DeepSeek’s V4 Coding Leap, Strava’s IPO Move & Silicon Valley’s Aging Craze | Jan 9, 2026

0:13Welcome, everyone, to the information's TI-TV. My name is Akash Pasricha. It is Friday, January 9th. We have a busy show to close out the week. First up, the information published exclusive reporting that DeepSeek is just weeks away from releasing its new flagship AI model. We will discuss the news with our Asia Bureau Chief. We're also exclusively reporting that Strava has filed confidentially for an IPO. and NVIDIA-backed cloud provider Lambda is in talks to raise$350 million ahead of its IPO. We will break down both of those stories. We have a great lineup of guests for you. Dan Shipper, CEO at Every, will join the program to talk AI and media.

0:55And we are also sitting down with the Zach Raytano, CEO of ROW, who will talk about the company's new GLP-1 pill launched with Novo Nordisk. And it is Friday, so it's time for another edition of The Editor's Cut, and our weekend reporter will wrap the show talking about Silicon Valley's latest obsession, learning your biological age. It is a jam-packed show, and so let's get right on into things. The information published exclusive reporting that DeepSeek is expected to launch its next-gen AI model in the coming weeks. Joining me now with the details is our Asia Bureau Chief, Jingyang Jing. Welcome back to the show.

1:32It's great to have you here. Thank you, Akech. I'll be back in the new year. So what do we know about this new model and the DeepSeq news? Yeah, so the new model is going to be called V4, which is a successor to V3, which was launched in December 2024. And this new model features very strong coding capabilities. DeepSeq employees conducted testings based on their own internal benchmark, and those testings show that V4 outperforms both, you know, anthropic and open-ass models in coding. Obviously, the proof is going to be in the pudding. We won't know if it really outperforms until, you know, we see what it is.

2:20Well, I mean, the first step is getting it out there, so that's the first step here. Talk to me about what is the difference between the V series models and then the R1 model that I know has gotten a lot of attention at DeepSeek. Yeah. So, V3 was launched in December 24, as I mentioned, and that model was already very good. It's a large language model, general purpose model, what we call, that already put the company on the map in the global AI community. And the R1, the R series, which we only had one generation so far is what we call a reasoning model. Basically, it can perform like it takes time to think and then perform a series of tasks in a logical sequence.

3:08And that was the model, by the way, which was released in January 2005, about a year ago, that, you know, shocked, that sent shock waves from Silicon Valley to Wall Street. That was what we now call the DeepSeek moment. Okay. I wonder how you think about DeepSeek's model releases in the context of its position in the AI race. Is it a similar sort of arms race as we see here in North America with respect to just trying to get the best model? I mean, is it working on anything broader than the models and applications that people can use? so DeepSeek as you know is like a big champion on for open source and then all of their models have been open source and when R1 was after VC V3 and excuse me after V3 and R1's releases it sort of started in China this intensified this competition and then really like raised the awareness of open source among the Chinese L1 developer community.

4:16And throughout 2025, we see companies from giants like Alibaba and Baidu to startups like Zhipu, Minimax, and Moonshot all rushed, raced to release their own open source model offerings. And however, in the same time, DeepSeek simply sort of has gone quiet. But, you know, if they do release v4, like we write about in the next few weeks around mid-February, that will be their first major model release after they became a global sensation. And then in the U.S., as we know, Lama, developed by Meta, was, you know, a big champion, is a big champion of open source, where I think most of the other model developers in the U.S.

5:04sort of have taken the closed source approach. And then the DeepSeek moment last year also sort of incentivized some discussions over, you know, if we actually should, in the US, if we should actually, you know, do more open source models as well. And right now, we haven't seen that really happening at a larger scale. And then DeepSeek plus, you know, the other Chinese model developers that have taken the open source route following DeepSeek's lead. has in the past year really sort of made China, Chinese open source model really the most used by number of tokens around the world when it comes to just, you know, open source models on their own.

5:52Great. Well, it is a story that we are following closely. Jing, I want to thank you for coming on. That is Jing Yang, our Asia Bureau Chief, here at The Information. Okay. Strava has filed confidentially to go public. That is according to a new exclusive story from The Information. I want to bring on our Deputy Bureau Chief of Finance, Corey Weinberg, to talk about the state of that business. Corey, welcome back to the show. It's great to have you here. Hey, Akash. Good morning. Is that the LA light that I got you for Christmas behind you? Yeah, I mean, it is, Akash, and I loved it. and to bring our listeners and viewers up to speed.

6:31The information does have a Secret Santa celebration internally, and I guess Akash was mine. I was very thoughtful that the LA Dodgers won this year, and I was also a little upset because, of course, they beat my beloved Blue Jays, but I'm a good sport, and so are you. You put it up, so thanks for doing that. I appreciate it. Totally, totally. the good good signpost for la all right so let's talk about strava which uh maybe you're a power user maybe you're not maybe we'll get there i don't know but you reported that the company has filed confidentially to go public this is a business that we have been writing about for a little while at the information what do we know about the state of that business yes i think Strava is going to surprise people when it's S1 comes out in a couple of months.

7:26The business has had one of those unusual stories where it's been around forever. You know, this company was founded, you know, when I was still in high school. So quite a while ago. And, you know, it has picked up growth as of late. So its revenue is between$400 million and$500 million was what we learned. But it's growing really quickly. It's growing over 50 % a year, which would make it the fastest growing or one of the fastest growing consumer apps on the public markets, should it go public. So that should be interesting to at least some investors. I guess it's kind of interesting with consumer apps.

8:11I mean, 50 % growth is staggering for any company. But I guess the risk with consumer is that if it's in, it's in. And, you know, he could have a good year. But then if it falls out of favor, then we don't know what's going to happen. Yeah, that's exactly. I mean, that's how I would think about the bear case as well. I mean, we'll see once we actually see more fully what the numbers in the story look like. But, yeah, it could be a fad. But right now it's in. And, you know, there are run clubs all over the place. You know, I just moved to L.A. Maybe I'll join a run club. but but you know it has like a decent sort of it's in with the fads right now people like running that's helped it grow beyond just cycling which was kind of its original use case so but you're right it could become not trendy very who do you think would be a good public comp for Strava out there I think the one that I've heard is Duolingo just in terms of subscription business for, you know, that is sort of making money off of people's, you know, sort of hobbies or side quests, so to speak.

9:23So that's one that's sort of come up in my reporting so far. And I think Martin, yours, our colleague, shouted that out last night as well. Well, and so one of the other things I've been thinking about is you talked in the story about, you know, how, sorry, getting a stroke. We got some noises here in the office, folks, but we're still going on with things. You had this piece in the story about, you know, the company being profitable. And we also wrote a weekend profile about the company a little while ago. And look, I mean, this is a business that I could have seen getting acquired potentially a couple years ago, just given the product suite that it has.

10:06Is this all just financial discipline that it's been able to acquire other businesses? Or what do you chuck that up to? Yeah, we don't know much about sort of whether there's been explicit M &A conversations. Obviously, there are larger companies that care about fitness and health that you would think would be logical acquirers. Apple, Garmin you've obviously seen private companies like Aura and you wrote about Whoop sort of take off on more of the hardware side so I don't know, I think it's logical to say this company is still pretty small, we called that out in the story, that revenue that top line revenue between$400 million and$500 million, that's not huge for a public company but we'll see we'll see how it trades after it goes public if it doesn't do well, I think it'll definitely get bought Okay, so let's talk about another story that you published this week.

11:02You reported, you had some new reporting about Lambda, the NeoCloud company. Tell us about what you learned there. Oh, NeoCloud. It's getting interesting. About this time last year, we were all gearing up for the CoreWeave, or maybe not all of us, but I was gearing up for the CoreWeave IPO, which happened in March last year. Lambda is a bit, seemingly, its little cousin. It has a very similar business in that they're trying to become some of the next big cloud providers, but their key tentpole is access to NVIDIA chips. NVIDIA is both an investor and a customer, which is a similar dynamic to CoreWeave, and they're trying to go public.

11:46These businesses need a ton of cash. They have debt. And before it goes public, what we wrote yesterday is it's trying to raise at least$350 million in what's known as a pre-IPO convert, a convertible note. And so that we think is interesting and definitely previews an IPO. I got to be honest, Corey, it's kind of funny. You are seemingly the only person I know that gets more excited about the NeoCloud sector than about Strava. I thought Strava was the more interesting business, but I guess NeoClouds are like, I mean, they are very much in this moment. They're harder to unpack or they're more controversial maybe.

12:30Strava is, it is what it is. I don't know. It's a consumer app that people use and it makes money. Neocloud always generate a lot more conversation because people don't know what to make of them. They are more interesting and weirder kind of financial animals. Like I said, they have debt. They all grow really fast. You know, Lambda is growing sales about 80%. But I think the question with them is like, what's their IP? What is this company? You know, sort of doesn't have a lasting business strategy beyond just being able to capture the AI boom. Now, you had this line in the story that I want to read.

13:10You said the terms of the new financing would pressure the company to list publicly the deal, which is structured as convertible notes, would force Lambda to award participating investors millions of dollars of additional equity or cash interest payments if it doesn't go public within a year. I want you to explain that a little bit to us and broadly the point I'm trying to get at is you know these complexities in deal terms I mean is that becoming more common in this moment in AI talk a little bit about that it it predates that a bit I would would take it back about a decade now it really these kind of pre-IPO you know structured rounds became much more popular when there became kind of a bigger rush of capital into the private markets.

14:01There became a deeper pool of investors that were interested in getting a piece of companies before they went public. And an instrument, a financial instrument like this, it's designed to protect investors' potential downside as well as giving them a piece of, you know, a potential upside rocket ship. And And they're meant, you know, they're designed to be raised fairly quickly. And so it removes some of the haggling over something like a valuation. Great. Well, Corey, I want to thank you for coming on. That is Corey Weinberg, our Deputy Bureau Chief of Finance here at The Information. Okay. AI has seeded a whole new generation of entrepreneurs that are finding ways to reinvent old business models and also rethink how we use everyday applications.

14:50Dan Shipper is one of those entrepreneurs. His company, Every, is part media company, part AI app company. I want to bring him on to talk more about the future of his business. Dan, welcome to the show. It's great to have you here. Thanks for having me. It's great to be here. Tell us about everything that Every is, because there's a lot going on over there. There is a lot going on. You can think about it in a very simple way. We think about Every as the only subscription you need to stay at the edge of AI. And we have three main pillars of the business. One is ideas. two is apps, and three is training.

15:24On the ideas side, we have a daily newsletter about AI. We do vibe checks. So when a new model comes out, for example, we'll get our hands on that model before it comes out and we'll do a hands-on review of it the day it comes out. We have a lot of stuff like that in our newsletter. We have a podcast. That's the sort of ideas side. Then we have an apps side of the business where we have four AI apps that we build and run internally. Each is run by a single engineer. One's called Quora. It helps you manage your email with AI. One's called Monologue. It's a smart dictation app. One's called Sparkle.

15:55It's an AI-based file manager. And one's called Spiral. It's an agentic ghostwriter. So that's the app side of the business. It's all tools that we build for ourselves to work better on the edge of AI that we also give to our subscribers. And the last part of the business is training. So for people who want to go deeper than just using our tools or reading our writing. We actually do hands-on trainings. We do what we call camps where we, for subscribers, they do, we have live streams where they can see how we use cloud code, for example, or, you know, next week we're having the cursor team on to come talk about how to use cursor really well for AI coding in an enterprise environment, all that kind of stuff.

16:33So we have camps and we have an enterprise offering where big companies, we have an enterprise offering where big companies bring us in to actually train their people. And we bundle this all together in one subscription. And with the apps, I'm curious, is the goal here to compete with, and I'm thinking of email here because email is one where you've got these massive platforms. Is it literally competing with these AI email clients? I'm thinking of the superhuman and stuff like that, or is it a more niche offering? Um, I like eventually, uh, yes, I think it'll, it'll compete with, you know, when I think about the giants, it's like competing with Google or Gmail, like that kind of thing.

17:14Um, for now it works with Gmail, it works with superhuman. Um, and over time we're adding, adding more and more functionality. I think the, the, the market that we're going after is really high taste AI early adopters who want to use features at the edge. And those are people that, um, a company with a much bigger audience can't serve right now. And it's actually quite a small market, but I think it'll be the biggest market in the world in 10 years. And we're really, really focused on those people who want to use an agent like in their email for everything that they do. And that's where we're starting, which I think is very different from a superhuman or a Gmail where they have to sort of split the difference between AI early adopters and everybody else.

17:53Now, agents is kind of an interesting area right now because I look at my own use of AI and I certainly use it, but I don't know that I use as many agents as maybe I could be or should be. And you sit at this interesting junction of trying to convince people to use these agents and embedding that into your apps. I wonder what reflections you've had from building in this space about how you can convince people to use agents because it's a bit of a behavioral leap. And I don't know how you think about incentivizing that? You know, it's ultimately about trust. It's about having an experience where you use an agent for the first time and you're like, oh my God, I cannot believe that actually worked.

18:42And everybody has moments like that. And I think our job as people who make software and also people who write a lot is to help people get to that faster and make sure it's clear what the value is. It's not interesting to use an agent just to say, hey, I'm using an agent. I think in 2025, right towards the end of 2025 is when programmers really began to be like, oh, wow, agents are completely changing my entire workflow. Agents really started there. You can use cloud code for like 20 or it'll go off and code for 20 or 30 minutes or codecs can go off for hours. And that is really changing programming.

19:19And it has not quite gotten to that level for the rest of knowledge work. But I think 2026 is when we're going to see that happen. And the reason for that is all of these coding agents that are built specifically for coding actually translate very well into general purpose agents. And so they're becoming the foundation for general purpose agents that non-technical people are going to be using in 2026 and 2027 and beyond. So if I'm hearing you correctly, you're saying that the key to trust is having a product that works extraordinarily well. And until now, we haven't really seen the models be good enough.

19:54but in 2026 you see them finally crossing that that mark yeah you can think of um models as like having us having a leash and i think you know when ai when gpd3 for example which is the first like real consumer application of lms when chat gpd came out um when we started using gpd3 it was the leash was very short it's like one prompt one response and the response comes really quick and over time that leash has gotten longer and longer and it's the longest right now in coding applications and the leash is only allowed to get longer when the ai is going to reliably do something valuable with that extra time if it's not going to use something valuable it's not worth the time and i think what we're going to find over the next year or so is that leash for non-coding applications is going to get a lot longer because there's going to be a massive amount of value that you see, allowing it to go off and do something.

20:53And a specific example that I can give you that I experience all the time, which maybe people who are watching this are not yet experiencing, is I use an agentic browser. So I use ChatGPT Atlas. It's from OpenAI. And anytime I have to change a setting on a website, for example, I don't have to deal with that. And I get asked that. I get asked for that all the time because I run a company. We have 20 people. I'm always asked, like, can we invite someone to this? Or can you change this security setting on this app. And my browser just now does that for me. So I never have to look at a settings panel ever anymore.

21:25And that's one of those things where it can be a light bulb moment for you to be like, oh my God, this is life-changing. You also are running a media company, which in the era of AI is kind of interesting. And you do some great writing on your website. What have you learned about running a media company in this era that you might not have expected when you started out on this project that's a great question um yeah i i think of everything as being tightly integrated it's ideas apps and training and you know if you want to teach people how to like really live at the edge of ai you need to do three things you need to spread the word and that's the media part of what we do and the stories that we tell you need to um give them tools to use and and we give them tools that we use ourselves and then you have to get hands-on with them to like teach them uh for people who want to go deeper.

22:13And I think obviously media is a critical part of that, but also there's this really interesting thing happening where there's not a clear bifurcation between a media company and a software company anymore. Because you can now program using English, the line between writer and builder is starting to blur. And so everybody internally at Every is both a writer and a builder. So everyone who's writing an article is also contributing to building these apps because you can just vibe code your own product. Exactly, and vice versa. And so that's a big deal. But also another big interesting part of this is that because software is so cheap to make now, it actually has a lot of the same properties, like building a new app has a lot of the same properties as writing an essay or making a video.

23:02You can think of software as a form of content now. And so that's another reason why the software part of things is not separate from the media part of things, because for us, if a new model comes out, we can go Vibecode a new app to display its capabilities and then launch the app. And that's the equivalent of writing a really cool article about it because we're showing people, hey, we built this thing. And it's not an app that we necessarily are going to support. It's just like a demonstration of what's possible now. And I think that's really interesting. Great. Well, Dan, I want to thank you for coming on.

23:34That is Dan Shipper, co-founder and CEO of Every here on TI TV. Okay. If you live in New York, you might have seen the subway ads of Serena Williams injecting herself with GLP-1s. The company behind that ad is health tech company Rho. And this week it announced it is now offering the new Wigovi pill from Novo Nordisk. I want to bring on Zach Raytano, the co-founder and CEO of Rho to talk about his business. Zach, welcome to the show. It's great to have you here. Thanks for having me. So this is the pill that you can take. We're done with injectables. Are those going away? I don't know if we're done.

24:11I think this is a yes and. This is really exciting. That's right. But I mean, long term, I mean, the injectables, they'll probably go away, right? I think it'll depend on the individual. And the main reason is, so the amazing thing first to talk about the pill is I do think that it is going to usher in a new era of digital health, direct-to-consumer health. It's the first time a drug of this size and scale launches nationwide on-road. So someone can see an ad and go from diagnosis to delivery in 48 hours or less. And the cool part about this pill in particular right now is, one, it's more affordable.

24:46So it's $149 to start. But two, to your point, patients don't have to trade off between efficacy and the preferred form factor for many. So patients on treatment saw about 16 % to 17 % weight loss on the pill. That's going to be transformative. So it's a better form factor for many. It's more affordable and it's available today on Rho. The thing that you mentioned about injectables is you do see both current injectables. There are some where patients lose 20, 22 percent and there are TLP ones in the pipeline. Maybe people have heard of in Silicon Valley, maybe they've heard of Triple G or Reditrutad or Cagrosemma.

25:22There are others that are coming out in 26, 27, 28 and beyond where patients are losing 25, 30 % of their weight. And so you will usually see the pills be slightly less effective over in the fullness of time. So it really depends. What about side effects? Is there any difference in terms of side effects? They're quite comparable. They're quite comparable, the pill and the injectable. Yeah. And generally speaking, I mean, side effects are, I mean, where are we in that story of people seeing any kind of adverse effects here? Yeah, the typical ones that people see are, nausea or gastro challenges and issues.

25:59And in terms of the frequency, over in the clinical trials, it might be 30, 40 % of people. But the side effects, and it depends how wonky you want to get, they usually subside over time. They're usually what are referred to as peak to trough side effects. So it's when people start or they escalate in dose. And there are ways to mitigate those things and we educate patients and care for them along the way to help mitigate those, but they usually people can get comfortable and subside over time. I want to ask you a question about the broader healthcare landscape right now. It's been kind of a newsy week for the space because OpenAI has made this new deeper push into health tech with its new platform.

26:46And one of the interesting things that we've talked about on this show is OpenAI integrating with all these different platforms, not just in healthcare, but other apps and tools as well. But the broader question I have for you is, so OpenAI is going to make sense of all this data with ChatGPT, and maybe you can link up your MyFitnessPal or maybe your Apple Health at some point. And the business question that I have in my mind is, isn't there a risk that these companies integrating with ChatGPT could just end up cannibalizing their own business and the value proposition that they had in the first place?

27:23How do you think about that issue? Well, from our perspective, I think one of the... So first, I think it's a great launch. It's exciting. I think that healthcare, the established healthcare system, I think over the last 20, 25 years has had its chance to use technology and clearly has come up short for a lot of people. And I think this is just a preview of things to come. So it is really exciting. I think for us, the thing that we really see about AI, not only using it internally a bunch, but when you see right now in ChatGPT, oftentimes it'll caveat or other AI LLMs, you'll see it'll caveat and say, this is not meant to replace your provider.

28:00And so what we really think about is, okay, all of that data, how do we make it actionable? And at some point in time, right, we want to make that data actionable. I mean, that might be a prescription, that might be provider guidance, whatever it may be. So for us, we actually just think that the more patients are engaged with their health, the more that they can turn that unstructured into structured data and get feedback in real time, the better off they'll be. I think an interesting analogy here for Silicon Valley that's obsessed with how does AI really spread throughout healthcare, I think driverless cars are actually a really good analogy here, right?

28:35And the reason I say that is you have some pockets where you'll see these very controlled experiments where they'll be fully autonomous, right? So that's like a Waymo driving around Palo Alto or SF. you'll see similar things in healthcare, right? So an insulin pump is sort of the V0. It's deterministic, but it's making automated decisions, right, in real time. You saw in Utah this week, Utah is rolling out a pilot where refill prescriptions can actually be done entirely by AI. So these very controlled experiments and they'll roll them out over time. And then you do have the equivalent of what we've seen more in like Tesla, FSD or lane assist or parking or things like that, where it complements the human driver.