Ladies and gentlemen, welcome back to the European Technology Network, ETN. What a time to be live. Today is Tuesday, 12th of May. It is 3 minutes past 11:00. Boy, do we have a show for you guys today. Before we get into that, let me tell you about our friends over there at Airwallex. Leave outdated business banking behind with Airwallex, the all-in-one financial platform built for the pace of modern business: multi-currency accounts, corporate cards, expense management, online payments, all in one platform. Go to airwallex.com. Welcome to the Tuesday show, and what a show we have for you guys today. Just to round up some of the guests we've got coming on: we've got VB, DevRel at OpenAI opening the show; Christos Charalambous, GM UK and Europe for Airwallex—let me hit the trumpets for Christos. Following Christos, we have Henri, co-founder of the RAISE Summit, with some news to come. Following that, we'll also have Alex, the CEO and co-founder of Deel. They've got a bunch of things lined up to announce. They have announced them already; we're going to be running through a bunch of them, digging into what those guys are up to. One in particular that I'm very interested in is this essentially tokenization tracking for employees. Managers can now track how much token spend their employees are running through. It's the new craze: does productivity equal token spend? We will find out. Following that, we have Ferdinand Dépêche, co-founder and CEO of Augustus, who as of, well—
Yesterday, when it was officially announced, is the youngest person ever to lead a chartered bank. He is 25. He hails from Germany. He's going to be joining us and breaking down what is Augustus, what is this new form of banking that he's trying to pioneer. And then we have, coming back onto the show—we love to see these guys—Ethan and Jamie, co-founders of Icarus Robotics. These guys are trying to decrease the cost of labor in space, so no small feat whatsoever. What's that noise over here to my left? I hear clinking and clanking.
Just dropped coffee everywhere. It's fine, don't worry. It's in the Framer mug, so we're good.
Let me tell you guys about Framer. Framer: build better, faster.
It's the all-in-one platform where enterprise teams design, build, and ship production-ready websites: no dev handoff, no bottlenecks, no creative compromises. We migrated ETN to Framer ourselves and had a great time doing it. You can too at framer.com/etn.
Indeed. God, it was fun migrating that website.
Fun job. Seamless. Such a good job.
Seamless as you like. Let's get into the news before we let on our world. It's been a banger weekend. It's been a banger weekend. RAISE, a bunch of other stuff. I mean, yesterday was a bit of a banger as well, to be honest.
We got some OpenAI stuff. We got some Thinking Machines stuff. We got some Musk final trial stuff. We got—
Anthropic big news, which I don't think was big enough. But anyway, we're going to get into it. Let's start on the OpenAI news because this is actually super interesting. So OpenAI now have launched the OpenAI Deployment Company. They came out with an internal article yesterday, and the title was "OpenAI launches the OpenAI Deployment Company to help businesses build around intelligence." OpenAI has agreed to acquire Tomorrow, giving the OpenAI Deployment Company experienced forward-deployed engineers, who—they were based in Scotland.
I'm pretty sure they were Scottish.
Let me just triple-check that. I'm pretty sure they were acquired by OpenAI.
Well, while you do that: OpenAI is launching the OpenAI Deployment Company, a new company designed to help organizations build and deploy AI systems that they can rely on every day across their most important work. Successful AI deployment is about empowering people and teams to do more. The OpenAI Deployment Company will extend OpenAI's ability to embed engineers specialized in frontier AI deployment, known as FDEs, into organizations working on complex problems in demanding environments. These FDEs will work closely with business leaders, operators, and frontline teams to identify where AI can make the biggest impact, redesign organizational infrastructure and critical workflows around it, and turn those gains into durable systems. Give it up for OpenAI. I mean, super interesting, right? OpenAI is moving deeper into the consulting, system integration layer. This AI race isn't now just whose model is better than whose, what benchmarks am I beating the other model providers on, and all of that kind of stuff, right? It is how do we actually integrate into these businesses. The bottleneck is now deployment; it isn't the capability of the model, which is a really interesting move. I think Anthropic specifically have done a very, very good job of making their model—their model is one of the best, if not the best. Opus 3.7 just is—I mean, I'm using Opus 3.7 almost exclusively over the others. And they've done a really good job of integrating into the developer ecosystem way better than I think Codex have. I mean, Codex have been doing pretty well, but it'll be interesting to hear VB's—
You can spar with VB when he's on.
—how the developer ecosystem is going. But I'm thinking, let's look back, right? Like, this is a really clever move from OpenAI, really, really clever.
So this is off the back of the news around them launching this joint venture with a bunch of PE firms. A bunch of consulting firms are also working on something as well, but OpenAI seemed to have beat them to it. But this comes back to when we spoke months ago now. In the words of Manny Medina, previously co-founder of Outreach, now founder of Paid, this year will be about realizing value. We've had the "we're all AI-first companies," and they've deployed it terribly, and they've used it terribly, and there isn't a conversion to how much AI are you using compared to your revenue, like, are we making a return on this? Same thing with the Omni founder after they raised their 120 million. He jumped on and said, "We are definitely a forward AI company. We are adopting AI in all of our workflows. But if I've got someone coming up to me in the marketing department and saying, 'Hey, I need 10 million to run this huge AI marketing campaign internally,'"—
"—using a bunch of AI content, a bunch of AI stuff," he was saying, "that better bring back 50 million worth of ARR."
Sure. So people are now looking for how can we actually get this into an enterprise well and see an impact on our top line, and the market is massive. All these conversations around the next trillion-dollar company will be a services firm masquerading as software.
Yeah. Like, this is that. These guys will be selling outcomes with AI at these companies, and they will—
It's not selling, "Hey, this model will help you do this if you—" and it's kind of an open book of, "Go and do what you like with the model. Go figure out what you want to go and do with the model." It's more, "Hey, what specific thing do you need solving in your business? Oh, I want this done, but I have no idea how to even go about it, regardless of how good the model is or not." "Okay, cool. We'll just put an FDE into your company for a few weeks." I don't know how long they—do they get contracted in for like two weeks?
Yeah, because I think it would run like a traditional consultancy firm.
I assume these projects won't take months and months. Like, I know consultants who have been on projects for—
Could be. Yeah. I think that this will be—I remember seeing something about the price point: it will be expensive. So I'm assuming that this will come for—
Only large scale. I'm talking FTSE 100, I'm talking NASDAQ.
Yeah, but it will trickle down over the next kind of 5 to 10 years, right? Where it's—I mean, I would say close to five.
So this Ben's take we have here speaks to this.
We can get that one up first. We've also got a great one from Aaron Levie there, CEO of Box. But Ben Bajarin says: "Large enterprises will be well served by the OpenAI Deployment Company, but the vast majority of companies are not large. They will want and need AI deployed inside their businesses, too, at a different price point. It's a big market, very big." And it makes sense, right? If I'm OpenAI—
That's what Mark Cuban's been talking about, right? It's like, "Hey, if I was a college kid at the moment and I didn't want to go into work, what I would do is learn how to use the best models in the best way and integrate into mom-and-pop shops, as they say in the US." To go into these companies—they sometimes might not even have internet, their website might still be backdated in 2005—
Writing on pressed reed sheets.
"Yeah, let me just fax that to my accountant."
Some of these businesses, especially in the US, are still doing that. And so they obviously can't afford to use the OpenAI Deployment Company. It's not like every mom-and-pop shop is going to be using the service. But it opens up the opportunity to leverage, "Hey, OpenAI are putting a shedload of money into this. They've just acquired a business doing consultancy. They clearly see the opportunity to move into that services market. Why don't I just do that at a very small scale and build a $10 million revenue business, which in the grand scheme of things is nothing?" College kids, go out and do a consultancy business in AI.
Mark's been saying it for years.
It's genuinely true. So I think it sits adjacent to a lot of these smaller AI rollup shops that are just going—like imagine your hometown right now. Picture your hometown. Think about whether it's a village or whatever. Take a smaller area of that. I'm from a village in Manchester.
There are a bunch of companies there that could benefit from using AI. Maybe there's four or five people on the team. Maybe the top line's not that high—we're not talking hundreds of millions—but it's a business that makes money. There are people doing this already, right? It sits adjacent to the AI rollup thing where I'm going to buy some sort of small mom-and-pop shop legacy business, AI the hell out of it, increase the margins, and turn it into efficient. But doing one of those on a very small basis isn't really worth my time, so why don't I just roll up 10 of these, put it in one big thing, and then live off the profits or flip them?
Yeah, yeah. Great point from Ben in terms of there will be a huge market for this. I think we'll see a load of boutique shops set up that are doing something like this because they'll just make cash hand over fist.
But then also, not to cut you off, but the next element to that will be boutique private equity rollup firms. I guess they can be called private equity firms, but you can buy a business that's not highly leveraged and go and scale that thing without the need of employees. That's obviously been talked about for a while, but this is another step into that world.
I think it sits adjacent there. But this thing's just going to print. It makes sense for OpenAI to focus on the highest premium companies possible because the ACV for the similar amount of work, the average contract value, will just be—
We're talking contracts probably like tens, if not hundreds of millions for some of these companies, I imagine. It will absolutely print cash, and I like the fact that they've managed to beat Anthropic to market with this statement, because I know for a fact that they'll be cooking up something similar. We've seen the headlines: they're partnering up, I think Blackstone was mentioned, like a bunch of these other PE firms and also these consultancy firms. But this thing just has to print. And the company Tomorrow was a Scottish company. I'm just going to get the name right.
Yeah, Tomorrow. So they say here further down: "We've also agreed to acquire Tomorrow, which will bring 150 experienced forward-deployed engineers and deployment specialists to the OpenAI Deployment Company from day one." Great win for the UK in general. This was based in—
Do you know how big the acquisition—
No. Maybe could you Google and check it? I couldn't see any figures on Google, but—
No, I can't. I will happily Claude it, though.
Okay. But yeah, it makes sense that the labs are doing this because the labs can just capture the whole thing. But this was something that needed to happen; it's something that's waiting to happen. I think that those guys are going to absolutely print cash because, as you say, I'm a large, maybe publicly traded company. On paper, I'm telling the board that we are AI-first, but in reality, they're probably not as frontier-ready as other companies, but for their size maybe doing okay. And you approach the OpenAI Deployment Company and you say, "Hey, this area of our operations—make anything possible." You can literally make anything possible. Think about it.
Yeah. I guess. But 150 experienced FDEs all working literally at the frontier, with the backing of one of the frontier AI labs.
It's obviously a premium service. You can make magic happen. This is as close as magic as you're going to get through the magician deployment company. But that's what I mean: you've got a large publicly traded company and they say, "Look at my operations. Staff here, our supply chain is incredibly inefficient. We're doing this, this, and this at the moment; we want to bring our efficiency, quotation marks, down by 30%."
"We're going to pay you to create that outcome using AI." And that's what this whole thing will be about. What happens about—let's now—I have maybe a number here, but just quickly, what happens to the software companies then, right? I haven't seen any comments on this announcement where software is dead or whatever, and I would have thought you would have seen that. Because if you see literally one of the top AI labs, the frontier AI lab, going into services, buying up a company who already are probably doing very, very well, and directly going against software and saying to enterprise customers particularly, who are these software companies' biggest customers, "Hey, you don't need this third-party software. We can just build it in-house for you exactly how you need it. And it works better because it's integrated with all of these different—I'm working directly in your company. It's not third-party where I only get 80% or 18% or whatever percent of context. I get all of the context. I can see everything, and I can condense it all down."
Tyler Denk has actually been really vocal on this recently, basically saying they've just allowed Beehiiv to plug into—to be able to build MCP servers on top of it. And so you can now plug Claude Code into Beehiiv, and you can create your own dashboards, you can understand more analytics and everything. He wrote a tweet the other day: he was like, "I'm addicted to building dashboards." And then I looked at my workflow and I was like, "Oh my god, I'm scrolling X while I'm waiting for Claude to build my dashboard for me. I'm building a dashboard, too. I'm addicted, too." And it's true, right? You can see all of the data. The software might not have all of the context of the whole business, but you do, and the AI can now.
And therefore the FDE does.
Go and get the FDE to go and build very custom software just for your company.
Yeah, there will certainly be enterprise contracts at risk from this deployment for sure.
It's weird that I haven't seen anything about that. But anyway, Denise Dresser, OpenAI's Chief Revenue Officer, framed the mission as helping companies bridge the gap between AI capability and real operational impact by integrating AI into the infrastructure and workflows that power their businesses. Should we take a look at Aaron Levie's take?
Yeah, pulling it up right now. So Aaron Levie, CEO of Box, says: "The need and opportunity for professional services and FDEs to deploy agents right now is massive. Every tech wave offers a new era of consulting and tech services requirements. Moving from analog to digital led to a massive wave in the '90s. On-prem to cloud, same in the 2000s. This is going to be a far greater scale than others. The reason is that agents fundamentally change the underlying workflows of an organization. Unlike most prior eras of technology where it was a change in medium of the service being delivered—on-prem CRM to cloud CRM—agents will rewire the business process itself. And unlike upgrading a tech system, business processes are full of idiosyncrasies. Every industry will have its own variance, and every department within those industries will have variance as well. Not to mention the bespoke difference between firms: bringing agents to marketing in CPG will look different from marketing in healthcare. Bringing agents to sales in a B2B software company will look different from a car dealership. And none of the change is easy technically. You need to first modernize your infrastructure and data and make sure it's ready for agents. Access controls, entitlements, and permissions need to be mapped in a way that works for agents and people. You need to make sure agents have the right context to work with. You need to consistently eval and maintain the agents when there are model upgrades. And you need to drive the change management of the process itself to figure out which parts the people do and what agents do. That's an insane amount of technical and domain-specific process work to be done to make this all happen. Huge opportunity for new service providers as well as internal teams and roles to emerge to help drive this change." 180k views, very typical for Aaron Levie. He's always writing gold, but he's right.
Surprising it didn't do more, to be fair.
Yeah, same, actually. But there's a point within that that is starting to become a bit more of a recurring theme, specifically from our conversations most recently with Friedrich, co-founder of Pit and also V Technologies. What they're doing with Pit—it's your product managers, essentially a product shop as a service for your company. They are also selling outcomes within the product org. But he's mentioning here that each company will have a bespoke workflow. Each company will have a bespoke way in which they use that workflow, and that can be consistent across departments, consistent around industry, around product, around levels of seniority, right? There is always—maybe you're at company A, I'm at company B, we both use Salesforce—
—and we're both in sales—
—but we probably use Salesforce slightly differently. There will be something that we do differently compared to each other.
So the whole premise of that—
Just workflows within Salesforce will be different, but you're kind of selling different workflows together. You're constrained in the way that things work on Salesforce, on the software that you're using.
So the good thing about what—
You just got used to doing that.
Yeah, 100%. And this is then superior in a way where these people can come in and they can build this outcome around your specific workflow.
You don't have to drastically change your workflows as well. I think that's the main thing, right?
You won't have to change them really at all.
Apart from taking elements out and making it more efficient. But it won't be like, "Okay, you guys need to adapt to how this—"
"Instead of clicking this button first, now you need to click this one to go to that one."
In the most simple terms, right? All of that is just wiped out, because that's probably what all of these enterprise businesses are worried about. Salesforce is just so heavily integrated within their business that they don't want to rip it out because it's like ripping the heart of the business out, because 8,000 of their employees use it every day. But now it's like, okay, well, we can just build one that's very similar to the function of Salesforce or whatever—I'm just using that as an example—and nothing really changes. It just says your name in the top right corner instead of Salesforce's name.
And it looks a little bit better, and it functions way better, and has more context.
Yeah, yeah. It'll be on the firms that decide, "Hey, we need to move super quickly on this, and who cares about the change." If people don't like change, you're, I guess, going to be left behind.
Yeah. On the acquisition numbers, apparently financial terms of the Tomorrow acquisition haven't been disclosed publicly, but one number that has been widely reported alongside the deal is the $4 billion initial investment figure. But that figure refers to the OpenAI Deployment Company itself, not the Tomorrow purchase price. The launch of the OpenAI Deployment Company kicked off with a $4 billion investment, with the Tomorrow acquisition giving the company a base of 150 forward-deployed engineers. That is from Constellation Research, so who knows how true that is. But yeah, I want to touch quickly on the name as well: OpenAI Deployment Company. We were talking about this last week, and we're like, "Why is China naming all these incredible companies incredible names—"
It's like an investment bank.
—and the Western world just isn't doing it?
Yeah, that one I mentioned was—
The Chinese something bank or something.
Yeah, but there was another one in the US, it was like Western Cowboys Industries or something. They raised yesterday, it was like 220 million.
That 20 million addition was probably just for the name.
"Yeah, here's an extra 20 because you guys are really cool."
Makes complete sense. Can fully see that.
Exactly. Anyway, should we move on?
Yes. Before we do that: Polymarket, the world's largest prediction market. Stay informed and profit from your knowledge. Go to polymarket.com. Let's get up the next piece of news we have on, and it is Thinking Machines. Big news.
They're back. They are so back. Everyone—well, not everyone thought Mira was down and out, but in the past—well, one, no one's heard anything from them in ages. And two, there were a lot of headlines around people leaving. People were joining and then immediately leaving from other hyperscalers. But they've come back. They've come back with their new interaction model. This launched yesterday at 9:24 p.m. We can watch the demo video. Essentially, this is a model in which it can interact with those engaging with it. They've put some benchmarks in there, which we can go over, but I think maybe to start we can just watch the very short demo clip. It's about two minutes; we don't have to watch all of it.
Let's talk about the caption first: "People talk, listen, watch, think, and collaborate at the same time in real time. We've designed an AI that works with people the same way. We share our approach, early results, and a quick look at our model in action. Boom. Go."
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."
Got it. I'll say "friend" whenever one of them walks in.
Cool. 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.
Friend.
Hey Rohan, I heard you're talking about our amazing interaction model. I have a few things to add, but to make it interesting, I'll do it in Hindi. Can you translate into English in real time for my friend and for the audience?
Absolutely. I'll translate as you go. Today we're taking a look at our preview model.
It can react as you're engaging with it, right? So it can identify that this person has walked into the frame and announce "friend."
The delay is super, super small.
There were benchmarks that they released. I can't remember what the name of the benchmark testing that it's called, but I think it's in the piece. Pause that. And it was quicker—theirs was recorded at 0.4 seconds, which is quicker than the current ones that have been put out by Google and I think OpenAI as well. But they basically released a thread on everything.
Today's models are apparently turn-based. So when you finish talking, then the model talks, then you talk. While the user is typing or speaking, the model perceives nothing. While the model is generating, it can't take in new information. Thinking Machines argues that this is a bandwidth bottleneck. The system has two coordinated models sharing context: an interaction model maintains continuous real-time presence with the user, and then a background model runs asynchronously for deeper reasoning, tool use, and longer tasks, which results stream back into the live conversation. So the interaction model and then the background model—does that not mean that you're using double the compute at the same time?
I'm not sure how the technical elements play out.
Probably naive, yes, I don't know. But they position it where it can interject. Typically, you'd be chatting about something and the model wouldn't cut you off. You can just cut it off however you want to do; as soon as you make a noise, it stops.
This one can apparently interject. It can kind of cut you off sometimes. It's meant to be more humanlike. And they've put out a bunch of threads of examples of what it's doing. So here we have: the interaction model can track when Lillian here is thinking, yielding, self-correcting, or inviting a response. There's no specific built dialogue management system. Again, comparing it to traditional models in which if you ever make a noise or try and cut it off, it'll just immediately stop talking. So it's meant to be more humanlike in that regard. This is still in beta; it hasn't been publicly rolled out yet. They're testing it with a few select partners. There were a bunch of comments online where they were like, "Labs should start hiring professional actors to run their demo videos to make them more engaging."
No, I think it adds a human element, the fact that they're stumbling and the fact that I would stumble when I talk to Claude. I'll stop talking midway through talking to Claude and then be like, "What am I going to say now?" and then go and say it. There's a nagging example here that I haven't actually watched, so we can watch this live right now.
Sit up straight and you'll be golden. You're starting to slouch forward. Try pulling your shoulders back. Much better. You're upright again. Tilting your head like that—try keeping it level. There's that forward lean again. Sit back into your chair.
So that really is super sensitive. That's really cool. It's like visual and audio models combined in this regard, which is really interesting.
Who's that French developer that made that model?
He's like the French—or no, I think Pieter Levels is French. Is he? I don't know.
He's a French guy, like a French Pieter Levels. I have no idea who that is.
He's got hundreds of thousands of followers. I can't remember his name. Anyway, he recently launched this model. He was on Starter Story the other day. Give me a sec.
They've shared a bunch of examples here. You know who I think this would be perfect for is the Synthesia team. They're rolling out avatars to FTSE 100, NASDAQ companies.
Marc Lou, you know Marc Lou?
So Marc Lou recently released an app, basically, that tells you if you're slouching or not.
Do you remember? His launch video was him running on a beach or something, I'm pretty sure. And it had almost this exoskeleton of your shoulders and your arms and your chest and then your head. And whenever you would move out of the way, if you only moved out very slightly it would go orange, and then if you were slouching at your laptop it would be red. And you had to sit there and it would use your camera and show you how you're sitting and what's good and what's not.
Yeah. Oh, cool. It's pretty similar to that.
Yeah. Well, I would love that in my life. Very often—I mean, my setup at home is dreadful. My desk setup is great, but my chair is just a wooden chair. I think it's for a garden, to be honest.
The chair is the most important bit.
It is. I know, I know. But very often I'm slouching.
That's fine, as long as you're getting good work done, doesn't matter. Anyway, let's move on. Base, the infrastructure for the onchain world. Base is built to empower builders, creators, and people everywhere to build apps, grow businesses, create what they love, and earn onchain. Go to base.org. Let's move on to OpenAI versus Musk, the final week. Okay, this is your domain.
We've got a few pieces to go over here. Firstly, Satya Nadella in that picture—he looks like a gangster. A lot of eyes on him post-Sam Altman and Mira Murati texts being leaked, in which they were trying to contact Satya during the firing of Sam Altman. They were trying to contact Satya and he just wasn't picking up the phone, or he wasn't replying on text or email. People were saying essentially, "Oh, Satya is aware of the fact that if there ever was to be any sort of litigation or event in court, they would have a right to access this information and he would have his name getting involved in these conversations." So everyone's saying he will only do these types of conversations over the phone. A lot of memes were thrown out there, a lot of AI cartoons were made around Sam Altman trying to reach Satya Nadella, which is funny. But he takes the stand. This is coming towards the final week of the Musk versus Altman trial. We also saw Brockman sit down and get pretty grilled again. But Satya steps in to testify about Microsoft's partnership with OpenAI and how that relationship was working. And there are some really interesting quotes in here. After Altman was briefly—sorry, this is by Angel Au-Yeung at the WSJ. After Altman was briefly pushed out of OpenAI by its board members in November 2023, Nadella said he pledged his support to potential new leaders, but also wanted to ensure that Altman and others didn't create a competing company or join Google. "That was obviously very concerning to me," said Nadella. "Given all of that competition, I just wanted to make sure we could hang on to the band that created all this technology." Musk's team has sought to paint Microsoft as a powerful company exerting control over OpenAI's fate due to its billions of dollars of investments in the startup. Musk, back in the day when all this was happening, was tweeting out tweets along the lines of "Microsoft owns OpenAI."
"They're completely constrained by Microsoft. They can't do anything. They literally own them." Musk attorney Steven Molo pointed to a news release that OpenAI's board published shortly after Altman's firing in which they said he hadn't been consistently candid in his communications with the board, hindering its ability to exercise its responsibilities. "Pretty definitive statement as to why they fired him, right?" asked Molo. "Yes," replied Nadella. Sutskever, an OpenAI co-founder, also gave brief testimony Monday about the role he played in Altman's firing and disclosed the value of his OpenAI stake is worth roughly $7 billion. It also came out—we can talk about this, we have some more notes on this—another report on Ilya's—essentially it came out, it was relatively widely known, but he spent essentially a year planning the ousting of Altman as well. But shortly after Altman was kicked out and they were replaced with—what's his name, the previous CEO of Twitch—Brockman actually left following that; he went with him. But Musk is suing OpenAI and its leaders, as we all know, Altman and Brockman, for allegedly manipulating him into giving tens of millions of dollars to the nonprofit organization. This is what we've touched on. This is the whole premise of why this trial is happening. Essentially, it's been changed from a nonprofit to a for-profit. He's put in something around like 130 million or something. He's saying that this isn't fair. The tides have been shifting back and forth, but ultimately I think the group consensus is that Musk has a claim here; Musk is kind of on top. During Nadella's testimony, he made clear to jurors that Musk never reached out to him to express concerns when Microsoft made its initial investment in 2019 to OpenAI, nor when Microsoft announced its exclusive license for GPT-3 in 2020. "Does Mr. Musk know how to contact you?" Microsoft attorney: "We have each other's phone numbers," said Nadella, glancing at the jury. "It has nothing to do with control," said Nadella on Monday during his testimony. "It had to do with just the strategic partnership, which gave us the rights to the IP." So this is in regards to a quote from Nadella who goes—attorneys for Musk have pointed to an interview that Nadella gave shortly after Altman's short-lived firing, often called "the blip"—it's referred to as the blip in budget testimonials. Nadella said, "We are below them, above them, around them," referring to OpenAI, the relationship of Microsoft with OpenAI. And then he's come out and said, "It has nothing to do with—"
Said, "We're omniscient."
Yeah, literally like they are omnipresent. They are—
Omniscient is all-knowing, omnipresent is all-present.
Omnipresent. Present, all-present. Everywhere.
Which is just a nuts quote to get brought up in the middle of the thing. And yeah, he said it had to do with just the strategic partnership which gave us the rights to the IP, but not about the control. Lawyer Molo pressed the Microsoft CEO on his influence on OpenAI, including selecting new board members after Altman was reinstated as CEO of OpenAI post-blip. A group chat exchange between Altman, Nadella, and Microsoft Chief Technology Officer Kevin Scott flashed on the screens in the courtroom. Man, I would love to be in this courtroom. So annoying that we can't see anything that's happening. You can actually listen to live audio, by the way. On the TV screens in the courtroom, Altman proposed a number of names as potential new board members, including Sue Desmond-Hellmann, who Nadella voiced approval for in the group chat. Desmond-Hellmann joined OpenAI as a board member in 2024. Sutskever, a prominent—we all know who Sutskever is—he recounted how in 2023 he told OpenAI's then-board that Altman's dishonesty had created an environment at OpenAI that was not conducive to creating safe artificial general intelligence. He came on and said Altman was sort of playing executives against each other, he was being dishonest, he was manipulating people.
This is before he got kicked out.
Yeah, building a case as to why Altman should not be in the position he is as CEO of OpenAI. It also came out further down in the trial that Sutskever, who was previously working at Google, during his testimony shared that the search giant Google had offered him $6 million a year to stay in 2015. So that's obviously mad money now, but even back then it was mad money. But he didn't take this. He ended up going on to found OpenAI. But they were offering him that much money just to stay, which is crazy. Now we've also got another write-up from someone who was present at the trial. This is a thread on Brockman from Gil Alfaro, and it's beautifully written as well, I must say. It kind of reads like a bit of a novel, which I'm going to go along with because I like it: "In the federal courtroom in Oakland this afternoon, the weight of OpenAI's founding promises hung heavier than usual as Ilya Sutskever, co-founder, former chief scientist, and one of the most brilliant minds in artificial intelligence, took the stand." And I love how they mention this again because this came up in another article: "Dressed in a light blue button-down shirt—" He's always dripping, you know? He's always mogging. "—unblinking gaze, Ilya appeared dejected about his estrangement from the company he helped build. He barely made eye contact during his roughly one-hour testimony, speaking in a commanding baritone that carried the quiet authority of someone who once poured his life into the mission." This does seem relatively broken-coded. "Elon Musk's legal team called him as a witness, even attempting unsuccessfully to treat him as hostile due to his massive financial stake. Yet what unfolded was a masterclass in the tensions that have defined this trial: the original nonprofit dream versus the billions now at stake. Ilya recalls early recruitment meetings with Elon Musk. Musk's lead lawyer, Molo, began with the early days, November 19th, 2015, roughly 3 and 1/2 weeks before OpenAI's public launch. Molo walked Ilya through these formative recruitment meetings with Elon. 'Did you have regular meetings with him?' Ilya confirmed they did. The questions turned to departures like Andrej Karpathy leaving for Tesla's self-driving efforts. Ilya admitted he was 'definitely a little bit bothered, but at least it wasn't the end of the world.'" Now, this part is really interesting. Ilya comes out and says how much he rated Musk as a CEO, how good of a CEO he thought he was. And this is funny because now they're suing each other, right? In 2018, an email from Ilya to Elon Musk, already admitted into evidence, Molo read the key lines aloud: "You quickly—"
pushed me out of my academic comfort zone. With time, I grew to appreciate the vast depth of your strategic insight. I can think of many examples. And then we come to the final bit. It helped that we have the most overwhelmingly competent person in the world helping us accomplish the task. The task being, you know, leading OpenAI. Ilya didn't dispute the words. This was brought up in front of Elon. Ilya explained that he viewed Elon as by far the most competent, the strongest founder CEO in the world during the 2015 to 2018 fundraising period. When pressed on the strategic insight reference, Ilya said it was about making high consequence decisions. Well, the mission statement from OpenAI's December 11, 2015 announcement was displayed, which was a nonprofit artificial intelligence research company. Our goal is to advance digital intelligence in the way that it was most likely to benefit humanity as a whole, unconstrained by need to generate a financial return. And this is a core theme throughout, and especially when they were questioning Nadella. Nadella sits on the board of another nonprofit charity, and they were kind of using that against OpenAI as an example.
Well, it's a comparison, right? How do you run that? Do you suddenly turn that one into a for-profit company? And then Nadella's like, well, you know, there are—
Do you raise multi-billion dollars for this?
Are you able to raise multi-billion dollars because it's a nonprofit, because all the investors aren't going to make any money?
Yeah, literally no. And I think it's on the board of some sort of educational conglomerate. It might even be a university, I can't remember, but they're using that as an example. And then he starts explaining that sometimes a nonprofit will have a for-profit arm.
Sure, but obviously pales in comparison towards it would be a smaller part of the business, as opposed to OpenAI would be a massive for-profit conglomerate as opposed to—
I assume the function of the for-profit would be purely to help run the nonprofit, to fund the nonprofit.
Whereas this for-profit is to fund the bank accounts of many investors.
Yes. And so two final points to share here, one of them was like a really big reveal in the trial. Ilya revealed his stunning $7 billion OpenAI stake. The mission of OpenAI is larger than its nonprofit or for-profit structure. He said firmly he opposed Elon's early proposal to merge OpenAI with Tesla, calling it something that would kill the dream. When asked about Musk's push for majority control, Ilya described it as feeling aggressive given Elon's other massive responsibilities. He also revealed the staggering personal price he turned down from Google, which is the $6 million in 2015. And in a moment that underscored the scale of what OpenAI has become, Ilya confirmed under oath that his ownership stake in the company is now worth roughly $7 billion, which is insane. This final bit has an amazing quote in it. The judge was asking about the progress that they were making in AI since 2018, not just at OpenAI, but generally. Judge Yvonne Gonzalez Rogers interjected with a rare question showing her own curiosity about the technology's leap. 'What was the technology like at that point?' she asked, referring to 2018. Ilya's answer: comparing AI then to now was like the difference between an ant and a cat without massive funding. He noted there is no big computer. That made a bunch of memes already surfacing basically of this.
Why a cat? I guess a cat is more adept than an ant.
They're great builders, though.
They are, and they're incredibly strong given their body weight.
So have they actually gone down?
Maybe a cat-sized ant would be pretty formidable.
Well, this is an interesting question actually. How many—yeah, it's kind of like how many humans to take a gorilla? How many ants to take on a cat? Do you think—
About 2,000. I imagine if ants are like there to go and, you know, kill the cat.
That's quite a lot though, right? 2,000.
I don't know, cuz they just need to engulf the body of the cat and it's game over.
Can we check? Can we check—
See what the general consensus is on ant versus cat.
2,000 seems like a lot of ants, but I guess it depends on the ant as well cuz you have worker ants.
Normal ants. I don't know what they are.
Fun thought experiment. Let me think through it. Typical ant can carry/lift maybe 10 to 50 times its body weight. But killing is different from lifting.
Realistic ways an ant would kill a cat are venom/formic acid or overwhelming bites, not brute force.
A swarm. It would swarm the cat.
Let's just—give me a number.
Just say to pick up a cat because you don't want to necessarily kill the cat.
Okay, pure lifting problem. Much cleaner. Average cat is 4.5 kg. A common ant is 3 to 5 mg and can roughly carry 10 to 50 times its body weight. Let's say each ant can contribute 50 mg of lifting force.
Oh, it's going to take 90,000 ants. So, in theory, around 90,000 ants can lift a cat off the ground if they all pulled their weight perfectly. But realistically, you've got a surface area problem. A cat's body doesn't just have room for 90,000 ants to physically grip it at once. They'd be stacked on top of each other, and only the ants actually touching the cat can lift the cat. The cat's outline gives you maybe a few thousand square centimeters of grippable surface.
You need like a flatter cat.
Yeah, or a lying down cat. So, the practical answer is more like a few thousand ants doing the lifting, each needing to handle a much bigger share of the load, which means in reality, ants of normal strength can't lift a cat at all. No matter how many you bring, you hit a surface area ceiling before a strength ceiling. Anyway, super fun.
Fair enough. No, that makes sense.
Now you know. Now you know, everyone.
Let's move on to something else. It's good that we got that down.
Yeah. Airwallex. Leave outdated business banking behind with Airwallex, the all-in-one financial platform built for the pace of modern business. Multi-currency accounts, corporate cards, expense management, online payments, all-in-one platform. Go to airwallex.com. Let's talk about the Anthropic news. This is really, really interesting. And I'm not sure why this didn't go more public, but Anthropic have addressed the unauthorized stock sales, essentially everything that's going on in the SPV market, buying secondary shares of Anthropic. Obviously Anthropic is a private company. You can't get direct access obviously unless you are an employee or an LP into a fund that has invested in said company. And this is from Gabriel Shapiro. He said, 'I am surprised more people are not paying attention to this update from Anthropic on its stock policy. This seems like a potential bombshell.' Before we do that, let's quickly just read the pictures. Let's go to the pictures. 'Unauthorized Anthropic stock sales and investment scams. We are sharing important information about investments in Anthropic to help protect individuals from potentially invalid transfers or investment fraud. Transfer restrictions. Both our preferred stock and common stock are subject to transfer restrictions contained in our bylaws. Any sale or transfer of Anthropic stock or any interest in Anthropic stock that has not been approved by our board of directors is void'—which is the key part and we'll get to that—'and will not be recognized on our books and records. This means that if someone—'
I'm trying to fit it in the screen.
We're just going to not have this up with you.
Just don't have that. It's fine.
'This means that if someone purports to sell Anthropic shares without proper board approval, that transaction is invalid. The purported buyer would not be recognized as a stockholder of Anthropic and you would have no stockholder rights.' They go on to say, 'We do not permit special purpose vehicles'—which is SPVs, which is basically everything. If you're going to do some form of secondary stock buy, you would use an SPV and a manager of an SPV. I'm kind of surprised that there isn't less regulation.
I'm surprised it's taken this long to come down on it. Everyone we've spoke to that we've looked at, whether it's token sales, whether it's like—
I can't remember who, maybe it was Robinhood who initially did like the Anduril / OpenAI one and then Anduril came out—
There's a whole market of them, right? There's like Forge, Sidecar. There's a bunch of secondary market platforms.
So the value then is purely the perceived value of that ownership that isn't even real.
Some companies—Anthropic are just saying all of it's void, which is actually very big news. To get access to shares in private markets, you go through an SPV, and most of them are legit, right? The person running the SPV has direct access and they can create their own little market of many LPs buying into their shares. And it works super, super well.
And this is kind of what it goes on to say. Anthropic is one of the only three pre-IPO companies alongside OpenAI and SpaceX to reach a trillion dollar territory on the secondary markets. And that's obviously huge. So there is going to be a clampdown, and I wouldn't be surprised if you see OpenAI and SpaceX particularly before their IPO clamp down on this as well. Anthropic's transfer restrictions embedded in its corporate bylaws now nullify the share transactions on secondary platforms. The main thing here is the void thing. The policy covers direct sales, beneficial interests, forward contracts, special purpose vehicles, and tokenized securities, which is what you were just talking about. Any transfer without explicit board approval is declared void, which will not be recognized on our books and records. And void is like the key thing. On the actual post, he goes on to say, Anthropic is threatening that—and they might be bluffing. If they're not bluffing, what litigation is likely to ensue? This can get into really esoteric areas of corporate law that depend on exactly how the transfer restrictions are drafted, as well as the language around how violations of transfer restrictions are treated. For example, if they are merely voidable, then downstream buyers can assert various equitable claims or defenses. But if they're just void ab initio, then in some jurisdictions that forecloses equitable defenses. Because the statement from Anthropic has said, if not approved by board of directors, it is void and will not be recognized, that basically makes a blanket statement: no.
Well, people are making insane amounts of cash trading these.
Yeah. Well, because the secondary markets have a price gap. In the secondary markets, Anthropic is valued at over a trillion dollars, and the actual—what was the most recent valuation of Anthropic? It was $350 billion, right?
So there's a price gap between Anthropic's primary funding rounds, which are about $300 to $400 billion, to their secondary OTC market. And that's obviously driven huge indirect interest. I think those secondary private markets are incredibly unregulated. It's like someone's just set up that type of thing to trade this stuff on there. People that are buying into it aren't buying any Anthropic stock. They're buying the opposing person's bet essentially. I'm pretty sure they're not regulated.
No, but most of them actually own selling secondary shares worth of stuff.
But what are the broader implications, right? Voiding these transactions at scale could freeze secondary trading activity in AI shares more broadly as buyers reassess the enforceability of transfer restrictions at other high-profile companies. Watch for whether OpenAI, SpaceX, or others follow suit, and whether the Delaware courts ultimately uphold Anthropic's position when, not if, litigation follows. Will you see more of this kind of stuff happening? Probably. Particularly when there is so much hype around all of these large players, especially before they're about to IPO. Everyone wants a slice of the pie, and therefore they're going indirectly through it. Do these people who they're going through actually own the shares? Probably not. And then you're going to have that added litigation or restriction element where the shares that you've bought are completely void.
A lot of people are going to be unhappy with that.
Let's move on to this last bit of news that we have at the bottom. We can touch on that actually. I do want to do the Michael Burry thing.
Let's do that quickly. We got 15 minutes.
Okay, nice. I can get that up in a sec.
Okay, while you do that: Framer, build better sites faster. The all-in-one platform where enterprise teams design, build, and ship production-ready websites. No dev handoff, no bottlenecks, no creative compromises. We did indeed migrate ETN to Framer ourselves, and it was super easy. Go to framer.com/etn.
We don't have to pull it up, but let me—oh, there we go. Everyone knows Michael Burry, the Big Short investor. He is claiming that the end is nigh and we are about to see a massive sell-off, a massive reversal given the current conditions that we're seeing in the stock market. Burry essentially writes, 'The end is nigh.' That's the warning from the Big Short investor Michael Burry, this being the stock market's blistering rally to record after record in recent weeks. Burry said he believes the market could finally be on the edge of the major decline he's been warning about, telling readers on his Substack in an early morning post on Monday that history seems to be repeating itself. For those wondering, the Substack is paid and does very well. 'With what is happening in the market in the last week, I had lived this before, suddenly dawned on me,' he wrote. 'The NASDAQ 100 complete reversal.' The former Scion Asset Management chief pointed to the major rally in recent weeks with indexes climbing higher on strong earnings and optimism for a peace deal with Iran, even with no agreement yet in sight. He gives some figures about the S&P performance year-to-date, like it's just basically absolutely mooned since the 29th of March. 'I'm calling something. The market has jumped the shark,' Burry wrote later, speculating that the conflict with Iran, the recent surge in oil prices, or a potential contagion event in private credit could all be triggers to a coming decline. There's a lot of chat about this guy calling another big thing that's going to happen every three months, and he usually is wrong. But he does address the fact: 'I know that people will call me the boy who cried wolf,' because a broken clock is right twice a day. I remember in 2021 where he publicly called out a massive Tesla short position.
And then Tesla stock mooned as soon as he announced his massive public short, but then about a month later, it crashed and it fell off. So it always seems to me that he's always slightly too early. Not that everything he says is always true or always happens, but it does seem to kind of happen just a little bit after he claims.
He's become the guy to call it first, right? So he's going to have to be early, otherwise he's late.
Yeah. And he also references the semiconductor stocks that everyone's watching moon right now and everyone's getting rich. Burry pointed out in particular to the tech and semiconductor sectors, which have been on a tear amid fresh deal-making and strong earnings over the past month. He references iShares Semiconductor ETF is up 65% year-to-date, one of its best ever rallies. The NASDAQ 100 is up 16% in the last month alone. In a post last week, Burry flagged an analysis from analysts at BTIG that showed the market by some metrics had eclipsed dot-com era moves. According to the analysis, the top 10 performers in the NASDAQ 100 were up an average of 784% over the past year. That's insane.
That compares to the average 622% gain across the top 10 performers in the year leading up to March 2000, the month before the dot-com bubble began to deflate.
So he's saying it's bigger, it's worse, it's scarier than what we saw in the dot-com era. He pointed to several charts that showed the complete reversal of the major stock indexes following the dot-com bubble in 1929, labeling each peak with 'you are here.' That's such a crypto thing to do. 'The market environment looks so familiar to the months leading up to the burst,' Burry added, suggesting that investors consider selling big momentum stocks or reducing their exposure to the tech sector. What about you? You going to be selling your positions? You going to be selling your stocks?
No. You've got diamond hands. I'm going to HODL. I have time.
Time in the market is better than timing the market. Am I right?
Burry acknowledged his reputation as the boy who cried wolf in markets, referring to his long history of making bearish calls on the market. He argued that many of his social media posts in recent memory have been framed by the media as stock crash warnings, though he wasn't calling one.
That's the problem, right? Whenever he comes out, it's an opportunity to get a scoop on being bearish on the market before it actually crashes or whatever. Fuels the fire.
He goes, 'I am now a meme for the number of times I have called a crash.'
He's going to be right at some point.
So yeah, Michael Burry telling you to sell everything. Do you take his advice? Do you not? I'm kind of enjoying this wee little ride that we've been on. I am going to sell my Apple stock, though.
You're going to sell it? I would have sold Apple stock ages ago.
I'm not that bullish on Apple nowadays.
Not really, no. I'd rather be exposed to some form of AI infrastructure industry that's doing something a little bit AI. I just feel like it's a waste of allocation.
Damn. Cool iPhone you've got there.
Not necessarily bearish, but you know.
Cool MacBook you've got there as well.
Yeah, look, they're doing what they do well, right? But they've been doing that for 25 years. Let's talk Helsing real quick.
How much time do we have before VB comes into place?
Not that long, 10 minutes.
Okay, nice. Before we tell you about Helsing, let me tell you about our friends over there at Framer. Framer, build better sites faster. The all-in-one platform where enterprise teams design, build, and ship production-ready websites. No dev handoff, no bottlenecks, no creative compromises. We migrated ETN to Framer ourselves seamlessly. You can too at framer.com/etn. Let's talk about Helsing, or we could go into the Town Square.
Let's go to the Town Square. Let's talk about the Signal tweet because I think it was a really cool tweet. 'Raising capital is more of an art than it is a science. The binding constraint at the early stage is almost entirely narrative compression, status signaling, reading the room, which is way closer to stand-up comedy than finance.' What's your take?
Yeah, I like it. When we raised—obviously we didn't raise that much money, right? But whenever everyone asked, 'How did you do it?' Raising $150k for a podcast that didn't even start, or a live show, was kind of impossible. That's why it was fairly interesting. And they were like, 'How did you do it?' And it's like, well, it is literally about reading the room, right? It's a power law. That's kind of what I took from that.
Yeah. It can't be a financial decision really cuz you don't have any metrics to look at, right? It's just a bet on the person, and you need to be talking to the right people as well. You spend days or weeks chasing someone who would just never ever back you anyway.
Sure. The thing that came to mind was the picture of the jester dancing in front of the princess. There's definitely a little bit of that.
I think it was Signal who replied to that tweet cuz it went viral again on the weekend. The meme with the jester dancing in front of the queen, and the VC is the queen and the jester is the startup founder. Someone, I think it was Signal, quote-tweeted the tweet and said, 'Only raise when you can be in the seat of the queen.' You want VCs to be dancing at you, you want them to be the jester, you want to be the power person.
Indeed. When you are the queen, it is way easier and way better because you have so many more choices of somewhere to take capital.
Message: be the queen. Indeed. We've also got—this is quite funny and an interesting thought experiment. Growing Daniel tweeted from the past. He says, 'Scheduling this tweet on—that'll be 11th of the 2nd—90 days from now. Hello from the past.' This is a quote tweet from Nikita back on Feb 11th. Prediction: 'In less than 90 days, all channels that we thought were safe from spam and automation will be so flooded that they will no longer be usable in any functional sense: iMessage, phone calls, Gmail, and we will have no way to stop it.' I think if I remember correctly, this is probably around the time of the Moltbook craze.
Which seems quite recent, but I guess it was around then. But it's interesting, right? Because if we look at the channels—
