02

  Well, I think there's two different things here too that are important, right? So one of them is going to be the startups that are building on OpenAI. So that's one category of things. Then there's the enterprises that are building on OpenAI and kind of trusting them. So it doesn't matter the scenario. We're on this podcast, we don't know what's about to happen. We're talking here, we're not refreshing Twitter, we don't know what's happening in the moment. We don't know what's going to happen in an hour from now. But no matter what happens here, let's say they stay at Microsoft or Sam and Greg and everyone goes back at OpenAI and they come up with governance structures. Microsoft, Satya and Sam will have to be going to the enterprises and trying to reestablish in trust. Like this has got to be a thing. The largest enterprises in the world are going to be asking governance questions. They're going to be asking like, you know, what happens? Microsoft, do you have access to these things? Do you not have access to

• 11:06Speaker 02

  these things? What's going to happen in a future scenario like this? And if they can't answer those questions credibly, enterprises are rightfully going to be suspect of what's going on. And they're going to have to be able to answer those questions. And it's nothing that they're going to be able to avoid.

• 11:20Jacob Effron?

  My feeling is there's like a lot of companies that were using OpenAI and they said, okay, at some point we'll figure out other models to use too, or at some point we may be building our own open source models and we'll be able to switch between them. But it was this future question, and I wonder if this becomes a catalyst to say, okay, now is the time that we really need to figure this out because we're going to get questions, right? Like, you know, if you're serving a product to your customers, those customers are going to start asking questions, just given everything that's unfolded these past five days. And so it does feel like a, you know, a pretty big moment on the enterprise side. I also want to hit the startup world. I mean, I think, you know, back to OpenAI being on this huge, just like hot streak, you know, the kind of breadth of what they were doing was just, you know, it was amazing, right? They were competing, I feel like with

• 12:02Jacob Effron?

  every, every startup out there and a ton of different things, whether it was the consumer products and chat GPT, the, you know, core LLMs, they obviously had text to speech, their Dolly. There's no shortage of things they were building. And I feel like if you were a startup, you know, Friday afternoon, you know, your, your odds just got a little bit better, right? There's a little bit of, of weakness in what otherwise was just this incredible, incredible story. I don't know, Jason, if you have any thoughts on that?

• 12:29Speaker 02

  Yeah, two things. I think if you ended up, if you were an application using AI, and obviously, if you were directly competing with OpenAI and any sort of thing, obviously there's an angle now for you. But the question is, still, who is your model? Like, are you using OpenAI? Or what's your, what's your model answer here? So this is, goes directly to your point. Now, on the flip side, let's go down to the model providers and, you know, pool size and the model provider where model plus a platform plus some applications built on top of our own version of this. But let's just say that my all my entire inbox or my chats, texts, whatever from VCs was, this is good for you, right? This is good for you, right? This is good for you, right? Literally, that's all it was.

• 13:09Jacob Effron?

  We always say it's an industry of original thinking, right?

• 13:11Speaker 02

  This is true. So like at two different layers, you have two different reactions. So at the at the foundation model layer, we don't directly compete with OpenAI. We're going a slightly different path, but even so, it's hard to disambiguate in the moment because if you don't have a full understanding and nuance of the differences, it's going to be difficult for anyone in the world to understand that. But in general, the reaction was the landscape's wide open again. You know, if this is as rocky and faulty as it's going to be, there's massive opportunity out there for folks. On the application side, though, again, you have different sets of questions to answer. It's, yeah, there's opportunity possibly for you here, but you've got to shore up your relationship with the intelligence layer.

• 13:47Jacob Effron?

  Yeah, and you even saw it in startup CEOs, right, posting, you know, I think across the service area that OpenAI covers, like you have perplexity, obviously, which, you know, in many ways is a chat GPT competitor. And, you know, that's, that's something that that could be really interesting, given what's happening. You know, you have obviously the foundation model companies trying to hire a lot of these folks this weekend and also fielding a lot of inbounds from enterprises. But even, you know, like 11 Labs and text to speech or an ideogram and image, like there's so many different companies that do compete with parts of OpenAI. And I think this, this definitely changes the playing field around that a bit.

• 14:20Speaker 01

  Oh, my God, you guys, I've actually been checking Twitter this whole time, and it looks like Sam is back.

• 14:27Speaker 02

  Yeah, I mean, we knew that was coming. That was the direction this was coming, right? Like, if

• 14:31Speaker 01

  I'm just, I'm just messing with you. I'm just messing with you.

• 14:33Speaker 02

  Oh, God, come on.

• 14:37Speaker 02

  I would literally put money on this at this point because like I feel that this is the direction it has to go. Like, if you're sitting here looking, you know, I say that, but this is not a rational actor territory type of situation all of a sudden. But it just looks like he's, he would, he and Greg and everyone's going to end up back over there with a new governance structure in Microsoft. Like, it's better for Microsoft. Satya is going to come out looking good no matter what because Satya is good at this and he looks like he's going to have a coup. But like, realistically, too, Microsoft has some questions to answer about $10 billion investment without governance structures. But in reality, it's better for them to have an independent OpenAI. Microsoft with an independent OpenAI with a relationship there is a much better place than for them to be inside of Microsoft. It just really is.

• 15:17Speaker 01

  So a question for the, for the VC and the former VC.

• 15:22Speaker 01

  Which is the, what's the best outcome for startups? I mean, Jacob was just saying that maybe the playing field got leveled a bit this weekend. I don't know if it's more level with Sam at Microsoft and not back at OpenAI.

• 15:35Speaker 01

  I don't know if it's better for the startup ecosystem theoretically if the OpenAI people had scattered to the winds and done their own companies or joined everybody else's startups. Curious what you guys think in terms of just sort of that macro ecosystem, what, what the different outcomes would, would play out like.

• 15:50Jacob Effron?

  You know, people have been talking a lot about where value accrues in AI and like where the default is. And I think a lot of people think this is something that's just going to help incumbents, right? And so I think what I was struck by, I guess, in your question is, I thought the default for how this could have played out is that like Microsoft, Amazon, Google, Databricks, Snowflake, like they just are the place that everyone turns to, to go and, and use some of these services. And in many ways, OpenAI was like the example of the company that wasn't those folks, that net new company. And obviously they did, they had a big Microsoft relationship. A lot of people were buying it through Azure, but it's still, a lot of people were going to OpenAI directly. And I think there's a lot of trust that's built up, you know, that they had built up that the whole ecosystem benefits from. That you don't just need to go to your cloud provider to, to get all of these services that like net new companies can provide some of this.

• 16:31Jacob Effron?

  And so I think, you know, an independent OpenAI obviously is a formidable competitor for anybody, but also like creates a precedent of, you know, working with, of, of going outside of like the, the vendors you might typically go to for that. I don't know, Jason, what do you think?

• 16:47Speaker 02

  So I think no matter what, the past, whatever hours that we've been in on this at this point is good for startups. Because what it effectively does, did is it introduced ambiguity into a situation that everyone thought there was clarity. And it's anything like VCs really like is like saying, oh no, this is clearly this, this entity over here. But now this shook confidence, no matter what, no matter what happens now, there's got to be shaken confidence. It also introduced the fact that there can be different angles that coming into here. Now of the two, you know, this being inside Microsoft is better from a competitive standpoint for everybody else in reality, because there's different angles to be played.

• 17:27Speaker 02

  It allows Amazon, it allows Google, it allows third-party mega enterprise vendors to actually have a narrative now, interestingly. The, on the flip side, you know, if it were to go back to OpenAI, I think, as I said before, I still think that's a massive gift that was given to all the startups in the space.

• 17:47Speaker 02

  And, but it's going to manifest differently over time. And so the model provider ones, they have more of a gift. The application providers, they have less of a gift, but they need to react sooner to it. It's kind of a weird, weird economy here. I will say one thing about this, though, is that of all the things that I said over the weekend or kind of reacted to here is the one thing I will say is this makes Microsoft look great on a lot of fronts and every other hyperscaler out there look terrible. And the reason why I say this is Microsoft has been, had this relationship. Questions about their governance structure with their investment aside, they reacted, they moved, everybody else was sitting there. And to me, this is actually the most interesting and fascinating part about all of this is,

• 18:31Speaker 02

  you know, I'm building a startup, et cetera, I'm fascinated by this. Where the hell is Google? Where is Amazon? Oracle, what are they doing? You know, you know, you don't need to get started on IBM.

• 18:41Speaker 01

  And to their credit, by the way, it was Salesforce. You could argue Marc Benioff has been as vocal as almost anybody. I don't know if you guys saw the exchange he had on X slash Twitter, but he, you know, he co-tweeted an employee and was like, you know, come join the team that has Slack and Tableau and all this and build Salesforce Einstein GPT, blah, blah, we'll match any salary. And the guy wrote back, sorry, but I don't think the, the company that runs Tableau is going to build an AGI, which was a pretty brutal burn, but like I give Benioff credit for being out there. I completely agree with you, Jason. Where is Google? Where is Sundar? The fact that they've been so quiet is crazy to me.

• 19:24Speaker 02

  Well, so as a person in the space too, it's also pretty clear that nobody in those areas has an answer to this, which, you know, to Satya's credit, he was able to go to Microsoft Research and say.

• 19:37Speaker 02

  If you could, why aren't you? I'm gonna go do this thing over here, whereas everybody else is still saying, we can. The answer is you can't. And it's not because you don't have the smartest people in the world. You have amazingly smart people. It's your organizational structure is not allowing you to do this. This is the advantage that Microsoft had. And this is mind boggling because they're doing the exact same thing all over again at all these places. We're gonna do it. We got four different internal teams building these, these models. One of them's gonna win. Yeah, three years from now, when you're, you know, everybody at this, the, the top of that company's fired. You know, that's, it, it boggles my mind. It really boggles my mind. If we talk about value destruction on one side of the fence with OpenAI possibly going out of business.

• 20:18Speaker 02

  Talk about, talk about what's happening with those hyperscalers and the fact that Microsoft is outside of the US government, probably the most important entity in the world in five years, if they continued on this path. Microsoft is just, they're killing it on all fronts.

• 20:30Jacob Effron?

  Jason, you obviously were, uh, were, you know, acquired into Microsoft before. Like, do you think it, like, could they have, I mean, who knows where Sam ends up? If they had ended up in Microsoft, like, do you think they could have run this thing and, and, and actually continued moving at the pace they've been moving?

• 20:44Speaker 02

  I mean, they're not gonna let, they're not gonna treat Sam like a normal employee. They're not gonna treat this division like a normal division. They're gonna do whatever they need to do to make it possible. So even when we joined Microsoft, you know, you've got to remember too, like, we, we, we didn't have a celebrity CEO, celebrity founder at GitHub. The founders are no longer involved in GitHub. It was Nat who came in from the Microsoft side, who became our CEO. Uh, and Nat, you know, Nat and I collaborated on this, but Nat took an extremely strong stance, which I was very appreciative of, which was, no, we're not integrating into Microsoft. We're using, we're still using G Suite, we're still using Slack, we're still using all those things. And, and, you know, that's the right move. You, you don't want to be integrated. You want to be your own separate entity.

• 21:25Speaker 02

  And now, over time, unfortunately, Microsoft forced integrations into GitHub, and that changed GitHub. So, you know, learning from the mistakes of the past, if they didn't do that to, you know, whatever division this thing is, ends up being called, I think they could be successful. That said, it's still Microsoft. There's still elements of the highly political Microsoft at play. There's still elements of can't help but be old Microsoft at play. But if anyone could probably resist it, it's Sam and that unit and how important it is. So I, I think they could be successful, but I still think independent OpenAI is a more important entity to Microsoft than a, a division of OpenAI.

• 22:05Jacob Effron?

  I'm struck by also, I mean, obviously, Google had such a collection of amazing researchers in DeepMind and Brain, and a lot of folks have had to spin out of there, right, to start companies because they were, in many ways, it's obviously a lot harder when Microsoft directly employs people in terms of the limits you have to put on what they can ship versus having this arms-length relationship. And so, I think we've already seen that movie play out almost six years ago or so with a lot of those folks leaving Google.

• 22:29Speaker 01

  Yeah, I mean, if you're the Microsoft AI team, are you rooting for Sam and Greg to go back to OpenAI because it's easier to deal with them as partners and arms-length than not ask comparisons in your annual review? That's one meta, I'm sure that's like question number 1,000 in terms of importance, but I do wonder for Microsoft employees, which way they would fall on that question.

• 22:55Speaker 02

  That's an interesting question. I thought my favorite tweet over the weekend was when Satya announced it, I forget who said it, but it was a simple one, I think he was in the .net group, said, "Hey, wait a minute, I thought we didn't have any headcount." It just killed me. It just killed me.

• 23:12Jacob Effron?

  I mean, Alex, one thing I'm curious about is it feels like, as we're talking here, and who knows what will happen by the time in between this getting recorded and released, but it feels like there's a sense of inevitability around Sam coming back at some point. And I do feel like the narrative has been pretty well-controlled over the last five days. Like, I'm just curious, what's happening behind the scenes? Like, in terms of, how do you think about, like, I feel like it's being run quite effectively as well on his end. What's your kind of take on their strategy and what they've been doing on the media side throughout this whole process?

• 23:46Speaker 01

  It's something I've been talking to people about today, the sort of meta story game here and the comms perspective. I think, one thing that's interesting to note is that boards of directors don't usually have their own comms person. They're usually using the company comms team, and if the company comms team is loyal to your CEO and your president slash chairman, you can't use them. So, in this case, the board of directors of OpenAI seem like very smart people. I don't really know them super well, besides having met Adam D'Angelo a few times, the Cora CEO, but I don't think they're comms experts. And I think that There are two main scenarios that I think are sort of most plausible,

• 24:30Speaker 01

  which I can't tell you the full answer to, which is one, something so bad happened or involving such trade secrets and technological something that they would rather fall on their swords and just be dragged publicly for days and let the narrative be all about Sam. Or they did not expect this backlash, this attention, and it's just a mess over there, a total mess. And so Occam's razor would probably say the mess, but yeah.

• 25:03Speaker 02

  I agree with you in that, for what it's worth. When all of us read that memo, I was in Europe when that came out. So it was evening for me and I was pretty tired and jet lagged and I'm reading this. All I could think was, wow, you don't say that. You don't say that in a press release about what's going on and you're letting go of your CEO. So you default to, oh, something went down, something egregious. And then when they had the second statement, which was, I believe, the Saturday, the next day, and it was backtracking on that, and then there was the internal statement, which was, I forget the wording of it, but it was a definite backtrack on this.

• 25:44Speaker 02

  You really start to say, somebody, somewhere is really messing up this comms. And I don't know, you don't know details, you don't know exactly what's happening internally, but you cannot release the statement you released publicly on Friday and then have these other statements that sort of basically walk it back. That comms experts know this, anyone who's been in C as a C-level knows this. So that was the, of all the things that happened outside of Sam being released and Greg quitting, those were two incredibly shocking things because they were jarring. That first statement was actually really jarring.

• 26:16Speaker 01

  Yeah, and I think that's a good point. And I think it's sort of like if you're going to open that box. You got to deliver more. And if you feel like you can't publicly, you've got to do it backchanneling. And even the backchanneling doesn't seem to have happened because Greg immediately sided with Sam. Mira Maroti and Ilya defected over to the Sam and Greg's side. And so it just looks like if something were really that bad, why did seemingly 99% of those involved immediately go to Sam's side, right?

• 26:48Speaker 02

  Yeah, I think the Greg, of all the things that happened from a reaction standpoint, the Greg quitting, stepping down, but I'm not sure if he was released from his duties on the board, if he stepped down and said, I don't know the sequence of them.

• 26:58Speaker 01

  They had announced that he was going to be removed from the board chairman role and just be president. And he was like, no.

• 27:05Speaker 02

  So obviously with the speculation from the announcement and what they said in there was, you immediately go to the egregious side with Sam. You say, hey, let's give the benefit of the doubt. But you just don't say those things in there. But then when Greg stepped down, he should have had full detail of it. They would have talked about that. Even in leaving him of his duties, they would have told him what, like, they should have told him. Like, let's just say that what they should have done. They should have told him what was going on. So the fact that he would have stepped down in that moment to that would indicate to anybody publicly that, wait, what the hell just happened? Because it can't be as egregious as that. Because Greg wouldn't put his personal reputation on the line like that. So anybody who's been through this for multiple cycles immediately sat there and goes, okay, handbrake, what the F is going on?

• 27:49Speaker 01

  And if it was really the doomsday scenario where there's a, you know, they have created a god and the god cannot be controlled, shut it all down, you wouldn't expect Ilya to change his mind over the weekend, right? It's been reported that Greg's wife pleaded with him and people were upset with him. But if this was like the world was at stake, you would hope that Ilya would have stuck to his guns, right?

• 28:12Speaker 02

  And this is where everything, any plausible scenario, and this is why I think we're kind of like in this weird limbo clown car style, you know, letting go of an open AI CEO.

• 28:23Speaker 02

  You know, the president co-founder quitting is none of it makes sense anymore because the statement doesn't match with what the other party was doing when Greg was doing. And then Ilya, if it was a God, can't come back and say, well, now I've defected, I overreacted, all that sort of stuff like, like, so now you're sitting there and from the outside observers and all of us just eating popcorn on the sideline saying like, well, at this point, who knows anything? So let's just sit here and just refresh Twitter for a little while.

• 28:47Speaker 01

  So can I, can I ask you guys a question? I know a lot of the audience are probably, you know, customers, potential customers or people who are not as necessarily as obsessed with this as us. Like, is this bad for the AI industry? Like, is this embarrassing for everybody? Or is it contained to just OpenAI? Because if I'm the government, I'm like, who are these jokers? Like, if this, if this is the top company in the whole category and they do this clown car that you just described, Jason, what does that say about everybody else? Like, I would not feel great as a government official.

• 29:18Speaker 02

  My view here is very strong on this point, which is, this looks terrible for the EA movement, which I'm very happy about because I think that that's just the silly thing. And also, it goes to the alternate structures of these entities. I think all enterprises and the government should be talking about the fact that these things were nonprofits or public good companies and the way that they're constructed and the way that they're incentive aligned and the boards and all that sort of stuff. This is an actual real conversation. I don't think it makes AI look bad. I think it makes a subset of things look really bad. And we don't know the story. And this is like in the moment, right now, on a, whatever it is, Tuesday afternoon, we don't fully know the story. But I think anybody who's going to walk in and start crying, you know, shouting

• 30:03Speaker 02

  wolf in the future is going to end up in a camp where this starts to be a thing. So on the positive side, because of the way my orientation is, I know I'm biased in this, I'm happy about that. Because we're, we're, we don't want to be slowing down on these things. We want to have real pragmatic, adult conversations. We don't want to have these scientific or sci-fi-ish sort of conversations. But on the negative side, that also means that anybody who does cry wolf in the future, like Ilya, like Ilya internally at OpenAI, what if they, there was no God this time, but there is one in the future. You're right, do this once with GPT-2, now he's doing it again. Let's say that. Actually is a scenario. He can't do it a third time. It doesn't exist. He's got no credibility to call, hey, we just created the god. So what happens?

• 30:41Jacob Effron?

  Yeah, no, it's a, it's a good point. I mean, Alex, one thing I am curious for your take on is you've obviously been doing this a long time. Like, is tech just crazier right now? Like, I can't, we've, we had like, you know, the SBF thing last year, same time, SVB happened, now this, like, I don't know if everyone just feels in every moment that things are crazy, or is like, this a particularly just insane period of time to be in the tech ecosystem?

• 31:02Speaker 01

  Bam, that is such a good question, Jacob. I, I think that you could easily argue that it is crazy and that just the world is crazier. Like, maybe we all lost our minds in COVID and this is just continuing to, to come out in different ways. Um, maybe it's that tech, you know, everyone pays more attention to tech, but yeah, it feels like, you know, we had, the past, we had scandals like the WeWork scandal, the Uber scandal, something like that. And it was sort of like one company and a lot of Silicon Valley could be like, oh, well, VCs didn't even invest in that, or, oh, you know, that's a consumer app and, and we're an enterprise company, we're in the cloud, whatever. It feels like the entire tech world is just obsessed with this story, like this, you know, maybe, maybe the biggest thing is that succession ended and we all needed a new season of succession.

• 31:48Speaker 01

  But it's, it's crazy to me to see how much this story has just kind of captured attention across the board. Um, I don't think we've seen this before at this scale.

• 31:58Speaker 02

  I, I agree. I don't think we've seen it at this scale either. I think like we've seen microcosms of it, but I also think that like there's a celebrity culture that comes with tech now too. So, and Sam and OpenAI are, you know, basically Jacob's on the other side of this, uh, trying to figure out how to invest in things in the moment because the way things are, there's, there's a lot of little celebrity making here and people are paying attention to one specific sector of tech right now. It's AI. Everything else is kind of dead.

• 32:23Jacob Effron?

  And in fairness, the future of shared office spaces and, uh, and, and ride shares doesn't quite impact the future of humanity as much as, uh, as AGI. So there's maybe a little bit more at stake, uh, or it feels that there's at stake for, for some of these things. Um, and I'm curious who you both feel walked away, you know, whatever happens here, uh, obviously it's going to be some damage control for OpenAI to do. Which like specific companies do you feel like walk away from this as, as, as winners or, uh, you know, at a better place than they were Friday at, uh, at 12:29 PT?

• 32:52Speaker 01

  Microsoft is interesting because I do think in some ways you could argue they're a big winner. And I think, you know, we even wrote a story to that effect, talking to a bunch of folks at Forbes yesterday on Monday. That said, I also feel strongly like if they could just rewind and have none of this ever happen, they would still prefer that. So I think like saying that they're the big winner here is overly simplistic because they want stability and predictability. And I think this was just like a huge distraction. Also, I just know for a fact that Satya hates doing press. And to have him have to make the rounds last night, like, I don't know if you guys watched the interviews, but he was not happy.

• 33:34Speaker 01

  So I think Microsoft, in some ways, in some ways, Microsoft wins, but like they would also not do this over again. So that would be my first one. Jason, do you have a better one?

• 33:44Speaker 02

  I would have said the almost identical thing, which is Microsoft's going to come out of this looking like a big winner, even though there's going to be internal conversations about governance and their investment and all that sort of stuff, rightfully so. But it shows how good they are at this too, because even when there should be some question, they can turn this into a thing. They would, and they very much would have preferred this never happened. But they're going to turn it into a win because that's what they do. I think other big winners in this are basically anybody who's competing with OpenAI directly. So Anthropic, obviously, is going to, is going to gain some share and some mind share, at least. Search engine query is going to go up, but anybody else who's kind of like, was sitting in a space saying like, hey, we're adjacent to this. You know, us or others like us are going to get some tailwind out of this.

• 34:28Speaker 02

  And it just, it just so happens that like, if there's a chink in armor somewhere for somebody and the OpenAI is clearly supposed to be the best of the best, but they show some vulnerability, that's always going to go to the other folks in the space.

• 34:43Speaker 02

  I think, you know, I think in another way, some like folks like Databricks and others can turn this into a win if they're aggressive. They don't need to be public. They just need to be more aggressive with their enterprises on this. I do, and I'll go back and I'll say this too, that I think the big losers in this all are going to be other hyperscalers. I just think that they look like they're absent. They're absent more than anything. We don't know what they're doing. Is there anyone at the wheel of those? And I'll answer this one other way. I think Sam and Greg as well, but Sam comes out a big winner in this. I don't know the story and maybe it's going to evolve differently and I'll be wishing I'd never said this, but right now he holds.

• 35:21Jacob Effron?

  Every single card as I can possibly see it, I can't, I can't think about how he does not have leverage in every single situation at the moment.

• 35:28Speaker 01

  Yeah, I, I think that's fair. Um, I just want to also throw out like, it feels like a million years ago, but OpenAI was in this kind of like, our philosophical debate or argument happening online with the open source community, but a week ago, two weeks ago, I do think that if they play their cards right, Mistral and some of the leading kind of open source companies should benefit from this to some degree, right? Like instability with the leader of, of one of the big model companies should be good for the open source community. But also, I, I'm not sure like if we would see that publicly as, as Jason noted, like that might be a more behind the scenes effect here. Um, can I throw out one conspiracy theory winner as well?

• 36:10Jacob Effron?

  Definitely.

• 36:11Speaker 01

  If Sam went back or does go back to OpenAI and Ilya does not stick around, um, whoever gets Ilya, uh, will probably be a winner. Um, and so especially if that's Elon and Grok, I don't think anyone I know takes Grok seriously.

• 36:29Jacob Effron?

  It's starting to feel like an NBA podcast when we're like, you know, talking about individual researchers and what, what team.

• 36:34Speaker 01

  So if Sam gets traded back and Ilya forces a trade to Elon, then Elon's franchise comes out a surprise winner in this, but who knows if that would ever happen. Mark my, if it does, you know, take me to Vegas.

• 36:47Jacob Effron?

  No, I mean, I think, I think you're totally right on the open source models. And I feel like, back to that earlier point of like, I feel like the space was waiting for a catalyst for a ton of people to try things beyond OpenAI. Um, and you know, Jason and I obviously have seen a lot of these like AI infrastructure companies and a lot of them rely on, you know, if you're just running OpenAI, you can kind of just use the APIs and there's some stuff to do, but there's a lot more to do, uh, if, if you're running your own open source model. And so I feel like there's a lot of these companies with really cool tech out there, whether it's helping you switch between models or, or helping you kind of, uh, you know, build, fine tune your own models. And I, I think all of them are kind of winners because they're, they're starting to be a little bit of a catalyst to, to, uh, you know, to adoption of that. Um, and so I think it's, it's a really interesting time for those companies. Whereas, you know, maybe before you, yeah, you could swap between models, but nobody was really that interested in swapping. It's like they were just using GPT-4.

• 37:33Speaker 02

  So on this too, because you, you know, I've been living in the open source world for the last 20 years. Um, I have some, I have some thoughts here and I just would like to offer this to any. Open source model company founder listening to this is figure out how you make some money. Like, please, like, your training runs are really expensive. Like, figure out how you make some money. The model is not the product. It is not the product. Figure out how you're going to make some enterprise money.

• 37:55Jacob Effron?

  I'm curious to, you know, obviously, one of the fun parts of this was, I feel like over the weekend, you know, there was a million different CEO candidates thrown out and, you know, then obviously, there was a lot of spoofs of CEO candidates that could be thrown out. Any that either of you were surprised weren't in there or that you thought would have been, would have been good picks? I mean, obviously, now we know it's probably not the best job to walk into when 750 of your 770 employees are out. But, you know, barring that piece of information, curious who you thought might have had their hat in the ring?

• 38:22Speaker 02

  I'll say, I'll say the one that was on there, which was surprising to me, which was Nat. Because knowing him well, he would not have been a fit for what the reasons they gave for booting Sam, which was he was going too fast. Like, Nat's not going to be a slow down, slow down guy. You know, Nat's going to want to go faster, too. You know, so that was a strange one to put on the list. Not that he would have been a bad one. I think he would have been a good one. He would have been a good CEO of that company. But it would be an odd choice if the reason you're booting Sam is because he's going too fast.

• 38:55Speaker 01

  I agree that Nat was the first name that came to mind because I just thought they would want someone with a lot of brand recognition and credibility in the space. So it didn't surprise me at all. Like, I was frankly texting people when that came out, being like, was Nat asked? Because it just seemed so logical. But I agree that it didn't make sense to the board's stated reasons. It just made sense from a, let's right the ship, let's not have everybody bail standpoint.

• 39:22Speaker 02

  Damage control one, because like you said, reputation, cachet, all that sort of stuff. But not for the stated reasons, which again, like, leads me to a different path of, like, just thinking about it.

• 39:32Speaker 01

  Yeah, so honestly, Jacob, I was unsure. I was shocked to see Alexander Wang on there as well, because that seemed like a conflict and I don't know how running scale would equip you to run OpenAI quite in the same way as well. And also, he's in the Sam world, so I didn't really see Nat or Alex defecting against Sam either. And it almost made sense to me just because he had been pretty vocal about safety stuff. And if that was the narrative that you wanted someone to sew things down, you need someone with maybe some founder credibility. And he was an OGYC guy.

• 40:12Speaker 01

  He obviously ran Twitch, which is a big company. So you could kind of force that one to make sense. But I had honestly thought they might go with like some well-known researcher or like person. Like if that was really the route you were going, you go with like a well-respected professor or like researcher who's gonna like bring it back to being a research organization, if that's what you're trying to do.

• 40:36Speaker 02

  And that to me is actually the challenge with all of this is that none of the storylines actually add up. It's like all these different, like we're watching this mini-series, all the different storylines, and none of them actually conclude and add up correctly. It's like, I don't understand why we're even listening to this side story.

• 40:54Jacob Effron?

  Yeah, we're gonna need some epic concluding episode that makes it all make sense.

• 40:58Speaker 02

  Well, I mean, if someone could pull this together and help me understand this, otherwise I just feel like I just watched, what is it, Lost? And that last season, you're just like, no, I'm so upset with myself for spending time on this.

• 41:12Jacob Effron?

  I mean, I guess if Sam does come back, like what does he, what do you do? Like, what does he have to do in that, those like first few weeks? I mean, I imagine there's a huge customer tour that needs to happen of reassuring.

• 41:21Speaker 01

  He has to go to Red Men, Washington.

• 41:25Speaker 02

  Well, he's doing that regardless, right?

• 41:26Speaker 01

  Yeah, fair enough. I think Sam has to do a customer tour for sure. I think in recent months, he spent a lot of time with world leaders. And, you know, based on reports, maybe like sovereign wealth funds and stuff like that. I think he needs to get back to enterprise customers for sure. And I think he needs to spend a lot of time with staff. But yeah, I mean, let's hear the CEO. He probably has a more, he would have a more tactical playbook. But from my standpoint, I saw Sam on Wednesday. He was on his way to or from an Apex thing, you know, with world leaders. And it seemed like that was kind of more the people he was rubbing shoulders with before this.

• 42:05Speaker 02

  Sam's gonna be Sam to a degree. Though the classic playbook here would be make sure you shore up your base. Like in base is gonna be enterprises, they have to be good with you. Microsoft and Redmond in your internal. Make sure you do that, kind of going into a hole for a little bit. Like I would limit the public appearances for a little while when it comes to like the non-open AI type of things. And just make sure that your folks are focused, building the right things, the distractions are gone, make sure all that stuff has gone away. You've gone to Redmond, you've gone to your enterprise customers. And also, he still has to close the tender offers from the investors out there to make sure that his people don't get basically screwed internally from what's happening there. So that's another thing that you would have to shore up with the other investors.

• 42:50Speaker 02

  Like you do all of that, those sorts of things, you know, that's what I would be focused on.

• 42:55Jacob Effron?

  Well, I guess, you know, to kind of wrap up, I'm curious, across all the different things that happened in the saga, obviously, I think the most shocking was probably just the news happening in the first. What was the biggest jaw drop moment for each of you in the twists and turns that have happened in the last five days?

• 43:10Speaker 02

  So obviously, yes. I saw it from a Delian tweet. Delian tweeted it. I'm on a low Wi-Fi signal, like 3G area or whatever it was. So the image wasn't coming in. So it was just like Sam was fired it open air. I'm like, ha, ha, Delian, good one. Let's figure that out when I get some more signal. But when I came through, I was like, holy shit, what the hell? But I think the biggest jaw drop for me was when they went in to negotiate effectively, when I heard the board said, well, we're in talks with Sam to come back. You're just kind of like, what kind of? Idiot parade is what's happening here at the moment. But then when Sam didn't come back, they dug in their heels and he had his demands and he didn't come back. That right there, that was the one I would say, everyone who asked me, I was like, of course, of course, that's going to happen.

• 43:55Speaker 02

  They're gonna go, he's gonna go back. It wasn't the Microsoft thing. It was that moment when he just, when he didn't go back. I was like, oh shit, okay, whoa.

• 44:03Speaker 01

  For me, for me, it was when Microsoft announced he was joining Microsoft and then he didn't join. Because the one group here that I kind of felt like had their comms down was Microsoft. And like would not make an announcement that could potentially move the market that would not be backed up. So, so when they started to play, have cold feet or sort of backtrack on him actually becoming an employee. That's where the story went from like, I can follow along to like, you can now just tell me literally anything and I will believe you. Yeah.

• 44:42Speaker 02

  Well, so, so this, this is the thing I'm in at the moment, which is like, I don't know. I don't know. Like, I want to hear the rest of the story. And obviously we're just having fun speculating and talking about all the different things and our perspectives and all that sort of stuff. But my wife's upstairs, I've been, you know, I was away from her for a little while, I was texting her all this stuff. And she's asking me what's going on and she's literally like, this is a new season of some show and like, what's happened now? Like, it was like, well, we, we got this new character, Emmett showed up and let me tell you about his past real quickly. And she's like, no, no, you're kidding. That's not, that's not true.

• 45:12Speaker 01

  Yeah. Man has some bad tweets. We need that meme from Secession of everyone chanting bad tweet. Yeah, that, that's apt here. So.

• 45:19Speaker 02

  It's like, what, what, what are we doing? What are we even doing here? It's like, good thing we're not actually building rockets, maybe to a degree at the moment, or surgical equipment, but like.

• 45:27Jacob Effron?

  Yeah, let's, let's get this, let's get this all settled out before like GPT five or six. Like, good, good that we're running through this whole thing at, at four.

• 45:34Speaker 01

  Can I say there was someone, I believe they were in India, like a business person there who tweeted at Sam and for some reason it like made its way to my feed. It was a response to one of Sam's like, I love, I think it was, he was like, I love my team and then Greg hearted it and something like that. And maybe Mira also responded with a heart. And this person responded and they were like, can you all just get on a WhatsApp group, figure it out and just update us on like the actual ending? And I was like, this person is right. You know, for 99% of the world, I think they're not having as much fun as us. And they're just like, can you just tell us what happened? Like, cut to the chase.

• 46:11Jacob Effron?

  Yeah. Should I keep having my product run on OpenAI or not?

• 46:15Speaker 01

  Does this affect my business? For now, it's affecting my business because all these people that I work with are seemingly obsessed with it. Yeah.

• 46:23Jacob Effron?

  Well, I feel like we could do another, you know, 10 hours of speculating how the story ends. But Alex, I'm already terrified that in the hour we've had together, you've missed like seven scoops. So I'll just say a huge thank you to both of our guests here, Jason and Alex. This was an awesome episode and hopefully the first of many of these emergency debriefs, though not under the same circumstances.