2025 Year-End Wrap-Up

Transcript generated automatically from the episode audio and lightly cleaned up (filler words removed, misheard names corrected where clear). Speaker labels were added afterward and may contain errors.

Rob May: Hi everyone, I'm Rob May. I'm the co-founder and CEO of Neurometric.

Ryan Eppley: And I'm Ryan Eppley, co-founder and CEO of RootAccess.

Rob May: This is the AI in NYC show. Today's show is our end of the year wrap-up. And we have a very special guest on today Jason Hiner who is the editor-in chief and chief content officer at DeepSeek. He'll he'll deepse the deep view. We're going to talk about DeepSeek but he is not at DeepSeek. He is at the Deep View. And he'll explain what that is, but it's an it's an AI related publication and we're going to have him walk us through some of the key AI stories that happened this year. So Jason, welcome to the podcast.

Jason Hiner: Yeah, thanks for having me on. I'm delighted to be here. It's fun. I'm glad you guys are at Rick are doing this.

Ryan Eppley: Yeah, absolutely. And before we go all into the craziness of the AI saga through your eyes. Love to hear how do you even get into becoming the chief content officer, the editor-inchief, you know, you have a lot of chieft, but it really comes down to you've been writing on technology for quite some time now.

Jason Hiner: Yeah, for a couple decades, writing about covering the technology industry at multiple publications. So, leading publications, ZDNet most recently. I'm for that CNET and TechRepublic. So writing a lot about the business of technology is has been most of my career with a few exceptions. But when when I looked at what was happening in the last couple years, one of the things that was really important to me, my thesis of where media is going, media is there's a lot of disruption happening.

So my thesis about where it was going is it's going to be about deeper engagements with more specific audiences and a lot less about SEO. You know, SEO, Google has driven the media industry for a couple decades, right? It's no it's no no secret. And in the last couple years like that's changed a lot and people's habits are changing really even before AI but especially since AI you know AI overviews are making it so that you know zero click search like people go to search they don't actually click on sites like the ones that I've been working on right so business model under attack for sure and so I really wanted to work on that and I also saw what's been happening at AI in ZDNet when we relaunched in 2022 to when I went to ZDNet, we did this this brand relaunch, the biggest one in 30 years. And when we did that, we changed the brand, we changed the team, we changed the mission, and we created the ZDNet for the next decade in 2022. And when we did it, I said, at the time, I said within three years, AI is going to be our number one topic. And at the time, it was our 22nd most popular topic in 2022. And then three months later, ChatGPT launched, right? And so by January 2023, so less than six months later, it was our number one topic.

And so ZDNet was on this journey covering AI. And the more that I did that, the more I sort of got excited about like there's a lot of really interesting things happen. There's a lot of hype, there's a lot of, you know, challenges, but where there's a lot of disruption, confusion, like there's great work to do as journalists, a lot of, you know, fun questions to ask and ways to kind of help a the audience. And so when I got the opportunity, the opportunity came looking for me, not me looking for the opportunity with the deep view. It's best known for its newsletter and it covers AI, you know, exclusively. But the founder really has this vision of not just being an AI publication, but you know, AI may eventually become like electricity or the cloud, you know, the deep views of brand that's just going to help people understand, you know, where technolog is going. It's AI today and the longer term, you know, who knows what it'll AGI or whumo.

Ryan Eppley: Yeah. And what is the special sauce? What do you feel like Deep View does that's different from a lot of the other AI publications out there?

Jason Hiner: Yeah, there's a lot out there. It's true. So, what I love about the Deep View is really playing the long game. It's about quality. It's about analysis. It's not about fluffing up all of the, you know, things that that all of the AI companies are are are doing. It's really about bringing that level of filter and analytical viewpoint, critical thinking about these things. And so, you know, we tell it like it is. We try to sort the wheat from the chaff. And so I love that. That's like my mostly I built my career on that of like we want to be we won't always break the news, right? Breaking the news has some value, but we want to be the place that when the news happens, you know, people say like, I want to hear what what they have to say about it because they always have really smart takes on stuff.

Rob May: Cool. Why don't we get started? Take us a little bit through we put together a timeline here of key AI events that happened in 2025 as we're we're we're wrapping up the year. Why don't you take us through some of the key events starting in the beginning of the year and we can talk about the significance of those, what was most impressive, all that kind of stuff. Yeah. Send us back January of this year. January 2025 forever ago.

Jason Hiner: Yeah. What a what a lifetime ago for sure. Well that was the DeepSeek moment of course right so DeepSeek came out. It was this opensource openweight model that had trained on a fraction of the resources that that these other frontier models had. And it really was the beginning of something that would unfold through the rest of the year, which is that like, you know, maybe the future isn't these gigantic smart models, one model to rule them all. Because it's so expensive and the frontier models, you know, have a certain perspective. And that that thesis has only continued to sort of you know, propel us throughout the year. But DeepSeek itself coming from China trained on lesser hardware, trained on a lot for a lot less money that really put the industry, you know, on its heels a little bit to start the year. What was your all perspective on?

Rob May: What I thought was most interesting about it actually was the fact that so for people that don't know how programming stacks work, right? There's different layers, right? There's like binary code and there's like assembler code and then there's, you know, these languages like C and C++ and you come up to Python and the difference between these layers of code is how abstracted everything is below it, right? How much does something else take care of that? How much do you like when you write C code, you have to understand the hardware and what you're doing?

And DeepSeek, one of the things about it was everybody hears about CUDA and CUDA is the G is the moat for Nvidia. Y which is the software that helps you get your machine learning model onto the GPU. Well, there's a lower level language that CUDA that interacts with CUDA which is kind of like assembly language and it's called PTX. And one of the things I found interesting was that the DeepSeek team worked at the PTX level to optimize a bunch of stuff. And that was pretty cool and it showed you what I it just shows how unoptimized the AI industry is today, right? Everybody's chasing the next big thing. I think everybody's chasing sort of top of mind capabilities. Yeah. And so I think one of the things that you'll see and we'll talk about this later in the show in like 2026 and 27 is there'll be about the years of optimization and all that.

Ryan Eppley: But I mean it's supposed to be interesting to you. You kind of work at this lower level of software program. It's interesting because you took such a like technology lens on this. I think for for me because you know I have a lot of friends that work in either defense or their international relations. So the geopolitics to me was actually more of an exciting kind of moment with the DeepSeek drop. It was I had a lot of friends kind of like, you know, texting each other on Signal like, "Hey, are you are you checking out DeepSeek?" And it was like this weird thing of like, "Yeah, but like I can't let people know I am, right?" Cuz like, you know, America first like AI here in the States.

And I thought a lot of people either took a stance of no, I'm not trying it. You know, that's like unpatriotic to try DeepSeek. And a lot of people saying, "Hey, like look at their technological like prowess and their cleverness. They don't have the same resources. They've been literally barred from having it through, you know, export controls and things like that." But I mean, wow, it's spending a ton less energy to do some great output, you know, and so I think for for me it was more of it just showed us even though everyone thinks that they've solved so much within AI already, it is literally inning one and two yet, right?

So there's still so much innovation to be done on the technological side. For me, I think it was more of a wakeup call on maybe we're not as ahead as a nation as we thought we were. And then obviously there's there's more things that happen through the year as a as a reflection of that moment. But I think government woke up very quickly to wow, DeepSeek is incredibly effective. What what are we going to do about that?

Rob May: Well, they kind of whipsaw, right? They kind of initially said, well, we've got to we got to ban the high-end chips from places like China so they can't work on them. And then they kind of came back around. No, no, no. We've changed our mind in the fall. We want them dependent on us and not developing their own and pushing forward on their own. I you know, I think they're we've talked about this, I think, on the show at some point. This idea that like mobile phones in Africa because they didn't have landlines, they got much faster adoption. And so you always risk if you if you cut somebody off of a layer of technology when there's a discontinuity and there's a new jump to a new layer of technology sometimes they'll jump ahead of you just because of that that they don't have a legacy infrastructure to contend with. So it's been interesting to watch like nobody really knows what to do or where stuff is going. So the whipsawing has been interesting.

Ryan Eppley: So did you write about any of the technological innovation on DeepSeek? Did you do much of the reporting on like the geopolitics behind it? Like what was like kind of the your take as the editor but also just maybe some of the writing room.

Jason Hiner: Sure. So, so yes. So, we wrote about it from both the politics and the but less the politics but more you know the China US race is always a thing you know the technology race part of it less the politics. But we also wrote about it from the technological you know part of it. And one of the things that was really interesting to me is it was I remember talking to a large nonprofit in the US and they very quickly sort of banned all of their they have some geopolitical concerns and they they banned everyone you know in the organization from using DeepSeek because they were concerned about you know there there's the real concern about China you know from a from a cyber security standpoint and that. So that was that was not something I expected. So that was really that was really interesting.

And yeah, I think that that DeepSeek that this point that Rob mentioned of like the fact that they found and had to f find a way to do it more efficiently, really start started a boulder rolling down the hill that I think has continued to, you know, gain steam throughout the year and I expect to be a much bigger story in 26 because, you know, we had this later, so this connected later with a thing with Stanford study no it was MIT the MIT study that 95% of people deploying AI weren't making you know enterprises were were not making profitable deployments with it. That narrative I think is going to carry into the year of like okay we these things we have to make this cost a lot less for it to for it to work for us.

Ryan Eppley: Yeah, that was DeepSeek, right? So that's January started off the year with a bang. You know, lots of lots of opinions flying around both on the technology itself and you know, are you anti-American if you're into deepsek? And a lot of VCs were playing that game of, oh, you back them, are you allowed to back them, is that okay? And you know, I think Founders Fund came out very very aggressively saying like, hey, it's not a good idea to do that. You know, for for each's opinion, but then what happened next? What was like the big next moment within this this trilogy of the year?

Jason Hiner: Yeah, I mean we'd have to say like the O OpenAI's, you know, 01 the reasoning model. I think that was a little bit unexpected too. And I think also the industry was and the people involved you know you know with AI were trying to understand what is it, how is it different, you know when should we use it? You know, I remember you and I talking Rob and you saying like it it was for your company like the founding of the your company like it was a it was like a seminal moment like when you saw it happen it was like oh there's there's really something here to really unpack and do something with.

Rob May: Yeah. Yeah. And it was the it was the 03 version of that that launched in 2025. Yeah. Okay. They had had an early early version of it at the end of last year but it was the first like Yeah. But it was the first reasoning that worked well. Yeah. That people could do stuff. Yeah. And it's interesting, right? Because it's like you know for for like the people listening that don't understand some of these terms there's a whole bunch of terms post-training test time scaling test time compute all these things that they call it you know inference time compute which is just basically like once the model is trained and done what else can you do to the model to make it perform better and it turns out there are a lot of algorithms that that can control how you probe the model how you ask it questions what you do to it I mean this is what my company's built around so I know a lot about the space but it's a growing space and it's the way you know when you hear about chain of thought and beam search and these these technologies these are the ways that you get more out of the model you can get more out of bigger models you can get more out of smaller models

and so I think everybody went wow I agree with you Jason like this is early stuff we're going to see a lot more of this because people are just my friends at the like frontier model labs still believe like pre-training has so much further to go like just more data synthetic data new techniques and everything else but like this the inference time side of stuff is going to is going to be a big story in 2026 and 27.

Jason Hiner: Inference and like we're you know inference for yeah and for the audience like inference is when you're you're asking the model questions you know the interacting with it versus training is like when you're sort of building the model to like before you know send it out and it used to be you know I don't remember the exact stats but there was something like up until last year it was like 60 70% of GPUs were used for training right making these big models but now with the explosion of AI actually put into production you know it's going to become 80% % inference and so optimizing inference to make it cheap and efficient and cost effective and reduce the power and all that is going to become a big big issue

Ryan Eppley: and it's something that now that you mentioned in that kind of statistic it's we talk about you know New York City as like the applied AI capital it's almost such a moment where when people talk about the bubble or not seeing the difference of every resource is going into just training and getting them ready to oh wow people are using these now I thought when OpenAI's you know, 03 kind of dropped. Like people that I've known for years that are not technologists at all were like, "Hey, this is incredibly helpful for me at work now, right?" Cuz I felt like it it was making it more of a layman's onboarding, right? You didn't have to have like this be a codegeneration tool anymore. This was like, "Hey, like I'm going to dump in a file for you of any sort. It could be a financial report. It could be like a 20page kind of, you know, quarterly earnings. Is like it could be almost anything and people are like yeah this gave me a ton of insights that I'd otherwise send to like my associate to go read.

Rob May: Well, this ties back into, you know, one of the things we've talked about off and on in on the show, right, is like, are we in a bubble? And one of the things that I think is happening is like when you think about 03 and the time it can save you, like like you give it a list of something, right? Restaurants, people, venture firms, like whatever, and you ask it a question, go off and do a research on this list, come back with these things, it's pretty good. It might not always be 100%, but it's pretty good. And it's gotten better since the 03 launch in April. And when you think about the time that that can save and how that compiles through the economy, like of course we're not seeing that. I don't know why people expect that we're like there's these innovations and six weeks later they're in the productivity statistics, right? Like how about three years later they're in the productivity statistics say but you know a lot of people don't know how to use it but like we're getting there and these things are really really powerful

Jason Hiner: and then deep research around the same time. So we started seeing deep research you know released too and that is I think what the significance of that partly is for people for for the broad spectrum of people using you know AI it's their first taste of what AI agents will be where you're like like go off and do this thing for me and then I'll come back you know later like because because it can take deep research can take five minutes you know sometimes sometimes I guess I've heard it can take up to 25 minutes I've never given anything that's taken up to 25 minutes you know It'll take two minutes, but you know, it can do things for you.

I had this amazing example of it at one point this year. Somebody was texting my my son asking him about a job and it was this company that he'd never heard of. And so he he so he texts me and he's like, "Have you ever heard it?" You know, because it's a tech company. He's like, "You've ever heard of this company? Like are are they for real?" You know, and I was like in the middle of a string of meetings. I was like, "Look, I'm in a string of meetings right now." But so I sent it to Claude Deep Research. I said, "Let me let me check, you know, let me do a query." sent it to Claude Deep research, ask it like what is it rated? How many job listings do they have? What can you tell? Is the company legit? How are they regarded?

You know, and it came back with an amazing set of things, you know, it was about five minutes or whatever. It was by the time and I had a break in the meeting, went checked, got it, sent it to him and I was like, looks pretty real. You know, they had all the citations like this many job listings. Here's what what on glass door it's listed you know this and it was it was a great example of like okay this is what AI agents are going to be like you're going to be able to just give it a thing go work on it then you come back to it and then you know can send that off you know that workload or that that workflow to somebody that was pretty cool

Ryan Eppley: that's cool so that takes us to July we had a new event that put a new company on the map

Jason Hiner: boy yeah this is where so Grok 4 came out so xAI I Grok 4 came out and that was sort of the watershed moment where Grok 4 went from this like funny little side project to like oh this is this is a true frontier lab.

Rob May: It won on some of the benchmarks when it came out.

Jason Hiner: Yeah, it did. It did. It was beating Google it was beating Google open AAI and Anthropic on some benchmarks and so all of a sudden it was like okay this thing's pretty real and it's good at a specific set of things I think too. So because it's trained on xAI is from X, you know, Elon Musk and his companies. It has all that Twitter data. And that's a that's a lot of data, you know, a lot of real-time data. So I think of Grok as as similar to like Perplexity. If you want something that's more like real time, super timely, I think Grok and Perplexity are the ones that, you know, have the best you know, the best stuff for that. At least the best front-end interface too and Grok maybe the best data. So it it suddenly became pretty real and it's continued to be throughout the year too. Like 4.1 is still is still good. It's still up there on a lot of the benchmarks.

Rob May: Yeah. And they did a lot of reinforcement learning with human feedback I think to make that model better. And so that was that was an interesting direction that sort of proved where RL could go. But I, you know, it's gonna be interesting to see h how they monetize this, right? Because I've read some stories towards the end of the year that no, Grok has this or xAI has this like perception of like, do you really want to advertise there? It's kind of a cesspool certain areas of the, you know, so I don't know.

Ryan Eppley: It wasn't a seminal moment, but I think it was during the summer where it got extremely hot, no pun intended, with the time of the year when it was all about stealing talent from each other. Yeah. Right. This is right when you started seeing posters and flyers about people getting like poached from each other's like, you know, super intelligent labs and things like that. I just vividly remember it was like right around this moment where, you know, Grok 4 comes online. Everyone's beating the drum about we're getting the most talented people doing, you know, in Meta's case is like sometimes $100 million pay packages, which is outrageous, right? Looking like, you know, starting quarterbacks for a franchise. You know, how how did you all report on the talent wars? Because that was obviously a big moment and it continued, but it really felt like it took off this year for sure in the summer.

Jason Hiner: I think that there there became a point where there there really was an arms race for these things. Like you look at those developments in the first half of the year with DeepSeek, OpenAI and Grok. A lot of those were at the sort of model training you know level and it was becoming clear if you were even around the same time but you know it was mostly in the US an OpenAI Anthropic and Google game and Google was considered a trail you know they were trailing at this point right like they were they were sometimes a little bit of a joke like they would release a model and mostly people would laugh at the things that it was doing but they were still considered you know one of the big three and so If you were OpenAI or sorry, if you if you weren't one of those three, if you were X AI or you were the straw or your Meta, you felt like and it looked and felt like you were really behind and Apple too, right? And Apple sort of sort of gave up the gave up the race. It felt like you probably weren't going to get there.

And so Mark Zuckerberg definitely feeling that is like, okay, forget it. We're gonna go and hire all the best, do what we can to hire all the best people, right? And they they made this crazy crazy, you know, move to go out and just offer keep people tons of tons of money. Interestingly, I've heard Anthropic CEO say that there were people on their on their team that got gigantic offers and didn't leave because they they refused to match offers. They really were saying like, "No, we're gonna we're gonna keep you know, we're not gonna let this ruin our our pay scales." And you have to think about the kinds of people that work on cutting edge AI. Like they want to be paid well for what they know. They don't want to like these people spend their weekends working on AI, right? Like trying new stuff. Like they're the difference to them between making a two or three million a year at Google and, you know, 20 or 30 or $50 million. Like it's it doesn't affect their lives very much.

Rob May: I so I have a friend who was on Zuck's call list who works at Google and his whole thing was like yes I got a better offer than what I make at Google like much better but I don't believe that Meta is as serious about AI and where it's going to go as Google is. So I like I want to stay here because I he believes Google will be the place that cracks AGI and he wants to be there when it happens and that's more you know he feels like he's compensated fine and that latter piece is that's more important to him.

Jason Hiner: Yeah. Once your bills are paid like you know it's not gonna have that much of an impact on on their life right like if you're he's well compensated you know a lot of these folks are well compensated the ones who left probably were like okay I want to do this for a little while and then maybe I can go do that startup I wanted you know funed myself with my nine figure paycheck

Ryan Eppley: I've been seeing a lot of even like series B series C startups starting to be aggressively outbound in public talking about their aqua hire strategies obviously aquaire has been around forever, right? It's kind of the easiest way to say you had an exit without it being like a big financial return. You know, it happened, but it wasn't publicized. Now I see founders saying, "Hey, we raised$100, $200 million, and we're going to go buy three to four companies that aren't trying to sell, but we just want their talent." you know, and I think this whole talent wars has, you know, been healthy in a lot of ways, but it's also been, you know, kind of distracting because it's really hard to build long-term success if you're continually switching people out of the locker room. It's hard to build a long-term, you know, winning team. And so, I' I'd been curious personally as I build a team of like, you know, how dedicated are people to the long-term vision of a company? You know, I don't know if if you've spent a lot of time thinking through the challenges of the talent war and the effects this will have and the ripples it will,

Jason Hiner: but yeah, I think the talent wars is, you know, as we get through, you know, Grok 4 into like the late summer, probably a moment that in 2025 will always be a defining moment for this AI race is the talent war itself. When when sort of these huge checks started getting written for AI, you know, engineers and sort of the downstream effects and then marketing it, you know, oh, we're paying this much for this person, you know, it was like a marketing, you know, opportunity more than it was like actually talent.

Ryan Eppley: Patrick Mahomes now wishes he was a computer science major.

Rob May: So, well, I mean, but it kind of ties into our next our next point about GBT5, right? And like they they kind of like I wouldn't say it was a flop, but it was disappointing. And does that have to do with the talent the brain drain that they had? You remember that they lost

Jason Hiner: so GPT5 when OpenAI released GPT-5 the there was not the big leap forward we'd seen with their you know three and four and even like 3.5 and so there was this big question of like what what happened why and what was going on and it was it was very clear in one sense because their two lead you know tech of the tech team had left Mira Murati and Ilya Sutskever and so it was kind of like Oh, do they not do they not have the goods, you know, to deliver? I don't I think that narrative is probably a little over a little over

Rob May: when you can raise two billion on 10.

Ryan Eppley: Yeah, maybe you can go start your own company.

Jason Hiner: Yeah, exactly. I mean, a lot th those two went and raised, you know, billions based on a couple blog posts, right? Which is pretty amazing. So, but they still have a lot of talent at at OpenAI for sure. But yeah, the but it it really did bring up that question of like, okay, maybe did they release it too soon? Do they not have, you know, the talent? Did they not know that it was

Rob May: Well, I think there's there's a lot of pressure on the recruiting side to be at the place that has the best models and is doing the best research, right? And then it's compounded at OpenAI. Like, you know, Anthropic has a different business model. They're more enterprise focused. They're closer to cash flow, break even, and all that. Yeah. So they don't have the burn problems OpenAI has. And then Google has a couple of cash cow businesses. So they don't have that. So OpenAI is really in a unique position.

Ryan Eppley: Yeah. And maybe we can get like a graphic up here at some point. But like when all of the different companies were starting to partner with each other, specifically OpenAI, right? You had the Oracle deal. Now you have the Amazon deal.

Rob May: You know, hasn't worked out well for Oracle though, man. That stock's way down since the announcement.

Ryan Eppley: It's been super super interesting. Just like I have to imagine you were all over this kind of backdoorsy where is the money actually parked and where is it going and what does the deal mean? You know, I've spent a lot of time trying to work through the cobwebs. What do these hundred billion dollar deals actually mean? Is it marketing? Is it actual long-term kind of like shareholder value creation? Like how have you seen these massive deals that just seem kind of like a web right now?

Jason Hiner: Yeah. And so, so I guess that's as we get to sort of the second half of the year, we start seeing OpenAI signing these huge deals, partnership deals. Oracle, obviously the Microsoft, they had to re, you know, up their Microsoft deal. You know, now Amazon recently, but they've they've done others. Nvidia, I mean, invested hundred billion dollars or whatever. Yeah, that was massive, right? So, I think some of this so and circular financing comes up a lot, right? Because it's kind of like I will invest 100 billion in you and you commit to buying you know 90 billion of my so it's sort of like is any money actually you know transferring I think some of it comes to the fact that you know it's tough for them to be a private company they're getting to a scale that it's really hard for them to be a private company there aren't there's not a lot of there's not $90 billion sitting around in a whole lot of banks right to be able to move and so but when you are a publicly traded company, you know, then those kind of numbers, you know, you can move around a lot more like with stock.

Ryan Eppley: The public would love to own a slog of OpenAI. I don't know why they don't IPO sooner.

Jason Hiner: I think there's like a lot of things that they're still sorting sorting out for their S1. I think their S1 is going to be very interesting to

Rob May: their financials are probably terrible.

Jason Hiner: Right. Right. Right. That that was my nice way of saying that. You know, I don't know if they're terrible.

Ryan Eppley: The Robin Hood Gen Z millennial crowd would just like buy that like just it's got meme stock written all over it. Wall Street bets maybe on Reddit. It's just I could also see it going like the antithesis direction. I can see a lot of people going well it would short it and I think that's what you get you get a lot of people short it.

Rob May: Yeah we had on an earlier episode I shorted Oracle for about six weeks based on the OpenAI deal with the and my thesis was that there would be a time when OpenAI announced they were struggling to fund raise and Oracle would be the company most impacted by that.

Ryan Eppley: Yeah so because I still don't understand like Stargate like I still am like still trying to figure out what that private public partnership at the largest level really means because that's right in the mix of this whole like you know interconnected web as well you know break it down for us. What is Stargate?

Jason Hiner: Stargate. So Stargate is this project with the US government you know OpenAI Oracle US government am I remembering that also SoftBank is part of this too. Yeah. So the interesting thing so and they are building a massive massive data center 500 billion $500 billion data center. They don't even call it a data center anymore. They call it an AI factory. Yeah. Which is essentially like seven data centers, you know, right next to each other.

Ryan Eppley: It's an AI, it's a marketing term, right? Like it's AG, it's like AGI. It's a marketing term. If we, if we've all played like Civilization enough, like a new piece of the tech tree has just opened, an AI factory, right? Like we just unlocked AI factory, right?

Jason Hiner: Yes. Yes. Totally. So, they are building a physical thing in Texas that is which seems an odd one like cooling is a big thing. So, like why would you put it in Texas?

Rob May: I put it in Montana.

Jason Hiner: Exactly. The coldest place you can find Alaska or something I don't know but you know you need good internet I guess so the interesting thing is they are building this you know big time and they are doing it at a scale which has not really been seen before right building data centers at scales that haven't been seen before but they started doing that and that was earlier in the year you know they've built bunches of it already and it has a lot more a lot further to go but and then you saw you know Amazon and Microsoft and other in Indiana just north of where I am. There are essentially seven projects from Amazon, Anthropic, Microsoft, OpenAI built all throughout there. It's a little colder so maybe that's why. But it's the these projects and then Atlanta is another place where there's a bunch being built. You know, I have a friend that work is it works in tech and he his whole neighborhood was went in and were sort of doing a petition against this being built because they didn't want their their power costs to rise because that's been happening in places where these are being built. So, the data center boom is also one of the huge stories especially as we've gotten to the second half.

Rob May: Yeah, we had a whole discussion a couple of shows ago about how the jet engine used jet engines are turning into power devices for data centers. I mean, it's crazy what's going on in that space.

Jason Hiner: EV batteries, too. EV battery. The demand for EVs hasn't gotten, you know, hasn't turned into what was expected. And so, you saw Ford announce that they were like the F-150, they they were going to end production. And instead of like stopping production of the batteries, they're going to sell them to data centers.

Rob May: Yeah. But it's interesting to see the like the density of these things as as we've as as Moore's law has caused this to not be able to grow chip density the way we would like. You're starting to see see it at higher levels of abstraction. Right? So now you're talking about more chips per board, more boards per server, more servers per rack, more racks per, you know, row in a data center and they're figuring out ways to do that. Nvidia's had in particular has had a lot of foresight in you know with the Mellanox acquisition and some of these other things sort of acquiring the networking high bandwidth you know capabilities to connect these things and I think that's going to be a big trend I think 2025 sort of laid the ground for that and I think the sort of memory and networking components will be a big thing to watch in 2026 for how these data centers scale you know the effect of like flops of AI.

Jason Hiner: Yeah. I wonder too probably the audience probably, you know, I wonder people in the audience probably wonder, you know, why do we need all these data centers, right? And so I think some of it comes down to they are the these companies like OpenAI, Nvidia, you know, Oracle, Microsoft, Amazon, you know, they look and they see that, you know, only about 13% of the population in the world use generative AI today, right? So when you look at that over the next few years, like there's a the growth curve is going to be pretty steep. And then the people that even use it, a big percentage of those 13% all they've done is just their first few queries, right? Asking a simple question that they would have asked Google, right, a few years ago. And the more people use it, they tend to use it more and more and more. And then when you factor in things like AI agents, you know, deep research, and you start using it for some of those things, all of a sudden that's where, you know, what we talk about inference comes in. And the demand for inference is likely to spike just enormously. And so today's computers can't contain today's data centers, these they could not handle, you know, that kind of demand.

Ryan Eppley: Yeah. Did you guys see the Micron announcement yesterday? So, so, so the day we're shooting the show, the night before, Micron announced like I mean they blew their numbers out of the water selling DRAM to companies on the Yeah. Every every, you know, person that builds computers like me in their home is like, "Oh man, I can't I no longer can afford that. I'm way too expensive. I can't compete with, you know, the likes of enterprises."

Rob May: But it but it shows like like what's interesting is you know you have the Michael Burry's of the world and these guys that are like oh this doesn't make sense and we can't justify it. But when you talk to the people that are in it doing it, they're like the demand is so strong for so far that we can see like we're like they're the Microns of the world, the Microsofts, the people buying data centers, the Googles, they're all in and you know to look at Burry's memo like Google's basically like our TPU v1 is still 100% utilization like seven years, eight, nine years, however old it is, like they're still using it as much as they can go. And I, you know, I think that's you know, it doesn't necessarily mean we're not in a bubble because it's hard to, it's still, there's still a lot of uncertainty around how you monetize these things and what they're worth and how efficient and productive it is. But in terms of demand, the demand is there and not going away. And everybody believes that.

Ryan Eppley: I mean, if we were sitting in 1780s, 1790s going into the turn of the century, like, we'd all invest in like railroads, right? Like that's a technology revolution. And I think we're in the midst of it right right now. So whether whether there is a bumpiness like every revolution over over time like you know I think the investments that when we get to this like fall period and the Stargates really starting to take effect and there's there's a lot of dirt being moved for these AI factories now I think it's very real you know and I think your your you know team's definitely written on you know bubble versus real you know we always ask every guest right like bubble not bubble like what's your take you obviously heard you know Rob's Rob's input what do you Bill, you know,

Jason Hiner: I do still I mean, Nvidia says that it has a 12-month backlog, right? Like like they they if they stopped if they didn't take any more orders today, they would hit their 2026 they already had 2026 taken care of, right? So, I think that does speak to the level of demand and this this growth curve that we're still at the very beginning of.

I the reason I the number one reason I don't think we're in a bubble with a capital B is because everybody's talking about it. You know, bubbles sneak up on you. Bubbles are like in the you know, I was I was covering the dot bubble. You know, I covered the you know, the 2012 2010 to 2012, you know, bubble. At those times, everybody was like, "It's all up and to the right as far as we can see for, you know, forever." And now we so we sort of just talked a little bit about that, but at the time, nobody was talking bubble. They're like, "It's just all it's just all up and to the right."

And one of the things that ju just structurally bubbles are usually you know involve the markets right where the markets are just like that typ typically happens when bulls just take they just run with it and then you know they drive the price of everything up so high but these are private markets these are private market with very smart investors doesn't mean that they always are right but they do a lot of care and research so I don't think there's a capital B bubble because usually one of the things a smart sort of finance person told me is like bubbles happen when the shorts give up like they can't understand the market and so they don't short anything because then the bears take off right with the and drive the price of everything up.

There's a lot of shorts that are shorting all kinds of things right now, right? So when you start seeing like people who short like you know Burry and that that sort of like give up because they can't you know understand what's happening that's trouble. But right now, when there are a lot of people that think we're in a bubble, we probably aren't in a bubble because they they're still they're still in the game. That said, I do think there are small bubbles just like there were, you know, in the past. There are there are in the there's a lot of cash that's been deployed by VCs over the past couple years. Like, not all those bets are going to pay off. The data center thing is really I think there's a lot of money being spent on that and the circular financing of it puts a lot of that circular financing is really bet on sentiment.

You know if OpenAI loses sort of sentiment that it's not as worth or that as much as everybody thought or that it doesn't have this strong up to the right trajectory. You know some of that house of cards will start to will start to fall and we could see some you know some ripples from that. So there's likely to be some backlash, you know, here and there, but I don't think it's a capital B bubble like the dot, you know, crash or the even the 20, you know, 920 yeah, mortgage crash.

Rob May: So let's talk about so our last story of the year from last month, November. Google launched Gemini 3. Kind of took back the crown on the model side pretty convincingly. And not just that, but like the narrative shifted from this this summer the narrative was ChatGPT is going to eat into their search business. Google's toast. Like what are they going to do to now like Wall Street's like Google is the best positioned and most vertically integrated to take advantage of AI and the stocks like on a tear since late summer. Yeah. Yeah. What were your thoughts on that? How did you guys cover it? What do you What do you think about Google going forward?

Jason Hiner: I don't think I've ever seen a narrative like change so drastically so rapidly. Which is which is really really interesting. I think some of the underlying realities are are still the same you know despite the narrative changing which is that you know Google is this big sprawling company that does a lot of things and so the reason that it got behind in the generative AI revolution even though it created the transformer right like it created the technology that underlies LLMs was because it still does a lot of different things, right? So, it's easy to not to lose focus when you have a company that does like this is what we do. This is all we think about every day. This is our our jam. You know, they tend to, you know, outrun big giants. And so they had lost they had lost some of that.

Clearly, they've got some of their mojo back. I think the part of that though is the fact that they took the deep mind team that was in London just working on AGI essentially over there and sort of pointed them toward like hey you essentially built all of this stuff like you know build us a model that's really good and I think my understanding is that's what's really driven them to this this breakthrough and if that if that continues to be the case I think they are they are the one that I mean the reason they bought DeepMind, the reason DeepMind agreed to be bought by Google was simply because of compute. You know, they had the access to more compute than anybody else. That's still true. So that's what I mean about like the reality is still the same that Google still has more access to comput and that's accelerating now. They have their own TPUs. They keep getting more advanced. I mean, they're just

Rob May: Yeah. And so that's what all those deals that that OpenA is out there trying to sign is they're trying to get more compute, right? Google doesn't have to go and sign those deals because they already have it.

Jason Hiner: Yeah. That puts them in a great to your to your point ultimately that puts them in a pretty amazing place to be not a great not a sort of unreachable moat or something but it is a strong position and as as they move forward when you think about where AI is going you know multimodal models world models world models that are based on physics are often trained on video who has the largest cache of video in the world I mean it's YouTube so it's

Rob May: they have a lot of unique assets the world models like they have you know Google maps they spend so much time building that out right you know their TPUs are on V7 they're very efficient now and one of the things that that Jason and I were talking about earlier is actually this idea that you know these these you have all these new AI accelerators that are GPU competitors you know SambaNova you have TPUs from G from Google you have Trainium from from Amazon you have Groq with a Q you there's a bunch of these chips and so they've they've all struggled like we talked about to from Doug O'Laughlin right at SemiAnalysis because of the software stack to get your models onto these chips with Trainium and TPUs going and working with these big companies that is going to ease a little bit right you know people are going to get used to thinking about different accelerators and what they should use them for and that could be a really big story in 2026

Ryan Eppley: yeah my my whole thing about Google Because if you are call it 30 to 40, right, you still have another 20 to 30 years of your like kind of primeish career. Yeah. I would argue that Google is the most trusted brand at least in America for that age group.

Yeah, it's cuz like I'm 32 and like when I think about a company that most of my friends would automatically like click the button to try something or go ahead and like do the new feature, it's probably like, oh yeah, I trust Gmail. Yeah, I've Googled since I was like in my computer lab in elementary school. The brand is don't be evil. Even though that's not really right well marketing marketing is very important but I feel like Google that from just from the brand positioning aspect they are going to get all the people in the middle of their career right now which are decision makers enough they might not be the VP of enterprise companies but they will be in 15 years from now but they are engineering managers they're the managers on the marketing team they're like doing the CFA work and you know they have enough sway and decision-m power but they're young enough that they're will look to as, oh, we're trying out a new technology in our company. You seem like someone that's like young and hip to this. And you're like, dude, I'm like in my 30s, like I'm not like a 19-year-old anymore. But you still have that youngness to you, but you can make decisions. And I feel like Google is just the right time for this decision maker to onboard new tools. They're going to trust Google more than a new tool in a lot of ways.

Rob May: I think AI has been Google's phase two moment. Like so you're old enough that you probably remember when Microsoft was the cool kid and then from like 2007 to like 200 13 14 people were like why would you ever go to work at Microsoft? Like that company's going to go out of business and then Nadella came in and really did a great turnaround there. And then and then it became Google became the like why would you go it was the place to work in the early 2000s then sort of like 2017 18 you're like why would you go to Google? Now I think that's come back. So maybe it's Meta now, right? So now Meta is in this like, you know, wandering in the desert phase of like what what are we doing? I mean, Darkhorse of course is Apple who's like bowed out and people have left, but we all know like their balance sheet is too big to not get involved at some point. What's going to happen with those, you know, Meta with Apple with the ones we haven't heard a lot about? Like where are they on the AI side? Where are they?

Jason Hiner: I know. So Apple, it's really interesting. You know what what we've seen the reporting that's been done is that they have had a lot of brain drain like this year most of the people that were on you know the ones driving a lot of the LLM strategy a lot of their generative AI strategy because they were going to try to do it on their own essentially I think they wanted to become a frontier lab right you look even to the first half of 2025 I think they were still on the track like we're going to be one of the frontier labs and they they appear to have gotten off that train you know that they've had a lot of a lot of the folks that were working on that have left, you know, have gone to I've heard that a lot of that team is now works at OpenAI. That was that was at Apple. That said, I don't I think we are going to get to a point in 25 where and perplexity is a good example of this where you know the frontier model part is going to be a little bit commoditized. It's not going to be super valuable. It will continue to be super valuable, but I don't think you have to be there to be a player, a big player in AI.

I think that you could be more at the you know, at the sort of application level of like integrating this and making it useful to people and do a lot. I mean, Perplexity has done that really well. It's a it's it is sort of the best AI search in a in a lot of ways for a lot of things. And I think Apple can take they can they can partner with OpenAI or Anthropic or Google. You know, the latest of the word is probably going to be Google. They will white box some Gemini, you know, stuff and integrate it. But they can still do some great work and makes use it to make Siri better, use it to integrate in directly into you know, those sort of billions of people that have their devices. They've got they've got the opportunity to do that.

Ryan Eppley: Seems so obvious that Siri would have been great by now, right? Just like shouldn't that be like the original AI agent? Yes. I just, you know, I feel like they they lost out on that big time. But, you know, Apple's they're going to do something impressive is my my take in the next year or two.

Rob May: We'll see. They're they're great at taking markets that are shaping up where people are doing stuff and like sitting back and watching and then coming in and being not the first company to do it, but the first company to do it right.

Jason Hiner: Yeah, that's kind of their strength to make it obvious, to make the thing obvious to the way a person would use it and integrate it into their daily life, their daily workflow. They still have a plenty of opportunity to do that. You know, I think they they need to do it, you know, in the next couple years because we're seeing that 13% is going to turn into 25, you know, probably pretty quickly and they need to be in early enough that they can, you know, be part of more people's lives.

Ryan Eppley: Yeah. So the whole Gemini and Google kind of taking back the crown was a big moment, you know, last month here in November. Is there anything that you are writing on now to end the year that you think is like something that is worth our audience hearing? You know, cuz it's not live, so they'll they'll hear in like a week from now, right? So is there anything that you're you're writing on right now that that's coming out in the next week or two before the year ends that you're like, I think everyone should really take note of this when it drops?

Jason Hiner: Yeah, I the year end is usually stuff goes pretty quiet, but I will say one for 26 that I think everybody and we touched on it a little bit already. I think world models is going to be the AI agents of 2026. AI agents were like the buzz of 2025.

I think world models could be that for 2026. AI agents will still be because we most people if you ask them like if you I was in a conference you know and I asked you know who who's uses an AI agent and there's like you know in a room of a hundred people there were like four right yeah but I think when you know two or three years from now like everybody in that room will and five years from now probably all of them will have at least five working for them but world models is the thing that I think is likely to be to get some of the buzz in 26 because and they've said this, you know, the nextgen AI, whatever you call it. I don't love AGI is a term. Super intelligence I like even less, you know, but I think what what they're really trying to get at is like next level AI that doesn't hallucinate as much, that has more capabilities, that can operate a little bit more independently.

And if you think about that, you know, it's going to it's going to need more than just language models, right? It's going to need like like you guys were saying earlier, it's going to need something that can understand you know more multimodal inputs, video, audio, photos, you know, mapping data, those kinds of things. And that's where I think world models could could be the thing that really drives the next, you know, level of innovation and the next leap forward.

Ryan Eppley: Yeah. I mean, you've you've spoken a lot about world models. You have a similar take big next thing coming up in 2026.

Rob May: Yeah. I would broaden Jason's take a little bit and just say that like I think moving away from LLMs, right? So, so people forget there's all kinds of models that we still use, right? People still use recurrent neural nets, they still use CNN's there's still a lot of things other than transformer-based LLMs. Statebased models have gotten more popular this year. And so I think I think 2026 will be a year where you start to see start to see world models. You start to see more usage of SLMs. Like you start to see some technologies that move away from transformer-based.

There's a lot of hybrid models that have come out. IBM has a key one. Actually I think didn't Google launched something I don't think we covered it here, but I think Google launched something that was Mamba plus transformer mix. You know they Google did a research paper last year on these recursive models which is a cool idea. So there's there's a bunch of stuff floating around out there and everybody's looking for an edge and that's going to encourage some people to try these things. So I think 2026 will be the a year when LLMs become just I think LLM have been the thing that people associate with AI and I think they'll become a thing. To your point earlier, we're still going to use the frontier models. They're still great when you're like, you know, I don't know what to do with this cat. Like there's still use cases for it, but I think you're going to see and I'm biased here because I run a company that, you know, is based on this whole thesis, but like but I think you're going to see people fragment into using the right model for the right thing.

Multi- multimodel systems, I think, you know, when I look at when I look at companies on the AI maturity curve, everybody starts they build a PC on a single model. Y then they build you know then they launch in production and they support like you know Google and Anthropic or open Anthropic they you know they have one they go to and then a backup but as they get sophisticated and they see more data they start to break out their workloads to more models and then you know the companies you see I can't comment on because there's a lot of private discussions but some publicly traded companies that are very sophisticated have you know 42 different models in production that they use some fine-tuned SLMs of their own some Frontier models from older models like GPT3 they just they know the right model for the right task and that's where everybody's going to go over time because this whole industry has to get more efficient to your point earlier in the conversation

Jason Hiner: I see this a lot I obviously you know you all talk when the deep the deep view covered you know that your the leaderboard that that you and your team did but that conversation was one of like a bunch of conversations that I had with people working on this idea from different from different vantage points that as enterprises start to use AI and start to use these models to do really drive efficiencies to drive new gains new innovations in the enterprise they come across this problem which is that inference is really expensive and so they're like how do we take this and do it but let's do it for a tenth of the cost and it comes down to we have to use more specialized models or we have to do caching or we have to do other things but we can't just run everything on sort of the frontier model on the cutting edge hardware because all this that gets really expensive really fast and we start paying a lot of you know token costs essentially and other other costs.

So, so I see this as one of the things that's going to happen a lot next year too. I would say using sort of the best model for the job isn't always going to be the frontier model. I mean, we've seen there's been some reports just in the past month that there's a lot of US companies that are building on DeepSeek and Qwen because they're open- source openweight models and they're a lot less expensive, you know, to run. And but even on running on older models, running on older hardware, there are a lot of workloads that don't necessarily need to run on those super expensive models. So you can do it for cheaper and get better performance in some cases, which is like incredible.

Ryan Eppley: So before we jump to our New York question that we always ask, let's end on one more AI question, which is what do what do you guys think is the most overhyped thing of a in AI in 2025? Like who who's the who's the faker that's going to be exposed in 2026 or or what's the what's the story where you're like everybody made a big deal about that but it wasn't a big deal you know I know the markets are saying a different thing but you know I have to imagine AMD is going to come out swinging in the next couple years like I feel like the 2025 thing is like Nvidia is just a unstoppable untouchable entity and if you just zoom out long enough and Look at other technological revolutions like Ford obviously was a massive car manufacturer and it kind of dominated, right? But other car companies came along and built up capacity pretty quickly.

And I feel like, you know, I have a lot of love for Nvidia. I've been building computers, you know, for 20 years since I was a kid in a basement. Like Nvidia is something I love as a brand, but like you know AMD I feel like is going to come and there there might even be others that that join this kind of race, but I think the biggest thing of 2025 is people thinking Nvidia for the next 10 years is unstoppable. Nothing is going to, you know, keep them from from being the kind of 90% shareholder. I think there's a lot to be said over the next couple years that AMD is going to come in and do a lot of damage.

Jason Hiner: Yeah, I think mine would be AGM and super intelligence. I already said it earlier, you know, I think that I think it's a little bit of a fool's errand, you know. I don't know. And it is a marketing term. I do respect like what Demis and his team at at DeepMind are trying to do, which is which is build what I think of as just like a really much smarter version of AI that can do things that that are even really challenging for humans to do. But and it's funny because a conversation we had, Rob, made me think about this because I think it's a good example. You know, you talked about when you think about models, think about people. Think about you take think take the smartest person you know. Like you don't go to them for everything, right? You don't go to them to ask like if you went you said Stephen Wolfram I think was the smartest person you've ever met. You like but you wouldn't go to him to ask like who do you think is gonna win the Heisman? You know this year you wouldn't ask him you know who wouldn't hire him to do my basic accounting like Yeah.

Exactly. So models that are kind of like that and even I think the way we think of like super intelligent models I think one model to rule them all. I think that idea has kind of been poo pooed in 2025 that that's not the that shouldn't be the end goal and it's not the best use of resources. And so I wonder if all this conversation around super intelligence and you know AGM starts to shift and we go you know what that maybe wasn't even the best path. Maybe maybe that's not the right path to think about it because domains in some ways need specific knowledge you know and not sort of one thing to know everything.

Rob May: Yeah. There there there are cases like like tax law, right, where whether somebody's smart or not, like you just need somebody that studied it and then has expertise in it. And yeah.

Yeah, you know, I think for mine, I mean, OpenAI is never going to sponsor this podcast because I'm like such a hater. But maybe they will. I maybe I'll be like, I just think I just think they're overextended. I think they're doing too many things. I think they're unfocused. I mean, they're launching a brow and like I get why they're doing it. Are trying to like boil the ocean and have their hooks and everything and be too big to fail. But like it's you know even at that size like if I talk to my friends that work at Google they still have capital constraints like they still don't have all the money to do everything they would like to do and they're still prioritizing things and I feel like OpenAI is not prioritizing things and I think no matter what size of a company you are you always have to do that. That's why they're not public.

Yeah. You can prioritize more things when you're bigger and you have more business lines and all that, but I you know I just I just think they're struggling with focus and you know I like if I was running that company I would probably actually stop fighting the model game and I would focus on user focused consumer AI. I think that's where they've built a great brand and that's what I would lean into. I that's the real asset. I don't think it's that they need to win the model game. I think they just need to be, you know, in that conversation. But I don't think but I think they want to win in multiple games and that's hard. That's really hard particularly when you're private and you're small and you don't have a lot of assets.

Jason Hiner: Consumers don't I think completely understand all the things that AI can do and the thing is if they they are really trying to become a consumer AI company and that's where I that's where I think it's challenge because essentially you're going to have to go up against Apple you're going to have to go up against Samsung you know even you're gonna have to eventually go up against Google which has a great you know consumer brand as you're talking about like there's so trust. Consumer brands are ultimately about trust. Like who do you trust to with your data, with your you know, so much marketing has to go into that, right?

Ryan Eppley: It's like you have to see a brand seven times before you even like have an open conversation with them, right? Something of that advertising. And you know, we got to imagine 2026 will bring better names to this whole AI kind of domain. Like what is going on? Obviously ChatGPT was made fun of so much about like the naming conventions with all their different models. Claude code got a lot of like, you know, hey, like what what's going on with Claude? I understand the, you know, purpose behind it. It's a great, you know, shout out to Vendor, but yeah, I have to imagine we'll get better names over the years, right? Like, you know, it's coming and ChatGPT and Claude, you know, you know, and I don't even like Gemini, to be honest.

Jason Hiner: Yeah, Gemini is essentially because like Gemini is twins and it's like Google and Deep View or sorry, DeepMind coming together. And so I made the opposite mistake you made at the beginning.

Rob May: No, it's funny. It's funny.

Ryan Eppley: Yeah. I think they all have clever underpinnings, don't get me wrong, but like you know, if if you're going to be a consumer, it's got to be a little catchier. Yeah. Yeah. We got a question teed up for you like we do for all guests. New York centric you know question. What is your least favorite tourist attraction here in the city?

Jason Hiner: This is I love the question because there there are a lot there are a lot of tourist traps in New York.

Ryan Eppley: So many sometimes they even have people outside the tourist attraction that's free to go to asking you to pay 10 bucks. I've seen so many homeless people doing that and it's actually really clever, right? But that just shows you how many tourist attractions and traps there are. What is the worst one?

Jason Hiner: It's true. It's true. You know, if if I can start with some of the ones that I thought would be the worst ones that ended up being the best. One of the ones that's always given as the worst is the Statue of Liberty. And I've only been one time. I went once with my daughter and like and I was like, I'm I'm sort of ashamed to say how much fun this was. Like it's great to walk around like amazing view of the city, like all of that. So that was the one I would that I hear the most. I hear New Yorkers tell me like proudly when they like lived in New York for 10 years and then they leave. They're like, "I never went to the I've never been I lived here for 10 years. I never went to the Statue of Liberty one time." Yeah.

Ryan Eppley: I think the Statue of Liberty is really cool. I love the Statue of Liberty. So, I'm sorry it got so much hate, but I'm glad you like it.

Jason Hiner: Yeah, I do. I do. I like that one. Same thing with with Central Park. Like Central Park also gets a bunch of hate. It's incredible, right? Yeah. The other one maybe on the list that I think maybe is deservedly is a little bit of Times Square. You know, that would be mine. It's so crowded. You know, like you're sort of feel like you're you're trying to get fleeced, you know, every every step you take, you know, there. And so, I think the theaters there are amazing. I've been to some of the theaters, you know, I got to see Hamilton. I got to see, you know, Lion King, with with my family and like they're incredible. But having to navigate through Times Square and like brutal middle of summer, it's brutal.

Rob May: Oh, it's rough. And then you think about some of the world's most valuable real estate, right, for for like commercial real estate. And what are the brands that are there? It's all the brands that are like everywhere throughout America, right? Like all of the franchise stuff. And it's like why would you like there's so many unique awesome New York franchises in food and shopping and all of that but instead of them being in Times Square like they're all the other places in New York and so instead you get the same thing. It's like if you came there's an Olive Garden, there's a Red Lobster.

Ryan Eppley: There's an Applebee's even a Margaritaville.

Jason Hiner: Exactly. You've hit all of us. So why would you come to New York and go to those things like you because you can pay double the price here, you know? Exactly. So, I think of that as like probably, you know, the one that I sort of makes me you know, sad that people come all the way to New York.

Ryan Eppley: Funny cuz we were we were chatting right before the show about it and like we both agreed it was Times Square, too. So, it's funny hearing you double down. Like, it's rare that all three people on on the show have the same one. When you live here, you don't go through Times Square unless you have to like like you're going to Broadway show and you have to like get off at 42nd Street or as someone that's a fast walker to like the death of me, it's really hard to be in Times Square. I'm just like constantly like, "Oh my gosh, like I'd almost rather be in traffic and I hate driving." So that says a lot about Times Square.

Jason Hiner: Yeah. For for 5 years, our office the office I worked out of was at 15th and Fifth. So it was down, you know, near NYU. And I really for most of my career had spent most of my career either in lower or Manhattan or Midtown. And when when I had those five years where I was there, for those five years, I never came to Midtown or Times Square except unless I had a meeting. But most of the meetings were never around Times Square. And I realized like there's so much of New York that's so you know amazing sort of outside of that and loved it and was so impressed, you know, by it.

Rob May: Yeah. Well, thanks for being on the show, man. This was really fun.

Jason Hiner: My pleasure.

Rob May: Definitely. And if you're not subscribed to the Deep View, go to thedeepview.com.

Jason Hiner: Right. That's right. Or subscribed at the deepview.com is a really easy way to just drop h. So that's how it starts.

Rob May: Yeah, check it out and hope you guys hope you guys enjoyed the show. Go watch the other episodes and have a happy new year. We'll see you in 2026.