Software earnings, GLP-1s, longevity medicine, OpenAI's hiring platform, & Nvidia's loop | Sep 5, 2025

24:45So the basic idea behind New Limit is that we may be able to make medicines that activate a special group of genes called transcription factors. These genes are really what set up those chemical marks in the first place as you're growing and first developing into a young adult. If we can use some of those same genes, we might be able to reset that epigenetic mark, that epigenetic code back to the way it was when you were young, restoring youthful function and age to disease cells in a way that keeps you healthier longer, rather than only treating you once you become very, very sick. And there are a bunch of reasons this is really only possible today, but I think the science is much more real and tangible to make people believe it is at first blush.

25:19Well, I have to tell you, I mean, you know, when I was researching your company, I was reminded actually, I took biology in high school, okay? And I remember memorizing DNA replication, transcription, translation. I mean, those are three big processes. There were 11 steps. Mr. Turner really hammered it into us. So I remember all 11 steps. It took me down memory lane. But the thing that I want to get into with is, okay, so you're building that on a scientific level. Are you building, is it a drug? Is it going to be a treatment? The whole idea here is that we're going to be able to extend healthy lives for people.

25:53So, you know, is this something they're going to take every day? Is it a treatment for people who are sick? What does it look like? Yeah, sure. So it absolutely is a medicine. So we like to keep our business model simple here at New Limit. We're trying to make medicines that keep you healthier longer. That's a very valuable product. Many folks will reimburse us for those. So there's no magic in the middle or questions about how one would eventually monetize something like this. We really see actually to cue off of your earlier discussion, the GLP-1s is one of the first examples of a medicine like this that might actually be a medicine for everyone that all of us might want to take at some point in our lives to add additional healthy years.

26:26And if you look at the economics of that medicine, it's really the first medicine that supported about a trillion dollars in market cap. So there are these medicines that when they benefit all of us, they treat the pathologies that will affect each one of us here and all of our families, can really be some of the most tremendous businesses ever created. So our dream would be to create medicines like this. And I think all you really need to believe to think that this could be a really compelling set of products is that there are other medicines like GLP-1s out there to be discovered, but perhaps these are not the only medicines of this kind.

26:54In terms of what the medicine will actually be like when you take it, we're using some technology that would likely be administered as an IV, and ideally these are relatively infrequent doses. So one of the beautiful things about the epigenome is that once you've modified it, it can actually be quite durable. The existence proof of this is that your cells today have different cell types because of the epigenome. So your eyeball doesn't spontaneously turn into your tongue or your kidney, and it's not hard then if you take that premise to believe, okay, if I'm able to rewrite these marks, they could last quite a long time.

27:22And so we're thinking initial versions of the product might be dosed every few months, ideally in the limit case. An IV drip. Yeah, through an IV drip, basically. And then in the ideal case, you can take that out to maybe once every few years. That's far out for now, and we don't have that data today, but I think it's entirely plausible given the basis of the science. And we've already seen... Sorry, I was going to say, Who would be a candidate for this medication? Yeah, so the first medicine we're developing is to restore youthful function in hepatocytes. That's just Greek for liver cell. It's almost funny.

27:52We even created a separate word for it. And so the cells of your liver actually help determine your metabolism, how you feel after you eat a cheeseburger or have a drink of alcohol, and also determine your energy levels day to day when you wake up. I don't know about you, but I wish I had the same energy level I did when I was in college. So the hope would be that we can restore youthful function in those cells. And initially, we'd go into patients who have a disease that looks like an exacerbated form of the aging that happens in all of us. And these patients really need a medicine like this.

28:19Their liver disease is really severe. Ideally, they would benefit very early. And then later on, this could broaden out to about half of all adults have something called metabolic syndrome after they turn 60. This is really a concomitant increase in the rates of heart attacks, diabetes, obesity, strokes. They all happen at the same time. And many of us have seen that in our family members or in our own lives, even if we didn't realize that there was a broader pathology underlying it. And do you have an estimate on how many years you could add or how many healthy years you could add to these people's lives?

28:48Yeah. So we get into the realm here of best guesses, and I don't want to make overly audacious claims. What I would say is that if we look backward in time, the first few longevity medicines that exist might be things like statins or, again, GLP-1s. If you look at the healthy economics behind them, the best guess is that those actually add a year or two years of healthy life to the median American. So we think if we're able to make a medicine of comparable impact, it's not totally unreasonable to believe that even just the first product we create could eventually lead to a couple extra years of healthy life for an individual.

29:18Right. And when can we expect this? I mean, it's a slow process. You know, everything in health tech and bio tech takes time. But when can we expect the first product to be on the market? Yeah, to be on the market is a longer question. We're aiming to be in the clinic by 2028. And then there are a number of contingent steps beyond that to determine exactly when that becomes commercially available and for who at what rate. But I would say that within a decade or so, I imagine there will be reprogramming medicines available to folks who might hear this, both from us and from others. I really believe this class of medicines is going to be as big as some of the biggest ones that are out there today, like antibodies, which are the plurality of all the bestselling drugs.

29:57There are actually more drugs today that do a billion in revenue than there are software companies, just to give you a sense of how big these classes of businesses can be. And I think epigenetic reprogramming is sort of the next iteration, the next platform technology that will be built upon. Right. Great. Well, Jacob, thank you so much for coming on the show. And next time you come on, and we'll have you on soon, why don't you come to us live from the lab where you can show us around the lab and some of the experiments you're running? Because I've seen videos of the blood bags, the frozen blood bags that are around the lab and stuff.

30:26It's very fascinating stuff. And so maybe next time you can give us a look at that. That is Jacob Kimmel, the co-founder and president at New Limit. Okay. Well, our next guest is a friend of the show. Sam Lesson is a general partner at Slow Ventures. The company counts Robinhood, Airtable, and Solana among its investments. He is also the husband of Jessica Lesson, our founder and CEO. Sam, welcome back to the show. It's always great to have you. Always fun to hang out. What's going on? So, look, I want to talk to you about some news that OpenAI is now doing a rival hiring platform to LinkedIn, which is like the latest of, you know, the dozens of products that it's working on.

31:05And I just wanted to ask you, I mean, what did you, this is the latest product. What do you think? Well, I think we have to wait to mid-2026 and see what it is. I mean, they're announcing this extremely ahead of anything real. And I think the interesting note is if you look at Fiji's blog post, you know, who's now running apps, a great person who put it out. I thought it was really funny. The tags on the post are global affairs community 2025. It's not product. It's not like the next important product. It's global affairs. Right. And like the reality is you have a company that is kind of the brand of AI disruption.

31:38They've been yelling about how for years about how AI is going to take away all the jobs in AGI. They're now in a political landscape. They got to launch some stuff. like a job center, right? To like make that. To save the, to help people find the jobs. Yeah, like look, this is not, Facebook I believe did a job space when Fiji was there. I don't think she worked on it. I don't remember, but like back in the day too. It's like, so I think the first thing I'd say is like, let's cool our horses a little bit here and see what they're actually doing and why. I don't like, that's part of it. There is a second part, which is like, look, what should we expect from OpenAI in the coming years in terms of apps and platform?

32:12You know, they now have a CEO of apps. most hyper growth companies end up in this cycle for a whole bunch of reasons where they launch a ton of stuff right like products and you know there's internal reasons for that you recruit a bunch of great people you gotta give them spaces to operate in like otherwise they're all falling all over each other on one chat interface um you know there's you know there's just historical reasons you know google went through this obviously with their apps platform and things like that you know 20 years later it's a business for them right um but for many years you know these were just kind of free products because they had a lot of engineering cycles and they were hoarding a lot of talent.

32:46You know, Facebook went through this and most of that stuff didn't end up working. Some of it did. So I just, I think we'll just see this period, especially as the technology matures, you know, you get the chat GPT five and six and there's kind of a flattening out probably what LLMs can do where you see a ton of products and a ton of experimentation and most of it will get chucked to the garbage bin and some of it might not. I was going to say, I mean, look, you know, We've got this table that I'll attach in the show notes of all the different things that the information is written about OpenAis working on.

33:16I mean, you've got social media, agents, browser, hardware, collaboration tools. It's all the things. LinkedIn, right? It's everything. And so my question for you is, is that a good approach? Is it good to sort of pursue all these different avenues? Is this just experimentation? Is it unfocused? What's your take on it? You know, look, I think there's a few things I give you. these are all layered things so like one from a pr perspective right like you know you got to feed the media beast right and like you know this is like the trump playbook right which is you just have lots of things and the reality is that you know most of them will be completely forgotten about and irrelevant some of them might not be even from a media perspective like you just want to stay current so like why not the second to keep in mind is like these are all you know an announcement in putting one pm and an engineer on something is cheap right and so the reality is if you're going to spend a hundred gajillion dollars on computers it's it's like kind of like free options to put them on it and i think the thing in mind which maybe people understand less but if you've been around a hyperscaling company you get is like there's also just a momentum to these things where everyone wants to do a thing if that makes sense right so you have a bunch of human engineers and human product people you're talking about the employees yeah like you've aggregated a ton of talent in like a pocket and like you have infinite resources and there's not infinite opportunities for growth.

34:37So you think about it, it's like there's just a momentum to these things. It might be hyper inefficient. It might be a huge waste of talent. But to some degree, it also just doesn't really matter, right? Like it's fine. Like let them all go do a billion things and overpay them for it. It's not that big a deal because the big bet is such a big bet that it's kind of all noise, right? And the other thing I kind of think about is, okay, so you have 100 engineers, let's say, that you don't know what to do with them. You give them old products, say, all right, split up into teams, work on some, you know, we'll advertise some.

35:07The thing that I think about then is how many of those products are, you know, start with a customer need and how many actually start with, we just want to build something cool. And I guess the answer to that is, well, you know. Well, this is the big thing. I mean, a lot of PMs and engineers who worked at big companies, right, leave and they're stunningly unsuccessful as entrepreneurs, right? Because the pattern of intrapreneurship or what does it mean to be a PM at a place where you have infinite traffic and can throw a bunch of out, you know, like, and have all these resources and whatever, versus doing it independently is massively different.

35:37The other thing people forget is like, there are things that actually are stunningly unsuccessful at big companies, but the companies are just so big that they like end up with millions of random users anyway, right? So like, look, I want to be clear, like, I actually think the education thing, the idea that like you want education and some sort of credentialing around AI, right, which is what OpenAI is pitching on the policy. It makes all the sense in the world. That should happen. You know, the idea that there's actually hiring opportunities around AI, like I obviously completely agree with.

36:07I started a company 18 months ago, two years ago, that does effectively use AI in smart ways to help people hire. I think I believe in those opportunities. History would say that generally the big platforms don't win those types of things, right? Long term, you know, and like they're not like the right place for it. It's not wrong them to do it. I just, I think there's like a very different style of entrepreneurship. That's completely separate from the broader point you're making though, which is like, yeah, you just, if you're a big company and you put a bunch of great designers, PMs and engineers in a room and like they're falling all over themselves because there's one interface and you hire a head of apps, you're going to build a lot of apps and like not all of them warrant that much attention, but because people want to write about them and talk about them and we're talking about them, they're going to happen.

36:51Right. And they're all mostly going to go away. So tell me more about this company. This company built this around AI credentialing? Well, we've done a little bit of that. The AI credentialing is one thing. Like there is like the Google Clouds certification program. And there is like the Microsoft and like people like that stuff. Super agent certified. Yeah, I got it. That's just like enterprise credentialing. You want it. You can sell it. There's like no reason to not do it. You'll have first parties to do it and third parties to do it. I think the more interesting thing is like you have Cluey and all these companies that go around being like, cheat on all your interviews, right?

37:25And like, you're like, well, that's a real problem, right? Because like, now I need to figure out how to unsheet on all the interviews. It's happening in schools too, and all sorts of things. So that's like one part of it. The second, even bigger part is I do think that we're in this moment, I'm not one of these people who thinks college is worthless, but I do think credentialing is on the decline and just show me what you can do is on the rise. And that means talent can come from anywhere. Anyone can learn anything from anything. And I really just care how good you are at a thing. That does change hiring.

37:52right? Like, and how you think about it. Because really, instead of saying, hey, I need a Stanford grad with this degree because it proves X, Y, and Z, and like, I'm using a third party. You're like, look, I don't care where you come from. If you're the best, prove to me you're the best. And like that, that is actually a meaningful, AI has lots of leverage in that. How do you design the tests, right? Like, because you have to be able to test people, used to just test engineers, you should be able to test a lot more people now. How do you evaluate at scale? Like, you know, if I get 100 ,000 applicants to something, how the heck am I going to review them?

38:23Like AI has a lot of leverage on that in like non-multiple choice questions, right? So what is the company that you're working on? Standard Aptitude is the company we started that basically solves these three problems. It says, if you have a job you want to hire for, it doesn't matter what it is. You talk to us about what the job is in LLM style, you know, tell us the story. We help you actually build great tests to then execute against that. Like actually, what is the test? You don't, you're the hiring manager. You have no idea exactly what questions to ask. We help you do that. We help you evaluate the responses so you know who to spend time with and attention on.

38:54And then we do a lot around anti-cheating, right, which is very much like almost stand and deliver style interviewing as opposed to like a multiple choice or a screen based thing, which is obviously you can cheat on in 2025. So it's basically like standard aptitude is this kind of full startup expression of it, which is in the future, you will not care where people come from. You will care to talk to and hire the best of the best, regardless of how they learned it and got good. But there's a whole different set of infrastructure you need to do that well. Right. That's going to happen. And I think standard aptitude is a great shot at winning it.

39:24But there's going to be like a bunch of energy around that. I just have, you know, we'll see where it comes from. And so this company you've co-founded, it's a portfolio company. Yeah, so I initially kicked it off and built VZero for my own personal use. As I've learned you do with many things. I love doing this. And then, you know, at this point, Slow Ventures, as well as Joe Longsdale's 8VC, have financed it. We have a CEO, we have customers. It's like off to the races. And like there are other companies doing this, but I think we're pretty excited about that specific one in the broader context of hiring.

39:56Okay, and so if this venture doesn't work, why might that be? Oh, so many reasons. As a startup, they usually don't work, right? But I think like one version is thematically correct, but there's going to be a lot of different people working on different versions of this, right, is what I would say is like the most obvious. Including OpenAI, maybe. I don't, I mean, we'll see. I think like, again, the idea, ideas are not proprietary, right, in the end of the day. And like, this is like an obvious area to be working in broadly if you know what's going and you believe in kind of the future of how these things are going to play out.

40:28Again, I'm never, I am never worried about startups competing with big platforms, right? Like I actually think that like, this is like a massive misnomer is like people like it's very hard to compete with big platforms. I actually think it's incredibly easy to compete with big platforms. So long as you're not doing exactly the core thing they do, right? Don't compete with big companies on their core business because they're very good at it. But do you want to compete with teams that are shaved off, that are kind of finding their independent verticals in a huge ecosystem and competing for resources and attention all day long, right, is the upshot.

41:00Right. Great. Well, Sam, we got to run, but thank you so much for coming on and sharing that. I think it makes a lot of sense. I mean, look, when I heard about this platform, the OpenAI hiring platform, immediately I went to, well, the resumes are going to be fake resumes and the job posters are going to be fake job posting. So what do you do? And I think your point to making people take tests, I think it's a good one. And so I'm excited to see. It'll be fun. Yeah. Go do your thing. Talk to you. All right. See you soon. Bye. Okay. All right. Well, for our final segment, if you thought NVIDIA's business was complicated, we are about to tell you how it is even more nuanced than you might have thought.

41:38A new story we published this week in The Information shows the circuitous nature of NVIDIA's business, and I want to bring on Anissa Gardizi to explain the details of what she found out. Anissa, two days in a row, welcome back to the show. Oh, you're just on mute. Just take care. There you go. All right, let's get into it. All right. You know, it's funny. This happened last Friday, too. Somebody was on mute last Friday. I think everyone's one person has got to be on mute on Fridays. Friday brain, but no less. We are the knowledge workers here to serve the audience at the information. Okay. I want to get into the story.

42:15Can you explain to us what Lambda is? Yes. Yes, Lambda is a GPU cloud provider. So that essentially means that Lambda buys GPUs from NVIDIA and then rents them out to developers, similar to how an Amazon, Google, or Microsoft might do that. And we call them a GPU cloud provider because they're specifically focused on AI compute as opposed to more traditional forms of computing. Okay, so they're renting out cloud compute. Now, what do they have to do with NVIDIA? Well, NVIDIA is their most important supplier. Other than the fact that NVIDIA does everything, but just walk us through the story that you wrote.

42:54Yeah, sure. So NVIDIA is, one, Lambda's most important supplier. Lambda could not rent out GPUs unless it was able to buy GPUs from NVIDIA. NVIDIA is also a Lambda investor. So, you know, has an ownership stake in the company. And what we reported this week was that Nvidia is also a Lambda customer. So earlier this summer, Lambda signed its largest deal to date, $1.5 billion across a few deals. And the customer is Nvidia. So, you know, three reasons right there why the companies are close. Okay. So Nvidia is an investor, they are a supplier, and they are a customer. Why is Nvidia doing all this with this one company?

43:40What's so special about this one Yeah, so we've seen this a few times before and written about this in the past. But over the past couple of years, as more companies have needed NVIDIA GPUs, NVIDIA has sort of looked around the room at the companies that offer GPUs. And originally, those big companies were Amazon, Google, Microsoft. And all those companies are working on their own chips. And the core of their business also is not to sell more NVIDIA GPUs. selling GPUs is one part of what they do. And so these startups popped up. A lot of them shifted from crypto, you know, renting out compute to miners.

44:19But these companies were like, oh, maybe if we rent GPUs, we'll be able to generate more revenue because AI is the current thing. And NVIDIA saw those companies and thought, you know, we should really partner with these companies because the core of their business will be getting our hardware into the market and they're not competing with us. And so - It seems like a marketing channel, really, for NVIDIA. I mean, it is. I mean, these are the companies that NVIDIA will mention on earnings calls. They get a spotlight at their annual GTC conference. Sometimes they're first to market with a new product.

44:52And these companies help NVIDIA in other ways, too. One thing we reported was that Lambda went to some Google Cloud customers who were using Google's chip, which we talked about yesterday, the TPU. and Lambda employees took on the work of helping that company shift to NVIDIA. So you can kind of imagine Google's not going to do that work. It's literally a sales team. It's literally a sales team. An extension of NVIDIA, yeah. Why doesn't NVIDIA just buy the company? I don't understand. Yeah, this is something that was on my mind a couple of years ago when we saw how NVIDIA helped CoreWeave, which is now a public company.

45:29I don't think NVIDIA would actually want to buy one of these companies. I think they'd rather have a hundred of these companies out there doing this kind of work. Yeah, doing their work for them. Okay, so that makes sense. One question I had for you is, and you said CoreWeb was another type of company that NVIDIA sort of had this type of relationship with. Is NVIDIA the only chip company that does stuff like this, or is this kind of common for big chip companies? I think it sort of started with NVIDIA, but now we're starting to see it happen at other chip companies. AMB is a good example. There are some similar companies in that ecosystem where AMD will invest in a new cloud provider, and then all of a sudden that cloud provider offers AMD chips.

46:09I think this is sort of similar to what we also talked about yesterday with Google, helping another cloud provider, FluidStack, finance a deal where it's going to have TPUs. But I think we sort of started to see this stuff with NVIDIA in 2023, and now other companies are just following NVIDIA's playbook. Right. And last question for you. Is there a risk at all in NVIDIA being so, I mean, it's a huge company. Lambda is a smaller company. You know, I don't mean to suggest that if, you know, Lambda's business starts declining or decelerating and all that, it would impact NVIDIA. But, you know, I am sort of wondering about, you know, if NVIDIA has its hand in so many of these different pockets, is there kind of a risk here at all to its business?

46:53Well, that's a good question. I think maybe the biggest risk here for NVIDIA is that, you know, these companies, usually when large enterprise firms go to get compute, they go to one of three, maybe four companies, Amazon, Google, Microsoft, Oracle, who have years of experience operating this kind of infrastructure. And so NVIDIA definitely is putting a lot of the eggs in the baskets of companies that have been around for far less time and worked with far less significant customers. But I think NVIDIA might feel fine doing that because they might say, well, these companies are primarily focused on our hardware.

47:31They're going to execute the best, but we'll just have to see. Great. Well, Anissa, thank you for coming on and explaining it to us. I'm sure there are many nuances that we aren't thinking about as well outside of this. And so we will have you on to discuss that even more. That is Anissa Gardese who covers the cloud here at The Information. Well, that does it for today's show. A reminder that we are live on this stream Monday through Friday at 10 a.m. Pacific, 1 p.m. Eastern. I want to thank Amazon Web Services, who is our presenting sponsor for this production. And I want to thank you for tuning in.

48:02We really do appreciate your viewership. I'm already excited to see you all on this stream on Monday. And so until then, have a great weekend. Bye-bye for now.