AI Profitability Problem, Silicon Valley Longevity Boom, MeritFirst’s Hiring Strategy | Oct 8, 2025

0:13Welcome, everyone, to the Informations TI TV. My name is Akash Pasricha. It is Wednesday, October 8th. We have got a great show planned for you today. We are talking with Coinbase's former chief product officer about his new startup. We've also got the latest on a new longevity-focused conference that is taking over Napa next week. We're also going to do a pulse check on how anxious founders across Silicon Valley right now are feeling right now as talks of a bubble persist. And last but not least, we've got a new funding round about a company that aims to make hiring at tech companies a much more merit-focused process.

0:50It's an exciting episode, but I want to start with a discussion about AI profitability. For a lot of AI companies, revenue has grown quickly, and we talked last week on the show about some of the downsides that can come with that. But the other big factor that many startups have to think about right now is profitability, and that is harder to come by for many of these startups. And so to talk about this, I want to bring on Aaron Holmes, our Microsoft reporter, and Ilya Kurnos, partner and CTO at Venture Fund SignalFire. Welcome to you both. It's great to have you. Great to be here. So Aaron, let's start with you.

1:24You know, I want to talk about profitability or in many cases, lack thereof in the AI sector. I mean, you know, we are seeing this challenge persist at all layers of the stack, you know, cloud companies, infrastructure companies, even the application layer companies. How are you seeing this play out in terms of the startups in your orbit and also the larger companies that you cover, the Microsofts and companies like that? Yeah, I mean, I think we've seen that there's definitely an appetite for, you know, AI applications across the board. But the challenge that a lot of these application sellers are running into is how to price them in a way that, you know, have healthy margins.

2:00And so I think like, you know, we've seen struggles among the AI coding software where companies like Cursor and Replit have had to, you know, completely overhaul their, oh my God. Keep it going. We're good. We're good. It's the beauty of doing a live show, folks. We're all getting used to it, but keep going. So you were talking about the coding companies, Replits, the Cursors. I mean, yeah, their costs are high. They are. And even like Microsoft GitHub Copilot, which was kind of the first AI coding tool or AI product in general to take off broadly, it was operating at a loss on each query as of roughly a year ago.

2:39And since then, we've heard that they've improved their margins. But I think the issue here is just that the cost of tokens have been relatively high. And that also relates to the cost of running these applications on GPUs being high. But the refrain that we're hearing is that costs will come down eventually, and that'll give margins some breathing room. So I think a lot of companies are still just waiting to see that happen. Right, right. Well, Ilya, I want to come to you. I mean, we've heard this talk about costs coming down. I mean, from what we've reported here at The Information, they kind of are not coming down as quick as some people hope.

3:13And in some cases, I mean, the model costs have kind of plateaued. Yeah, well, I think what's happening is the capabilities are increasing. So if you look at kind of an older generation model, it is true that generating those tokens is quite a bit cheaper now. But what's happening is, you know, for a lot of these applications, you want to be on the cutting edge on the frontier models to have competitive performance, because your competition is on those models. So it's a bit of an arms race. And also you're using a lot more tokens with these reasoning models that produce a lot more tokens. And so it's a bit of an arms race right now.

3:45So, you know, in some ways, yes, older models, tokens, token costs are coming down, but you're using more and you're having to be on the bleeding edge, which is expensive. I think the bet for a lot of these companies is eventually kind of things slow down and they're able to sort of take advantage of the hardware improvements and sort of costs will come down for the models that are actually competitive. So that's what we're seeing right now. Ilya, one of the things that I was talking to another venture capitals yesterday, and one of the things they mentioned I wanted to ask you about is they said, you know, companies are so focused on getting products out the door right now and shipping products to stay to stay competitive that you have a fixed number of engineers, for example, that are working on these.

4:30And if you have a choice between shipping product and then trying to make product profitable and more efficient and optimizing maybe some of the back end, the priority right now is on shipping. Is that the case for startups in your portfolio, in your orbit? I mean, are you seeing startups say, hey, we're not even focused on profitability right now. We just want to get stuff out the door. So I would say, you know, there's a difference between profitability and unit economics, gross margins, right? If, you know, it's certainly true that most startups are not profitable, right? You're actively, actively investing in growth.

5:05But the question is, you know, are you selling dollars for 80 cents or maybe 14 cents or whatever? And I do think right now, kind of the moment in time is such that a lot of founders believe that they need to get out there, grab market share, and then optimize later. And I would say for a lot of companies, that makes a lot of sense. You know, if you're an enterprise company selling, you know, 100K plus deals, I don't think the token costs are going to be the thing that makes the difference between having, you know, a great profitable business in the future because the model costs are actually probably a small component of the overall cost, right?

5:44But, you know, if you're more of a consumer company or a prosumer company, like a cursor that's charging 20 bucks a month to a lot of people, and what's funny there is, you know, I can be paying 20 bucks a month and using a lot of cursor, right? Using it very sparingly. And yet we both pay the same amount. I know they have tiers now, et cetera. But I think a lot of it also comes down to pricing, you know, because their underlying costs are per token. And yet they're kind of the price they're charging the consumer is a flat fee. Right, right. What you're seeing is this mismatch. And I think they're also, even at the higher tiers, you probably have rate limits and caps to what you can use because of this nature.

6:21I think at some point, they probably introduced usage-based usage-based pricing. Now they're going for simplicity in pricing because that's what gets market share. And that's what kind of people have become used to, frankly. So what I hear you saying is you're kind of more in the camp of actually the CEO of Replit who came on the show. He was talking about pricing a lot. And he was saying, look, costs, I mean, they might not come down the way people expect. We have to adjust our pricing to reflect the value that we think people are getting out of our services, and then also a way to actually become profitable in the long run, it seems like that's sort of the camp that you're in.

6:57Yeah, I would say it really depends on the application. I think coding is a good example of where every improvement to a frontier model is something developers want to use, right? So if you're a replet, you need to be on the best, smartest model. Otherwise, you're gonna lose developers. In another application, like let's say you're creating marketing copy for websites or something, I don't think that, you know, you need to be on the bleeding edge for those kind of applications. And you're probably more willing to optimize for cost in that equation. So the domain is what I would say. Right. Aaron, I want to come back to you as it relates to quality of revenue.

7:36We had this discussion last week on the show about a lot of deals that application layer companies sign with enterprises. I mean, it's not always the stickiest revenue in some cases. How much of a concern is that among some of the companies that you're talking to? Yeah, I mean, I think like what we've heard from those companies is, you know, we can pass on some of these costs to our customers because they are getting even more savings from using our products, often by like automating away roles. At the same time, yeah, there's a lot of AI startups that, you know, have, you know, reported revenue that might not be as recurring as they would hope.

8:16And so I think like the question is how sticky these apps become. I think another big question that a lot of these companies are facing is, you know, if my software automates away certain roles and companies, the headcount of the customers goes down over time. You know, how do you price that into your model? Which I think is also just kind of an untested question that a lot of a lot of these startups are trying to figure out right now. Right. Great. Well, I want to thank you both for coming on. It is a topic that we seem to talk about more and more on this show, and I anticipate it will get even more important.

8:48Ilya and Aaron, thank you for coming on the show. That is Ilya from SignalFire and Aaron Holmes, our Microsoft reporter and also a reporter on all things enterprise software here at The Information. Thank you. Being the chief product officer for a company as big as Coinbase is the dream job for many people across the valley. But in 2023, our next guest left that role and quickly started a company focused on AI agents. Now, I will say the jokes about a crypto executive pivoting to AI basically write themselves. But I want to bring on Surajit Chatterjee, the founder and CEO of AI agents company, Emma.

9:23Emma, Ema, we're going to get clarification from him on how to pronounce that. He's coming on the show for his first time. Surajit, welcome to the show. It's great to have you. Thank you. Excited to be here. Is it Emma? Is it Ema? just corrective corrective it's actually an acronym enterprise machine assistant ema what was it enterprise machine assistant ema okay now says emma not ema well i you know enterprise sounds like an agent to me right i mean that that seems to me what it's all about uh very quickly talk to me about what ema does and then we'll we'll talk a little bit more about agents at large Yeah.

10:02So we call Emma Emma. So it's Enterprise Christian Assistant, but it's pronounced as Emma. What we do, our goal was to build an universal AI employee. And we think we have built the first universal AI employee. So it just doesn't answer questions in the enterprise. It actually takes actions, dynamically plans, understands enterprise context, reasons over all kinds of structured and unstructured data, connects to all your legacy systems, take end-to-end kind of entire workflow automation. That's what Emma does. We have pre-built AI employees for functions like customer support, HR, sales, but our customers can also take the framework, take the platform and build their own AI employees, which are essentially multi-agent workflows.

10:55Right. So all of this, while maintaining enterprise compliance, security, high standards of deployment flexibility, and so forth. Right. So we've been talking about agents this week. of course, with OpenAI's Dev Day that happened earlier on Monday. And the question really becomes, as these larger AI companies like OpenAI unveil tools that make it easier for people to create their own agents, what happens to startups like yours and other companies that are sort of pitching these agents to help companies streamline their operations? How is what you're building different than what OpenAI is pitching their own customers, saying, hey, you can build it yourself?

11:36Absolutely. So look, OpenAI continues to increase the overall overton window of what's the art of the possible. But from there to actually getting a deployable enterprise software, enterprise automation software that works is a long path. What we have been able to achieve is making enterprise automation really simple for our enterprise users. Subject matter experts, they don't need to have a degree in machine learning, an engineer. They can just come and conversationally build brand new complex workflows, agentic workflows that can connect to their entire set of systems of records, the data sets, the databases, and dynamically plan and take actions.

12:30That's the difference where taking something that's possible to actually making it work at enterprise scale. How do you think about pricing to ensure that the unit economics of your business is going to be profitable in the long run? Absolutely. Look, we price everything based on usage or outcome, not based on seed, which is like popular pricing scheme for SaaS applications. In fact, I think most SaaS applications today or systems of records are facing a disruption because of that, because most of the seats are unused under each line. Everything is based on outcome, based on usage, and based on the value we are creating for the customer.

13:13Everything we do is based on kind of actual improving the ROI. and our customers are largest of enterprises, you know, customers like Hitachi, customers like KPMG, the largest pharmacy benefit management company in the US and so on. And so how have your gross margins fared then with these large contracts for these agents that you're selling them? Very well. Look, we are not... Are we talking like traditional gross margins, you know, 70, 80 % type of neighborhood or, you know, give us a ballpark. here? It's traditional growth margins, and we are able to achieve that because we are not in the business of building large foundational models.

13:59We are building smaller models. We are leveraging all models. In fact, we have developed homegrown technology that we call Emma Fusion, which is a model of models. It's a meta model that sits on top of every model out there, 100 % LLMs dynamically optimizes every enterprise task and how it executes those tasks. And we have proven that it's like 1 20th the cost for kind of typical enterprise benchmarks and with a higher accuracy than most models out there. It basically leverages everything out there. So you've been able to reduce your cost base by using this model of models sort of technology that you've built?

14:42Got it. Well, Surajit, it is a fascinating space. I mean, agents have certainly become the thing that everyone loves to talk about. And I appreciate you talking candidly about profitability because it is something that we are really trying to wrap our hands around on this show. Thank you for coming on. That is Surajit Chatterjee. He is the CEO and founder of EMMA, a new agents-focused AI company. Thank you for that. Well, we spoke yesterday about how longevity and consumer health is the latest craze in tech. And Next week, there is a big conference in Napa on this exact topic that our next guests are starting.

15:17Dr. Jordan Schlain is a longtime concierge doctor and the founder of Private Medical. And John Battelle is a longtime media entrepreneur. The two of them founded Doc, the name of the event. And I want to bring them on to talk more about where they see this longevity movement going. John and Jordan, it's great to have you. Good to be here. Thanks for having me. How are you doing? I'm doing great. Well, I'm excited for this fun conversation. Look, the place I want to start is, I mean, John, you're a media entrepreneur. Jordan, you've been a doctor. I mean, how did you guys even meet in the first place and decide to start this thing?

15:52John? Well, Jordan and I have known each other for almost 25 years. As a matter of fact, I worked with his sister on, of all things, the Webby's, which you might remember back in the first dot com. Five word acceptance speeches, right? Yes. Right. That was Tiffany, Jordan's sister. but Jordan and I have been friends a long time. He was my doctor for 10 or 15 years till I moved to the East Coast and I hooked up with one of his doctors out in New York. But he and I have always traded notes and done a bunch of whiteboard sessions on problems. I was at starting companies and problems he was having with his business, scaling his private medical business.

16:34And that ended up about two years ago with us realizing we wanted to do something together. And that was There you go. And it started with me as I was at an event in Las Vegas with a longtime entrepreneur, Kevin Ryan. And we were lamenting how these healthcare conferences were just so commercial and so lacking in evidence that they weren't really healthcare events. They were more, you know, kind of performative commercial events. And now if you go to an academic event, you know, they're so inscrutable and hard to understand. So people don't go to those. but when you go to these big commercial events, you're not really learning anything about healthcare.

17:12It's like, what can you buy? And so Kevin challenged me and said, hey, Jordan, if you could invent something out of cloth, like what event would you go to? And I said, I don't know, let me think about that. And kind of then I called John, I'm like, can we whiteboard a little bit on this idea of how do we put something together that's both academic, but it's also ready-made for people to understand, focuses on longevity, and really addresses the innovations that are happening right now. There's so much noise in the market. And I do want to talk about that noise, and I want to stick with you because you're the doctor out of the three of us.

17:47So maybe you can shed some light here. There's been so much conversation around longevity in the last two to three years specifically. And we've got the bands, we've got the rings, we've got the documentaries now that are coming out. And the question I have for you is, Is this a result of any particular scientific or medical breakthroughs that we've had in the last three years? Or is it really just the thing that has captured the zeitgeist right now that people are kind of rallying around? Great question. I think there's a lot of factors that kind of all come into it. One is, let's take COVID just for a quick example.

18:22At one point in COVID, we all, I mean, whether we recognized or not, thought maybe we were going to die or somebody we knew was going to die. So we all confronted mortality or the specter of mortality for a minute or maybe longer. So we also appreciated that those who were healthy during that COVID episode were the ones who actually made it out alive and didn't get that sick. So there is this piece of the longevity movement that we are now thinking about living healthier and longer, number one. Number two is there's like billions of dollars of research over very long periods of time that are finally coming to fruition in terms of the biomedical research on whether it's CRISPR or proteomics or metabolomics.

19:07We're starting to understand biology more. And that leads into the third thing with this technology, which is the ability for compute to come into the fore. And so, but that's part of the problem too, is like everybody thinks this is an engineering, like longevity is an engineering problem. And I would say that the technology is like eclipsed the biomedical industry as the sexy hot thing right now, which is as it should. Well, right. Because I mean, everything biotech takes so much time, right? I mean, longevity is I can control this here and now. And so I certainly see why it's caught on. You were going to say something.

19:42I'm coming back to you, John, but what were you going to say? Yeah, I was just going to say that longevity isn't a technology problem. It's a human problem. And what you need is you need technology and scientists and physicians all to come together. This is not like bring a bunch of engineers and solve the longevity problem. And I think that what we're trying to do at DOC is to bring together a community of technology, medicine, science, media, and really try to like, how do we bring science back into something that we all respect? Because if people are anti-science or science skeptics, I just say, how about that cell phone?

20:14You like that cell phone? That's science. Right? I mean, your car? That's science. Right. How can we be anti-science? So, John, the headline of the conference that you guys are putting on, it really surrounds this question. when, if ever, will AI replace doctors? And it's the question that's on the website. And so we should be asking the doctor these questions, but John, I'll come to you. What do you think? I mean, when, if ever, will AI replace doctors? We're opening the conference on Sunday night with that question that is going to be debated both by physicians, and Jordan is the moderator of that debate, and by technologists, including Reid Hoffman And Mike Krieger, who's the co-founder of Instagram and the product officer at Anthropic, probably the whole audience knows both those names well.

21:02It is a way of forcing and framing a conversation about the role of technology in medicine and healthcare. And I don't want to call the debate before it happens, but it's going to be a really interesting conversation. It'll frame the next couple of days where we talk about a lot of the issues of how technology impacts medicine. And also, back to Jordan's point, how we can get a narrative supported that is about a true north of science and evidence as opposed to claims and marketing, which is what we're seeing a hype cycle in medicine that feels very similar to the hype cycles we're all familiar with in technology.

21:48but we're talking about people's lives here. So I think we need to take it more seriously. Right. Jordan, is the goal with Doc to offer longevity-focused therapies? And in the long run, where do you see this business going? The long-term goal is to showcase, like there's so much, like I said, John said, there's so much noise out there. A podcaster, a celebrity healthcare podcaster has more influence over people's health than the scientist that's developing and delivering these solutions so we're trying to highlight what's real what works what's here now and try to create a community that will understand those things and promote uh those because as john mentioned in kind of one of the premises for this event is great marketing on top of shitty science will always beat you know uh shitty marketing on top of great science and so part of what we're trying to do with John's help and his understanding of media is how do we market science in a way, great science?

22:51Because most scientists don't want fame. They just want to tackle a hard problem, and they spend their whole life trying to solve one thing. And if some doctor does that, or some scientist does that, that benefits everybody, but they're drowned out because of peptides, the peptide circus, then we lose. And so we want to try to elevate and make scientists the cool cats again. And so, I mean, John, let's come back to you. You've started so many media companies. What is the business plan here for three to five years for Doc in making these scientists the cool kids in town? Well, there will definitely be more robust media coming from Doc over the next few years, but we have greater ambitions than that.

23:32As someone who started a bunch of media companies, very few get to the point where the information has gotten. It's not easy. and we know that there are other ways to do, you know, to execute the mission of supporting great science. So we'll definitely be doing media about great science, but we're going to be also creating a platform. We'll be making some announcements about that later in the year. So stay tuned. Great. Well, we'll have to have you both on later in the year when you make those announcements. Best of luck with the conference next week. It's an exciting topic and I'm excited to see what the output of those conversations are, not the least of which is that debate on Sunday night.

24:11You both got to come back on and tell us. But, yep, Jordan, you're about to say something. One last thing, which is a spoiler alert, kind of not really, is at our debate, we are going to have Claude and Claudine on the debate stage. Okay. So we are going to have AI participate in this debate. Okay, there you go. Well, send us some clips then. We'll put the clips on the show. I think that's, I'm excited and, you know, we'll have to tune in to the live stream if there is one. I know that tickets are probably pretty expensive. So you're probably not doing that just yet. But thank you again to the both of you for coming on.

24:47That is John and Jordan, the co-founders of DOC, the new conference that is sweeping through NAPA next week. Okay. Well, there is no question that the past few weeks have started to feel like things are getting very frothy in the technology sector. And back in August, we published a great story in the information about how that has actually caused a lot of anxiety for founders and startups as they grapple with more competition than ever. I want to bring on Todd Olson, the founder and CEO of Pendo, a data analytics company he started back in 2014. He has been watching this exact dynamic play out very closely, and I want to bring him on to the show to talk about it.

25:26Todd, it's great to see you again. Well, thanks for having me. It's great to be here, Akash. Well, I should say, the genesis of this segment was you and I met last night at the first Smart Capital Founders Dinner at the New York Stock Exchange, and we sort of hit it off talking about the current moment for founders and startups. And the question I asked you last night was, I said, how's it going? What's the vibe? And that question really came out of the story that we published, which is that people are so anxious right now, just given the stakes and the pace of technological innovation. And I found your answer pretty insightful.

26:00And so I'm going to ask you again, what's the vibe? How are people feeling around you in your world right now? Yeah, I mean, it's intense. I mean, it's a really, really different time, you know? And I think, you know, we talked about the word anxiety or anxious, and there's zero question. It's one of the most anxious times I've sort of felt in the company's history, and certainly in recent history, you know? And if you kind of like look back across, you know, the last few years, you know, you had obviously 2021, which you sort of had this peak moment within, you know, peak tech IPOs. You had significant funding.