Musk’s X Money Plans, Atlassian Buys The Browser Company, Google AI Chips vs. Nvidia | Sep 4, 2025

0:13Welcome everyone to the information's TITV. My name is Akash Basrich. It is Thursday, September 4th. We have got a busy, busy show for you today. We have a story about Elon Musk's progress on making X the everything app. We've also got SignalFire, a very cool venture capital firm, coming on the show to tell us about its data-driven approach to venture capital investing. Rogo is coming on the show to talk about its acquisition of AI spreadsheet company, Subset. We're then going to talk about Google's progress on its own AI chips. And finally, do not go anywhere because the head of product at Atlassian is coming on to talk about the company's acquisition announced today of the browser company.

0:54There is a lot going on. Let's get right into our first segment. Elon Musk has plenty of experience working with regulators with many of the companies that he runs. This morning, The Information published a story about the challenges that one of his companies has had with financial regulators. It is a story all about Elon's big plan to turn X into a payments platform. And I want to bring on Theo Waite to tell us more about what he's learned. Theo, welcome back to TITV. It's great to have you. Thanks for having me. So I want to rewind just a few years. Elon Musk buys Twitter. What were his ambitions for Twitter now X on turning that into something that was bigger than just a social media platform?

1:35So Elon really wanted Twitter to become what he called an everything app, which already exists in China in the form of WeChat, but doesn't really have an equivalent in the US or Europe. He wanted it to be a place that you can turn to for everything, paying your friend back, buying stocks, talking to your mom, finding a job. Basically, Venmo plus Robinhood plus Facebook plus LinkedIn plus a ton of other apps just all rolled up into one. And the core of that strategy would be payments because once someone trusts you with their money and, you know, lets you have that level of access into their lives, it's much easier to sell them on using other things too.

2:14So what happened to those plants? So in order to be a payments company in the U.S., you need to get these things called money transmitter licenses from separate regulators in pretty much every state. And XRD has gotten a ton of these licenses in 41 states, including California, Texas, Pennsylvania, over the past couple of years, but it still hasn't been able to get approval in New York, which is obviously an incredibly important state if you're doing anything in payments and finance. And one option that was discussed inside X was launching a limited version of the service in a few states and just doing a proof of concept and then going to New York and saying, look, we know how to run this kind of service, so you should give us approval, and that works in some cases.

3:03But we understand that Elon really wanted to launch in 50 states or none because he wants to go all out and thought it would be silly to do anything less. Why exactly was it that New York regulators were unwilling to give them the license? So for what it's worth, you know, New York declined to comment for this story, and, you know, they haven't really said anything in an official capacity about X. But my understanding is that But one of their concerns, you know, kind of comes from the premise of a super app in the first place. Like, think about the times when you've had to verify your number to log into your bank account or, you know, upload a picture of your passport to Venmo or things like that.

3:43There are these layers of verification that exist that companies don't like asking people to do but are essential to root out criminals and comply with the law. um that's kind of inherently a bit at odds with elon's vision of making x money as seamless as everything else on twitter where you just you know it's like sending a dm to send someone a bunch of money um and regulators are going to have a lot of questions about that um the other big issue is cost cutting and turnover you know elon famously fired a ton of people when he came into twitter and likes to run his company's lean um and mass firings don't really give regulators confidence that you have enough people on hand to comply with the law and are going to have consistent leadership.

4:26One of the questions I had for you is how much Elon really cares about this now. I mean, he's so distracted with XAI and humanoid robots at Tesla. I mean, is this even a priority for him right now? I mean, it's important to remember that he had this vision for Twitter when he acquired the company in October 2022, which was literally like right right before ChatGPT came out and kind of changed the whole tech industry's priorities. And he's also been involved with the election, all these other things, obviously. But this has been an obsession of his for more than 25 years. He first founded a company called X.com in 1999 that would go on to be morphed into PayPal and Elon was kicked out basically.

5:08So I wouldn't count out that he's still going to come back and be obsessed with this at some point. And for what it's worth, you know, X is still ostensibly moving forward with us. Right. Last question for you. We were talking this morning in the newsroom about, you know, what this could indicate to us about Elon's management style. You had some thoughts there. Tell me about what you were thinking. I mean, I think it just shows that, you know, moving fast and breaking things is his style and that that has, you know, worked for him in the past. But when you're dealing with 50 different state regulators, you know, that that can be a difficult approach to take.

5:43and I'm sure there are a lot of people in the company that wish he would have been a little. Regulators do not like they move fast and break things. They like they move slow and follow the rules. That's right. All right, well, Thea, thanks for coming on. It was a fascinating story and I think there's probably more to learn here. So we look forward to having you on again. That is Thea Wait, who covers all things Elon Musk at The Information. Well, SignalFire probably isn't the first name that you think of when it comes to venture capital, but it is a powerhouse fund that counts big startups like Grammarly and Rowe as part of its portfolio.

6:20And joining me now is Chris Farmer to discuss his data-driven approach to venture capital investing. I'm really excited for this conversation. Chris, welcome to TITV. It's great to have you. Thanks so much for having me. So, you know, you built SignalFire on this idea that you can use data to really pick the right winners in venture capital. And you've been at this for quite a few years now. I'm always curious about what data specifically VCs are focusing on. And so if we take 2025, you know, the past eight months, I guess, what data have you been really obsessed with this year that you think people should be paying more attention to right now as it relates to startups?

7:02Yeah. So from the beginning, we started the firm about 2013. So we've been applying data for a long time. It really depends on the sectors you're talking about. But obviously with AI, these are very key talent-driven companies. The people who have had a chance to see the future, who have been working on these foundational Gen AI technologies, are the key talent to follow. When you see that happening through the types of salaries that are getting paid by places like Meta, but just the sort of the huge scarcity of talent that has this experience. So a large part of the system from day one has been all about following the talent.

7:50Where that talent goes is a leading indicator of where the world is going. And it's sort of foundational to being able to execute on building these technologies. And so when you say data associated with talent, I mean, what kind of data are we talking about here? What are you tracking? So we track about 650 million people. almost every engineer in the Western world. So we're tracking everything from, it's like a modified page rank, basically how Google uses websites and hyperlinks in order to track. We're tracking people and companies as nodes in the network. You're basically tracking where people are getting hired and what companies people are leaving?

8:27Yeah, where they're getting hired, how fast they're moving up in those companies, any patents, academic publications, open source work they've done, academic pedigree, almost anything in the public domain. I mean, at this point, we're doing psychographic studies of them using LLMs of anything they have in the public domain. I mean, it's a pretty extensive system. And broadly speaking, given that you look at all this data as such a, you know, from a bird's eye view perspective, what are you seeing in terms of people moving away from big tech companies? Or is there a movement to move back to big tech companies?

9:01What are you seeing there? Well, there's certainly been a huge movement. We put There's been a number of new trends. I mean, one, there's been a migration back to San Francisco after a bit of an exodus post-COVID. I think people want to be in the action. They want to be at the epicenter of what's happening. And I think it's really drawn talent from folks that had gone to Texas or Florida or places like that. They needed to be back in the mix. And they're joining startups there? They're looking, I mean... Yeah, it's interesting that it's sort of complex. A lot of, there's fewer entry-level jobs being hired, but there's lots of people migrating back to startups, but also to some of the bigger companies like Google and Meta and whatnot that are really staffing up around these sort of formational AI technologies.

9:51So I think talent is dispersing within the Bay Area ecosystem, but there's a migration to the Bay Area from all over the world. Most people want to be in the action. So you haven't been able to definitively say, you know, in 2025, hey, there have been more people starting new companies or joining early stage companies compared to 2024, have you? I'm just trying to sort of get a snapshot here about where we are in terms of people leaving big tech or joining big tech. I mean, it's a bit of both. There's definitely a lot of folks joining startups and there's been folks from all over the world. The top talent has definitely clustered here.

10:38And so I think the net hiring overall has increased. So yes, there have been more people starting and joining companies in the last couple of years. but there's also been some hiring in certain areas at big company but they've also been shedding jobs for all. So it really depends on the function. So it's a hard question to answer precisely because it depends on the definition of what is an engineer and what types of roles and functions. But certainly there's been an increased concentration of the best technical talent in the area and New York as well. I want to ask you a question. You're a fundamentals guy.

11:17You root yourself in the data. What's your take on these AI startups that fetch valuations, 10, 20,$30 billion with no product, no revenue, just star researchers? What's your take on those? Look, I think that's a very hard way to make money consistently. As an investor, I think it's often an unnatural act also as a startup to start with this unbelievably high watermark of what a company is worth. It makes it, I think it has perverse impact on who you hire and what their incentives are. I mean, it sets the bar so unbelievably high. It's not sort of not a natural migration or evolution of a company.

12:02And so, you know, I think there's a fair amount of FOMO and some of these crazy valuations that are pre-product, pre-revenue, I think unfortunately will not end well. It'll be difficult to recruit and retain people because those expectations have been set so high. And the opportunity cost is very high right now for great talent. And so you're seeing even with these huge pay packages, people go to like a meta and then leave within months, even though they're astronomical figures that would just, you know, justify staying almost under any circumstances normally. So, you know, I really think it's a challenging time.

12:39I think the fundamentals of building a company are still the same as it's always been. You need to build around a core team, build a culture, figure out what the product is. I think we're in the mystery phase of AI, not the puzzle phase. We got to the place in SaaS where it was sort of a defined problem. You could figure out what it was like as an investor and whether there was product market fit in AI. I think we're all learning, and it's hard to see more than 10 feet in front of your face. So I think from that standpoint, you need time to evolve, time to figure things out and see what works, time to build a culture and a great team.

13:15And I think if you try to skip those steps with too much capital and too high evaluation, it usually doesn't end well. Last question for you. I just want to go back to what you said. You said the mystery phase, but not the puzzle phase. Just explain a little bit more what you mean by that. I mean, I think we got to the point where you could look at CAC LTV or the the customer acquisition cost or the lifetime value, customer retention, et cetera, and see whether a company was working. Right. I think today in AI, we're still trying to figure it out. Is it the foundation models where the value is going to accrue or is it the wrappers?

13:50So we don't even know what data to pay attention to is basically what you're saying. And then a new technology leap forward will happen every couple of weeks and we're moving in another direction. And then the pricing of tokens changes dramatically. Right. Right. And so, you know, I mean, it's very hard to make, you know, very large, high conviction bets in an environment where you don't have much visibility into what's coming over the horizon as a founder or as an investor, which I think, you know, I think that that is one of the things to your prior question. Right. It's difficult to over-raise and be overly top-taking valuations because we're all figuring it out, whether it's an enterprise or a startup or the foundation models themselves.

14:37Well, great. Chris, I'll tell you what, next time you come on the show, you told us about the people that you're tracking. I want to hear some more about that data, because like you said, if a data-driven investor is telling us that, hey, we don't know what data to look at in the AI era, then I think that leaves a lot of room for great discussion about what else we should be looking at. So next time you come on, we'll talk about that. That is Chris Farmer from SignalFire. Well, AI is certainly coming for all industries with its data, including the finance sector and professional services at large.

15:07But before that happens, AI needs to get a lot better at the simple task of generating spreadsheets. This week, AI investment banking company Rogo announced it is buying Subset, a company that does exactly that. I want to bring on Rogo CEO Gabe Stengel to talk more about this deal. Gabriel, it's great to have you. Welcome to TITV. Thanks for having me on the show. So why did you buy an AI spreadsheet-making software company? look at rogo we're focused on building ai agents for financial firms for investment banks for private equity for public markets investors every ai agent for a financial firm will need access to the same tools that real human employees have access to powerpoint excel and so on and the team at subset had spent years building an ai agent for excel for the spreadsheet a way to create financial models to do diligence in the interface that financial professionals already use.

16:01And so when we saw that team, their composition, the fact that they were bilingual, they spoke the financial services language, they spoke the AI language, we just thought it would be a perfect addition to our company, our team, and the product we're offering. So I want to take a step back. Rogo sells software that my understanding is you've got analysts and investment banks that could use this to basically replicate their task, I mean, even make it faster. I want to get to that in a second, but how much do you charge for this product? pricing is variable it depends on the type of firm we're speaking with and it also depends on the number of users we have some firms that have 10 seats and some firms that have 5 000 seats you know it's it scales a little bit and we're moving more towards a usage-based pricing model too like you said analysts can use this to speed up their work they can also use it to replace some of their existing workflows i can imagine a world where by the end of next year we charge for you know the number of excel models we create or the number of powerpoint slides we create rather than just a typical per seat SAS model.

16:58But I mean, I'm just thinking of compare, I mean, you know, Bloomberg terminals, for example, I don't know what it is. I don't have one, but I think it's something like$25 ,000. I think something like that. So are we talking like thousands of dollars or hundreds of dollars ahead? We're talking thousands, but it's far more similar to a chat. $25 ,000? Not$25 ,000. No, no, no. I was going to say it's far more similar to kind of the generative AI tools that you're seeing in market, whether that's a ChatGPT or a Claude or a Harvey, rather than a data provider like a Bloomberg or a Faxit or a CapIQ.

17:30And that's because for us, we are a workflow tool. We're not a content or a data provider. Okay. So, I mean, you're putting together these spreadsheets now with this new tool. You know, you can pull a lot of financial data, make the PowerPoints. Why can't OpenAI just do something like this? I mean, I think we heard that they were working on spreadsheet tools. Spreadsheets are a little bit different than just Excel, right? There's a reason that bankers don't use Google Sheets, I'm forgetting the name of their tool, rather than using Excel because it's actually not functional all the ways you need when you're doing a really deep financial model.

18:02Doing financial work is messy. Getting the data in is messy. There's a nuance. There's a type of analysis that the foundation models are just not going to focus on. You have to go very deep. It's not that different than the specificity that a company like Cursor or Dev and Cognition needs to take to solve software engineering. You really need to understand the financial services domain and build out both the custom model and intelligence layer, as well as the real tooling to do it. You know, finance is not just an IQ game, right? It's not that, you know, the smartest person isn't necessarily the best investment banker.

18:31There's a lot more that goes into actually doing that work. Right. One of the questions I had for you is you've got a lot of big logos using your product right now, many of which are large investment banks. You know, something that we've been tracking at the information are the extent to which big companies like this, They're very willing to try these products out with what they call exploratory budgets. And then there's sort of a risk that they sort of churn and fall off after. Are you seeing any of that? What are you doing to prevent that? Look, we always start with proof of values or trials.

19:01We want people to use the product, see if it's helpful for where they are. And we're very transparent about what it's capable of and what it's not. And then we want multi-year commitments to make sure that folks are bought in and actually enabling the platform across the firm. And what we're seeing is there's many firms where we're deployed, where Rogo is a part of new analyst onboarding, right? They have a day where they learn how to use PowerPoint, a day where they use Excel, and a day where Rogo is part of the curriculum. Did you have an example customer that's doing that you can talk about at all?

19:28Uh, some of these banks, it's hard to know what I can say and what I can't say, but there's a number of large banks on our website, like that have thousands of users on the platform and use it daily. And so we'll be able to talk more and more about some of these use cases and about how it's augmenting investment banking work in the months to come. But, you know, Rogo is here to stay at the institutions where to put. You know, one of the questions, and so I should clarify here. So these are these are multi-year deals that you've signed with with these banks. Yes. Got it. You know, I always like to ask founders about the risks to their business.

20:02And the question I have for you is, if Rogo doesn't work yet, why might that be? You know, I think there's a lot of execution risk. It's hard to, you know, build a product that's going to augment and automate more and more of a real knowledge work category. And part of that execution risk is integrating with all these existing systems. I think it's easy to look at a tool like Rogo or ChachBT or any Gen.AI tool and say, well, It's a chat bot. You know, how much can it do? The reality of all these agential tools is they're going to become more and more action based. A tool like Rogo is going to be able to take meeting notes, update your CRM, update the data log, update the model, and then send the email out to the sponsor saying this is where we are in the transaction.

20:41It's going to take action in the real world. All of those integrations, very messy. That's hard to do. You have to work with third party institutions, third party data providers, do the permissioning and the sort of content context passing between models. Right. It's a technology problem, but it's also such a deep and economically valuable problem that it makes sense for us to invest in. Because I just, I mean, you know, you integrate with all these companies, S &P, Capital IQ, Faxit, etc. I mean, I would have thought they could just sort of make one of these tools themselves, no? Well, look, I mean, when you're prepping for a segment, you want to have to do one chatbot to look at private companies, one chatbot for your internal data, one chatbot for a spreadsheet, one chatbot for PitchBook.

21:21You probably want one platform that connects to all your data and understands all your workflows. And if you're the CTO of an investment bank, you're not thinking I want to deploy 12 point solutions. You're thinking I want one generative AI platform that can be our future AI employee and understand all the work that we do. Right. And so last question for you, are banks putting up any walls at all in terms of giving you access to the data that they need? Because, I mean, this is all very confidential information. They can't just throw it around. We've written about corporate data wars. Are you facing challenges there with walls?

21:53Well, look, banks know that to get the most value out of generative AI, they need to start using some of their own data. Every bank is, you know, in a different stage of the life cycle of becoming comfortable with cloud or becoming comfortable with generative AI. But we have institutions that put, you know, a lot of their data into Rogo and use it. And I think you'll see over the next few years, if not even faster, almost every institution is going to get comfortable with connecting large language models to their proprietary content because that's where the edge is going to come from. Great. Well, thank you so much for coming on, Gabriel.

22:23It's a fascinating business. And what I was going to tell you, next time you come on, bring on some of your banker friends, and we can talk to them about how they're using the product. Because I'm very excited to sort of hear about how many PowerPoint hours and Excel hours are being saved. Happy to. Happy to. Thanks for having me. Thanks for coming on, Gabriel. It's great to have you. Bye. All right. Well, we had big news today out of Atlassian that it is making an acquisition as well. The company said it is buying the browser company for$610 million in cash. Browsers have become a really big topic in the era of AI, and so I'm really excited to dig into this one.

22:58I want to bring on Sanchin Saxena, Atlassian's head of product, to talk more about why the company did this deal and what his plans are for the business. Sanchin, welcome to TITV. It's great to have you. Thank you for having me. So look, I think people know Atlassian. It's a giant company, okay? People might be less familiar with the browser company. So tell us what the browser company is and what they do. Yeah, the browser company of New York is the pioneering company for the last five years, but they have reimagined what the new browser should look like in the AI-first era. So the first thing that they did, they launched Arc.

23:32Some of you might be familiar with. It's an amazingly well-designed browser to help you get the job done across multiple tabs, across multiple workflows that you have. The latest incarnation is called Dia, which is an AI-first native interface where you can chat with different tabs, where you can actually organize your work really, really well across multiple tabs and get the job done very efficiently. That's very different than some of the native browsers that you and I are familiar with, Akash, which are full of browser tab overload. And it's very hard to carry your context across those tabs.

24:05And so this is something like, it's like a sidebar, it's a chatbot I can talk with to find data that is in my email or my CRM open. and it can pull data across all those tabs and give me an answer. Let's take a specific example. You talked about email. I mean, if you look at your browser, email is probably the place where you spend a lot of time on. Calendar to know where to go. And probably you have some sales app or some customer support app that you're working with. In fact, in a traditional enterprise, about 11 to 12 different SaaS apps are opened by any knowledge worker. So what you can do is you can chat with a site chat in India, and you can connect the dots across those four or five browser tabs and get the answer you need versus switching tabs and figuring out when is my next meeting?

24:51What's the next email that I got? The chat interface allows you to connect all of those tabs and get you the right context that you need to make a decision or take an action in your workflow. And I've always wondered, like with these sidebar chatbot type thing, and they're becoming more popular, you know, perplexity, just put a comment. And I think they have a similar sort of tool. I've always wondered, are they storing the data? Are they pulling the data? How does the data privacy work with stuff like that? Yeah, great question. I think in the vast majority of the cases, it's very important to recognize that these data are customer data, and you want to act on them in the most secure, private way possible.

25:32By default, a lot of these browsers do not have enterprise-grade security. But Dia, for example, separates your work memory and your personal memory so you can actually have control around what you're sharing and how much you're sharing. Okay, so it does share it with the chatbot, but it doesn't store it? Is that the idea? No, so none of these models are using that data to be trained on. They're using data as a context to answer the question. Got it, got it, okay. Well, look, one of the things you mentioned in the press release is that Atlassian is hoping to create a browser that is designed specifically for workers using apps.