Altman’s Testimony, AI SPV Drama, Ebay Rejects $GME Bid | Diet TBPN

We have a great show for you today, folks. We have a bunch of guests and a bunch of news stories to go through. Of course, the trial is ongoing. Did you say that this might be the last week of the trial? I thought it was a 4-week trial, but it sounds like it might wrap up in week 3. Are they ahead of schedule?

Yeah, Mike Isaac said it might end this week. I assume just because they're getting through the, like, you know, depositions, whatever, faster.

Yeah, I mean, it seems like Ilya has gone, Mira's gone, deposition from Sam Sattia, and Shavon Zillis, and then Sam's on the stand right now. I think Mike Isaac is live tweeting it, so we'll run through that. Sam Altman took the stand in the OpenAI versus Elon Musk trial this morning. Just as a reminder, here's what's at stake. Per the Wall Street Journal, Musk is suing OpenAI and its leaders, Altman and Greg Brockman, for allegedly manipulating him into giving tens of millions of dollars to a nonprofit organization only for them to turn the AI lab into a for-profit venture. Musk is also suing Microsoft, OpenAI's largest investor, for aiding Brockman and Altman in their alleged deception. So big turning point in the trial was last Wednesday when former OpenAI CTO Mira Murati and former board member Helen Toner gave testimonies about the events leading up to the November 2023 failed board coup that were critical of Sam Altman's leadership style and candor. They call it the blip where Sam was out and then back very quickly with a couple other people stepping into CEO for just a few days. But last week, OpenAI's side also began to land some punches on Musk. Earlier testimony from Shivon Zilis and Greg Brockman also had already suggested Musk was not just defending a pure nonprofit vision. He explored scenarios where OpenAI might become a part of Tesla, where Altman might help lead Tesla AI, and where Musk could retain deep control. Brockman also testified that Musk supported a for-profit conversion if Musk could control it, including a version tied to raising money for his Mars ambitions. Yesterday, Microsoft CEO Satya Nadella and OpenAI co-founder Ilya Sutskever took the stand. Satya largely buffed OpenAI's side, defending Microsoft's partnership with the company and saying that Musk never contacted him to complain about the deal violating any agreement Musk had with OpenAI's nonprofit, despite Musk having Satya's number. Ilya testified that he spent a year compiling a 50-page document documenting Sam's manipulative behavior, but also said he never promised Musk that OpenAI would remain permanently nonprofit. Also, it came out in the trial that Sutskever's stake in OpenAI is probably worth around $7 billion, which probably complicated how the judge and jury feel about his own motivations. So the story is, since Murati, first the trial became a referendum on Altman's trustworthiness, then it became a referendum on whether the 2023 board was brave or incompetent. Then Microsoft came in and tried to make it look like it was the stabilizing partner. Now Altman has to personally answer the core questions hanging over the whole case, whether OpenAI OpenAI's evolution was a necessary adaptation to build frontier AI or a betrayal of the nonprofit mission Musk says he funded. The trial is expected to conclude this week, and you can—

Max Zeff pulled out a quote from Ilya. This is explaining why OpenAI has a for-profit, and he's— he, I guess, yeah, was playing, and I don't think intentionally playing into the meme potential, but certainly that's how it played out. Ilya said under oath, in a federal court. If there's no funding, there's no big computer. Max says, in the running for quote of the year.

If there's no funding, there's no big computer, and you need big computer if you want big AI. Sam Altman takes a stand in Musk versus OpenAI. Mike Isaac has a live blog going on X under Rat King. Let's run through some of what Mike Isaac is finding. OpenAI begins questioning Altman much in the same way that the plaintiff's side had questioned Elon Musk, establishing that Altman, like Musk, has been enamored with AI for years and wanted to build inventive things with it. This was a very interesting tidbit from that, that original, like, that viral quote from Sam where he's like, AI will probably destroy the world. That happened in 2015 and it was an answer to a question of like, what do you think the key problems to solve are? And then the next sentence in that quote that always gets clipped out is like, and I'm starting a company. Well, it's more of a nonprofit to work on this problem. Exactly. Yeah, but that doesn't make it in. But so they were clearly both very interested in building beneficial AI, although they ultimately butted heads. And they were also very worried about Google. Altman email from 2015: Been thinking a lot whether it's possible to stop humanity from developing AI. I think the answer is almost definitely not. If it's going to happen anyway, it seems like it would be good for someone other than Google to do it first. Altman said Musk once said he would potentially pass control of OpenAI to his children upon his death. I would love to know more about what that means, because if you fractured into like 20 different children at a certain point, like that can create a whole different dynamic of like succession, right? There's an old email that says Altman might have joined the board of Tesla as part of old AI discussions. Also came with a nascent threat of Musk doing this AI work inside of Tesla. Very, it would be a shame if you didn't accept my offer sort of moves. This Tesla offer is interesting, says Mike Isaac. Musk offers a board seat, Altman says, It was something he felt was to assuage concerns that Altman would have no direction over the development of AI if it were folded into Tesla. But Altman also said it appeared to be a nascent threat. If Altman had not accepted the idea, according to him, Musk hinted that he may have done work developing AI on his own at Tesla regardless. And Mike Isaac gives some more context here. He says, this sort of talk is fairly commonplace these days on the battlefield that is Silicon Valley. Mark Zuckerberg, Meta CEO, has in the past made overtures to companies he's interested in acquiring, though he is often more explicit about his intentions. If you don't take the deal, we'll come for you. Tony Soprano vibes. Altman is criticizing stack ranking of engineers across AI labs, something Musk loves to do across his companies apparently, and something very common across big co's like Meta and Amazon. And also Microsoft is known for the stack ranking. They almost like invented it or certainly popularized it throughout the '90s and 2000s, I believe. Altman says, AI engineering labs need more psychological safety. You have to be willing to let a researcher go off and try something random in the corner, very bottoms up. This is where the Deep Research project came from and a bunch of other AI breakthroughs came from. Altman says there was a meeting at Tesla during the evening about folding OpenAI into Tesla for AI research, and then a long, long period of time with Elon showing us memes on his phone. And apparently the court reporter asked Sam Altman to repeat memes on his phone loudly. I don't know why they didn't hear it the first time, but Rat King Mike Isaac is laughing at this.

One of us.

I am going to have Sam Altman stating memes on his phone into a booming courtroom mic playing inside of my head on repeat for the week. Memes on his phone in all caps. Getting hungry. He only had a banana this morning. Mike Isaac is suffering. His stamina points are draining. And he says, LOL, email between Altman and Siobhan talking about how to handle Musk and Altman telling him about a Microsoft investment. So Microsoft's going to invest. How, how should we tell Elon? Altman narrates the email verbatim. Hopefully it's easy. Cross fingers emoji because you have to read it out. So really funny to hear cross fingers emoji in the court record. Altman is going through his real first postmortem of his firing, appears to have gone through all 5 stages of grief multiple times over the course of 5 days. What are the 5 stages of grief? Stages of grief.

Cope, seethe.

Cope, seethe, mold. Denial, anger, bargaining, depression, and acceptance. Yeah, that is actually cope, seethe, mold, munt, something. The 5-day period is what is referred to internally at OpenAI as the blip, since it was a brief intermission, for what it's worth. Altman said, I had poured the last years of my life into this. I was watching it about to be destroyed. There was something appealing about going to work at Microsoft. I was also very angry, hurt, and upset. It felt like an incredible betrayal. It was definitely one of the hardest times in my life. Altman finally broaches the issue of his widely rumored untrustworthiness. Clearly there were misunderstandings and a breakdown of trust, he said, but with what comes of a bit of a practiced humility in his voice, quote, I was not trying to deceive the board. I feel badly for the misunderstandings, but that was never my intent. This goes to the heart of how Musk's counsel has tried to portray Altman across the entirety of the trial, a fundamentally slippery operator who says one thing to one party and something else entirely to others. OpenAI's rebuttal to that line of thinking has been to depict a board of directors at OpenAI rife with dysfunction. And as Microsoft's Satya Nadella put it earlier in this week, directors who are operating from amateur city. Interesting. Taking shots at the board. Okay. Sam is giving the full jury Sam treatment again, trying to bat back against Musk's picture of Altman as a serious liar. If I knew how difficult and painful this was going to be, I never would have tried. But I'm very glad I did. OpenAI is done. Cross-examination begins. Stephen Molo, lead Musk counsel, who has the flair for dramatic, will probably give us fireworks. And this is continuing, says, oh my God, Molo, are you completely trustworthy? Altman says, I believe so. Molo says, do you always tell the truth? Wow, this is getting heavy. It's all about Musk counsel painting Sam as a liar. Brutal. Altman is on the defensive, but taking more of a muted tone with some attempt at humility in his voice. It's very clear contrast with how Musk appeared combative on the stand. Interesting. Cross-examination is basically Musk's lawyer, Molo, reading off a list of questions saying, hey, bro, do you remember? Do you remember all this messed up stuff you did? And Altman saying, no, I don't know. Not true. No. Absolute chaos. Well, you can follow along @MikeIsaac. He has a whole thread and he's live posting. And I believe that there's a New York Times live blog as well that you can follow along with.

Should we talk about SPVs?

Yes, let's move over to SPVs. The special purpose vehicles are rocking the Valley right now. People are raking in the dough with SPVs. Yeah, so not everyone's happy.

So Yeah, we can actually go down a little bit and pull up this post, which has since been deleted.

Oh, really?

The post said a few days ago, simply brokering an Anthropic secondary deal made me more money than my entire net worth from working in my 20s. This is insane. It is especially insane because this is not legal. It is insane to post.

Is it not legal?

Yeah. You need a broker-dealer license to broker securities.

Yeah, that's right. That's right. Well, I mean, she didn't say that she doesn't have one.

She might. Very, very unlikely. It's very burdensome. Just the compliance to actually get your own broker-dealer. Yeah. Even there's people that I know that just do secondary transactions. They don't even have their broker-dealer license. They work under a firm that does. They basically like contract with a firm. So they're kind of like an almost like a real estate agent working under I'm smelling an intern challenge. Tyler, get your broker-dealer license. Figure out a way to use AI to bring down the compliance burden.

Slash goal. Throw that slash goal down. From the X-High. X-High 5.5 codex.

Yeah.

And try and get your broker-dealer license.

Yeah.

So, and so let's see what it does.

Yeah. Turns out that if you want to broker securities.

Yeah.

Transactions. Yeah, it is. There's a big regulatory burden. Yeah, for good reason. Right. We're talking about, you know, transactions that are at the scale of high-end residential real estate in this case.

Yeah.

You know, if this individual was maybe getting like a 5% fee on the deal paid in cash, who knows?

It could—

it could have been. Yeah.

Wait, there is another take here, which is that There's the term brokering, which requires the broker-dealer license, but there is also the format where you set up an SPV, the SPV takes in money from an investor, and then buys secondary from someone who has the right to sell it, maybe an investor who is not subject to the form that we saw Anthropic put out. And in that case—

We'll get to that.

We'll get to that. But hypothetically, There are SPVs that they are not technically brokering the secondary deal, but they are facilitating it, and they do take a fee, right? Because SPVs often have fees associated with them, and that does not require a broker-dealer license, right?

So it's possible. But the straightforward interpretation of the post is that they had some sort of side letter, which was like, if I can find you— yeah, I can find you a buyer. If I can find you a buyer for $100 million of your shares, you give me $5 million. Yeah, yeah. That happens a lot, but it usually and hopefully is happening through legal channels.

Yeah.

Dealers. So this seemingly prompted both Anthropic and OpenAI.

Oh yeah, I think this is what started it, right? This must be what started it. 'Cause that post really, really mattered.

More specific language on their site saying Anthropic said unauthorized Anthropic stock sales and investment scams.

Yeah.

And said, you know, basically saying any transfer or sale of Anthropic stock or any interest in Anthropic stock that has not been approved by our board of directors is void and will not be recognized on our books and records. And so, yeah, typically, yeah, typically like these are— there's so many ways to like transact without informing the company.

Sure.

Right. A common one would be futures contract, basically selling the right to the investment and the future performance.

Yeah. And so I think the futures, the economic exposure enthusiasts, or the futures contracts teams would say that, well, we didn't, there was no sale of the stock. We didn't transfer the stock. And it's not a direct interest in the stock. And so it doesn't need to be approved by the board of directors. And the board of directors would say, absolutely not, that doesn't count. You think you found a workaround and it doesn't count in this world. But that is going to be, you know, fought out.

Yeah, so the funny thing is there's these sort of digital asset equivalents. Yep, some of them put out comments. Like people making basically meme stocks around that are trying to track overall interest or the overall, valuation of these companies. They sold off, which is funny because I don't believe they're actually tied to any real underlying equity.

It was purely sentiment-based. And even in the case that they were like saying that they were tied, it was typically like 1 to 5% of the fund in Anthropic or OpenAI or SpaceX. And they were like already well disconnected from the fundamentals of the book value, right?

Yeah, so I'm trying to— think through how this plays out. And overall, I think the reaction from the internet was being more dramatic than maybe is necessary. Because already, if you're, let's say, an early investor in Anthropic, and you at some point sold your shares, you didn't go to the board and get permission, but you structured some deal to sell your shares.

You get your shares back.

That is so— so, win-win. Well, waiting for that. On one hand, neither party, if you bought the shares or you sold the shares, neither party is that incentivized to go to Anthropic board and be like, hey, I'm really sorry, we did this, it was against— because on one hand, there's probably some scenario where the investor or the Anthropic employee could get their shares voided or reclaimed in some scenarios. So they don't necessarily want to do that. The investor is like, well, I bought these shares, and now they're worth a lot more. So let's just be chill and let this play out, and we'll all forget about it, right? But there is a scenario where the seller tries to then make the case of, oh, actually, I'll just give you the money back because now the stock has appreciated so massively.

Sure, sure, sure.

There might be a weird— but what ends up happening is like, there's this legal tension, tension between both these parties. And then there could be some incentive. Again, the person that bought the shares at a lower valuation wants to just let it ride. But then somebody might be like, well, I would happily ride up another 20x or something like that. But then if this starts going— if once it goes into an actual complaint or a lawsuit, then it becomes public. And then you have this— third party in there, which is Anthropic, which is gonna just be like, hey, you guys have been messing around. This isn't cool. This is against multiple agreements and terms. So anyway, it's gonna be very messy, right? 'Cause there's already been—

These blog posts are not new rules. It's merely they are publicizing rules that are probably already in the spec document.

Yeah, because if you invested earlier, you're an employee. You should know all this, which isn't a surprise, right?

Well, it's possible that very early investors don't have transfer restrictions for some reason. I don't think so.

No, no.

It's pretty standard.

No, you always set this up as a company, because imagine you have an early angel investor, and your company does well, and they just sell it to somebody who you don't like. Like, part of the reason to be private is you can control—

I just mean, like, every deal is unique, and every deal gets negotiated points, and there might be at some point when some investor or employee was like, I'm not joining unless you give me this. And they're like, okay, yeah, we'll pay you less, but we'll give you this.

Yeah, maybe.

There's always like horse trading, like double trigger, single trigger.

Yeah, but it would be— it would be like—

I agree with you. I agree. I agree.

I agree with you. Who knows how many transactions there's actually been? I don't know. There could be like— when you actually look down through all the trees of SPVs and LPs, there could be what, what, a lot, 10, over 20,000.

It's not, it's not like taxicab driver telling you about the SPV that they got into yet, but it's like close to it. I mean, there was the story of the guy who was like selling his house for Anthropic secondary, right? Like there's a lot of examples of this.

Like, oh, that's bad too. Imagine, imagine the like record updates. Like, you'll be able to look at the deed or whatever and be like, oh, who's the new owner? Oh, you work, you're an early— Like, so that transaction, that whole transaction doesn't really work.

Oh yeah, and that's extremely public.

Yeah, yeah, yeah, that's what I'm saying.

It's way more complicated, yeah. Well, let's break it down for the Frog and Toad fans, the children in the audience, because Frankie over at Paradigm put it in terms even a child could explain, potentially a 4 or 5-year-old. So if you're familiar with just Frog and Toad and you don't know anything about all the buzzwords we've been dropping for the last 5 minutes, you can think about it this way. Frog and Toad, beloved children's book. Frog put the shares of Anthropic or OpenAI in an SPV. There, he said, now we can transfer these shares freely. But Anthropic can still exercise its transfer restrictions, said Toad. That is true, said Frog. The Frog and Toad, it's a great one. Ankur says, if Anthropic deems all secondary sales of Anthropic stock should be voided, does that mean the original buyer retains financial interest even after sale? Selling it away, lawsuit territory. Yeah, there's gonna be a messy thing. People have been talking about unwinding SpaceX AAA or SPVs for a long time.

Yeah, that's gonna be a billion dollar, that's gonna be almost a billion dollar industry.

Yeah, we need to get the Matt Graham clip up again. It goes viral every time we talk about SPVs. What does Mike Isaac say? He says, "Yeah, look, all the privcos draft this language to scare employees who don't know better from trading on secondary markets and from buyers seeking those shares." And yet, SPVs find a way. Funny to see the saber rattling in Twitter accounts doing hyperbolic posts though. And so it'll be interesting to see like how far does the legal implications, how far do they actually go? Well, over at a different AI lab, Mira Murati, who was just, was Mira on the stand? She was actually testifying in person or was she just video deposition?

When I saw her, which was last Wednesday, it was just deposition. I don't believe she actually has testified in the trial yet.

Von that was on the stand.

Correct.

Got it.

Okay.

Well, Mira is cooking over at Thinking Machines Lab, TML, launched interaction models with a delightfully concise YouTube video that we should play so that we can watch this. Mira Murati says on X, today we are sharing our work on interaction models, a new class of model trained from scratch to handle real-time interaction natively instead of gluing it to onto a turn-based one. Let's play the video.

Hey, I need your help with something today. You ready?

Absolutely, I'm ready.

What's up?

Yeah, so we're giving an announcement today and I've got two of my friends coming to help. Every time one of them enters the frame, I need you to say "friend." Look at those speakers.

Got it. Absolutely insane setup. Audiophile.

So we've got a new system for full duplex audio and video. Which means that you can stream input into it in real time and it can respond to you even while you're speaking to it simultaneously. How does that sound?

Sounds like a solid setup.

Full duplex with real-time interaction is super useful. It seems faster than the original voice mode, which was lamented by the viral Instagram Reel producers.

Can you translate into English in real time for my friend and for audience?

Absolutely.

I'll translate as you go.

Ajam, a model.

Today we're taking a look at our preview model. I saw or I heard about a version of this in China. That's a mask that you wear that translates everything you say out of a speaker on the front. And I was. I was just hearing this and I was just like, why is this not in America? This seems so sick. Like we hear about the AirPods, but then you— so the example was a mom in China teaching her kids English. And so she basically wants to be talking to them in English constantly, but she doesn't know enough English to teach them, but she wants them to learn. And so she will, I think, wear—

Put on the mask.

No, it's literally a Bane mask. It looks exactly like a Bane mask. Yeah. You merely adopted English. She was born in it, or born in the AI translation minds. So she wears this Bane mask that does the live translation out to her kids. Her kids speak back to her in English. Smart headphones translate back.

Is that like metal that wraps around?

I don't know. I couldn't find it. I was listening to it on a podcast, so I didn't have any visuals. But we got to find this thing and get a pair in America, because in theory, we could do the whole show speaking Chinese to each other, and the audience would hear Chinese, but we would be hearing English that we talk to each other. Isn't that cool? Powerful. So I would be hearing English on a massive delay, probably, but I would be speaking Chinese as it comes out of my Bane mask. And it was just a whole story. The whole thrust of the New York Times Daily was is like the optimism of AI, optimism in China around AI. Just tons and tons of examples of everyday people being like, oh yeah, like AI's amazing. I'm teaching my kids English. They're gonna have a great life. And like, I would not be able to do this before and now I can just do this. And there's like so many examples of that and it's the exact opposite.

David says the moment Jordi gets a live translation button, it's over. Yeah, we're not that far away from me being—

We're working on that right now actually. Yeah, for—

Are you serious?

I guess. But, uh—

Oh yeah, for guests.

For that mask, I would actually be very surprised if that's, like, real, or at least if the audio sounds very good. No, because, like, you can just look at, like, okay, what are the best, like, real-time translation models? What are they, like, API prices? They're, like, not super cheap. So you can't do it locally, which means it's somehow in the cloud, right? And just, like, even the best models are— there's still some delay. And they're just basically now getting to a point where it sounds like a real person and not like super computerized audio?

It has a clanker dialect.

Okay, so yeah, maybe, but just like getting the latency down is extremely difficult because people have been working on this for, I mean, just like Google Translate.

You don't think you can do it on device at all? What if you basically built a phone?

At some point you can, but I think it's still like very difficult.

Phone ASIC for this one model, you take the Llama 3 version of it, you bake it down, huge battery pack, you're wearing a whole jet pack full of batteries to power the H100 in the back. It's on-device inference, but they didn't say how big the suit is you have to wear that connects to the mask.

Like, yeah, you have to trail a whole rack for you.

No, it's like the Nathan for You, the Philly suit.

Yeah, but it's an NVL-72 that you're just like dragging behind you like a washing machine. Uh, anyway, GameStop is going to need to level up if they want to successfully take over eBay because eBay has has rejected GameStop's $55 billion takeover bid. The online marketplace— that's eBay— called the cash and stock proposal, quote, "neither credible nor attractive." The New York Times has a story here. The online marketplace rejected the proposal. This news should surprise no one since the odds it would accept GameStop's brash offer were infinitesimally remote. There is another media deal going on right now. Byron Allen is buying BuzzFeed. Investing $120 million in the digital media company and will become the CEO. This is a very interesting story because I don't know how familiar you are or the audience is with Byron Allen, but he had a fascinating career where he was originally— he jumped straight into late-night talk show host. He became a late-night talk show host and then eventually had this very interesting business where he would buy—

Zero to Yap.

Basically. He would buy the rights to broadcast on TV in certain slots, in certain hours, and then he would independently go sell advertising against the programming that he would put together. And so the money would flow from the advertiser to him, and then he would pay to air his content on traditional TV. Interesting. It's very interesting.

So he was on the hook for the airtime basically, but any increment, any difference was—

But he was paying for the airtime. And then he eventually bought the Weather Channel and a number of other sort of traditional over-the-air TV stations and sort of grew. He's also will be taking over the time slot from, I believe, Stephen Colbert post that changeover with a show called What is it? Comics Unleashed. It's a roundtable conversation. It's sort of just a podcast with a bunch of comics. Allen Family Digital, a company associated with Mr. Allen, will pay $120 million for 52% stake. And this is a huge jump from where BuzzFeed was trading. I think BuzzFeed was trading around like $40 to $80 million market cap or something. So it's like almost a 3x premium. And we can get into the share price in a minute. But yeah, I'm sure— There will be more developments on what happens with BuzzFeed. Uh, as the story continues.

Yeah, I'm so curious what the plan is. I don't think it's a good brand at this point. Certainly has name recognition, but I don't know anybody that wakes up in the morning and says, "Yeah, I gotta know what BuzzFeed is talking about today." Granted, there's probably a lot of people, and Byron, uh, is, uh, a media tycoon.

Yeah.

Now, that's why I'm curious. Like, what's the play here? No, I think he sees an opportunity. If you had $120 million to just build a new media property, you wonder, hey, what could you actually do? So is there some massive audience that's still sneakily more engaged? Maybe he sees a pivot to prediction markets. Oh, I don't know. I don't know.

That's a good paper for BuzzFeed to do. BuzzFeed.com. 50, why would you put that thing in writing? Photos that prove people are the worst. That's the number 1 trending article on BuzzFeed. 41 celebs people used to love and now can't even stand to look at. 22, they really go, they love the listicle. And they're getting longer. I was joking about 5 key reasons. That would be, that would never make the cut at modern BuzzFeed. 22 stories about boy moms. And their sons that will make you cringe into oblivion.