Today on Decoder, I’m talking with Hayden Field, The Verge’s senior AI reporter, about a question that’s been rocketing around the tech industry for the past week: Is Google losing the AI race?
That’s because last week Google announced a bombshell reorganization of its AI division, Google DeepMind. Jeff Dean, the company’s chief scientist, is leaving to form his own startup and DeepMind cofounder and CEO Demis Hassabis is stepping aside to focus on longer-term research.
You can read these moves, and Google’s reaction, in a lot of different ways. So I really wanted to sit down with Hayden to dig into some of the smartest analysis we’ve seen this past week, and what we think is really going on here.
Also: At its core, this is an org chart story — and what is Decoder about if not org charts?
Okay: Verge senior AI reporter Hayden Field on what’s happening at Google DeepMind and the future of AI research. Here we go.
This interview had been lightly edited for length and clarity.
Hayden Field, you’re The Verge‘s senior AI reporter. Welcome back to Decoder.
Thanks. It’s great to be here.
It’s always chaos when you’re here, Hayden.
It really is. There’s no week off. Every week something crazy is happening.
This time is a little bit different. The news happened last week. Google reorganized its AI division, Google DeepMind. Demis Hassabis, who was the head of DeepMind, has ascended into Google heaven where he’s now the chairman of DeepMind. He’s going to focus on bigger research. Jeff Dean, who had started Google Brain, left with a bunch of other people from Google to start a new lab that will run on Google Cloud.
This is a big reshuffle of Google’s AI efforts. It comes as Google is not competitive on the frontier anymore. There’s a lot of reactions to this news. I wanted to just go through some of that reaction with you — you’ve done a lot of reporting on what’s going on with Google, what’s going on in the industry — and try to put this all into context for people.
So let’s just start at the start. Google is by and large the best-positioned company to win AI. If you just look at what it is, how it makes money, its distribution power to put AI in front of people, putting AI in Search, its resources, it feels like it should have always been the winner.
As you reported from the Elon Musk-Sam Altman trial, everyone was afraid of Demis Hassabis the whole time. You look at all of these ingredients and you think, “Google should be the runaway winner.” Instead, I would say over the last week, a lot of people have said, “Google’s going to lose,” which I find fascinating.
Is that your diagnosis of what’s going on with Google? Is this just a catastrophe for its bleeding-edge efforts?
It’s really bad, but I do think people are forgetting how powerful Google is and how much they own. All of the integration they can do, all of the tools, all of the data they have on us — they have a huge safety net, is what I’m saying. A huge cushion.
Even if they make huge mistakes and are falling behind, I still think you can’t count them out completely because they have so much power. However, it’s crazy to me that they have that much power and all that going for the company, and they’re still not in the lead, let alone being behind as of right now. That’s what I think is crazy.
And Sundar and Demis both made that point in their statements last week. Sundar wrote that the company’s committed to being at the frontier and they’re focused on the areas they need to improve, which is CEO-speak for, “We know we need to get our shit together.” And then Demis wrote that the company’s “entering a next chapter” and that “it has the ingredients to lead from here, and I firmly believe we will,” AKA, “we aren’t yet, but we have the ingredients to do it sometime in the future.” Both of them acknowledging that is pretty crazy.
It’s not looking good for them, but it reminds me of someone with a trust fund. They have a fallback. Don’t count them out yet.
That trust fund is Google Search. They have a product that billions of people use and that they’re putting AI into. As you and I have discussed many, many times, that is a consumer product. Consumer AI has not yet reached a point of making a bunch of money for a bunch of people, but everyone sees that as the prize. All of the action is in enterprise AI right now, and you can see Anthropic is out to a big lead because of their focus on enterprise AI and in particular coding.
One of the things I want to come back to is whether Google made the wrong bet on multimodality and world models, which Demis is really focused on, instead of these core enterprise use cases and whether they even want to be at the frontier. They’re saying they want to be, but maybe they don’t. Maybe they just want to sell cloud services to Anthropic, which is a business that is growing for Google.
But hold that thought. I want to come back to that. Just put that in your brain. This is part of the framework that I’m thinking about. You just said Sundar said, “It’s the next chapter. We’re committed to it.” They’re making changes.
But Hayden, a few months ago I was sitting with Sundar Pichai and I asked him, “Hey, you just restructured this entire company. You got rid of all of your senior executives. You have all new ones. You made DeepMind.” He said, “Yeah, that was hard.” Here’s a clip:
Sundar Pichai: It was tough to convey it outside, but I pivoted the company to be AI-first. We had all the ingredients, so in some ways I felt like the Overton window had changed. People were adopting these technologies faster than we had expected. To me it was a way to go and actually express ourselves through our products, but I realized we had to organize ourselves for it. And going back to my earlier point, I realized we need a core model and a core infrastructure team to power everything we are doing across Google. A lot of my initial energy was to go set that up.
To get one AI team, we had world-class research teams in Brain and DeepMind and brought those together as Google DeepMind, which was harder than it sounds because it’s like saying, “Go put Stanford and MIT together and create a department out of it or a university out of it.” So I think we’re doing that well.
That’s Sundar saying, “My company wasn’t organized for the AI moment. We were slow to react to ChatGPT. I made this huge change to build a central infrastructure team.” It’s three months later, four months later, and the head of Google Brain is leaving. The head of Google DeepMind is becoming the chief scientist.
Now, Google DeepMind doesn’t have a CEO anymore. It has an SVP. What do you read into that? Is that yet another reset or is it just more of the same?
It’s another reset, but it’s hard because I really do think it’s both. It’s a result of the fact that Demis has never been interested in the productization stuff. He’s, like you mentioned, really into world models, research, curing disease, and drug discovery. And Jeff Dean is similar. So I’m not surprised.
If you’re being optimistic, you could view this as just the ripple effects of the conversation you had with Sundar, but I also don’t think this bodes very well for the company because, as we’re seeing with OpenAI literally right now, whenever there are a bunch of executive shakeups, people get nervous. People below them leave. People don’t work as hard because they’re thinking too much about things. Plus, a lot of the people that are working on AI at Google are there because of Jeff Dean or Demis. So what’s going to happen now?
They’re thinking about this solely on the page, logistical. “Okay, who is best positioned to speed up our productization?” Maybe that’s the best decision. But when you take into account all the minds that are leaving and all the minds that will leave in the next couple of months, it’s going to be rough.
There’s a lot of controversy, or I guess curiosity, around Gemini 4, which is not yet out. They announced it at Google I/O. They put all their emphasis on the 3.5 class models. The 3.5 class models are not at the top of the leaderboards. They’ve actually fallen quite far behind.
Does this strike you as the actions of a bunch of executives who think Gemini 4 is going to vault them back into the lead? Or do you think people are getting out because Google isn’t as committed or focused on being on the frontier?
As SemiAnalysis wrote in its blog, “These are not the actions of people excited about Gemini 4 Pro.” So definitely not. Google is pretty committed to being a frontier lab. It’s more, is it going to reach its goal? Though my answer to that a few months ago would’ve been that it was a frontier lab, right now it’s more of a wait-and-see. It is right now, but is it going to fall out of favor?
SemiAnalysis put it more heavily. They said in their release, “We believe DeepMind is no longer a frontier lab.” To me, it’s still got that designation, but in a few months that might be very different and I didn’t anticipate that changing so quickly. There’s a chance it could go the way of Meta where they suddenly get pretty behind and they keep promising things and then taking a while to deliver. Same with Apple, actually. We’ll see.
I’m glad you brought up SemiAnalysis. It’s an investment research firm. They do great work across AI and chips and thinking about how the math of all this should work out. That post really struck me as well.
I just want to reuse some other parts of it. There’s a lot of this in which The Verge has covered Google for a long time and there are echoes of this throughout our coverage of Google, and there are some parts of this which feel new to me — they’re new problems for Google.
So I’ll just read you the longer version of the quote you just said. As Hayden said, they wrote, “For all intents and purposes, we believe DeepMind is no longer a frontier lab. Google will continue meandering on and releasing models, but their odds of reaching the state of the art again have dropped to zero.”
This strikes me as an incredible claim. We see the model race twist and turn all the time. And to say, “Google will never once again hit the state of the art” feels like an enormous claim. That’s the part that feels new to me. Week to week in AI, anything could happen. Do you think there’s evidence for that?
That’s really putting the cart before the horse. It’s been a few months since we ourselves wrote about Gemini 3’s hype. It was winning the model race for a few weeks until something else came along. I definitely don’t think the odds have dropped to zero for them reaching state of the art again. In a way, they’re cutting out some of their research focus and just going all in on products. If they’re really going all in on that and they find a way to put all their efforts behind that, they will reach state of the art again.
However, what this will be is them constantly catching up right after another frontier lab. And then the other frontier labs that invest in research more and long-term stuff are going to be pulling ahead. It’s kind of the same thing as if you’re a writer and someone keeps copying you, they’re always going to be right behind you. They’re not going to be thinking of the next thing like you are.
That’s what I foresee ending up happening here. If you’re not really investing quite as much in the longer-term research and all these niche areas that, for example, Jeff Dean’s startup is going to be looking at, I feel like they could reach state of the art again. But it’s not going to be them coming in the lead over and over again.
That’s the part that felt new to me. I don’t really know how to evaluate that claim, especially as Google’s AI division changes leadership. I can’t read Sundar’s mind. I’ve tried many times. It’s very difficult to see what’s going on in his head. I don’t know if he wants to win that race or if he thinks Gemini is good enough now to make a bunch of money inside of Google’s products. I really don’t know how to evaluate, “They’ll never be at the state of the art again.”
The part I do know how to evaluate is the second part of the SemiAnalysis claim here, which anybody who has paid attention to Google will immediately understand and immediately agree with this, I think.
They go on to say, “Perhaps there’s some world in which Google reaches the state of the art again, but we think the odds are basically zero… The issue with Google was not Jeff Deam or Noam Shazeer,” who we should talk about, “but rather their extremely bureaucratic, painfully slow, and strategically timid culture.”
That’s Google. I literally sit with Sundar once a year and say, “Tell me about your culture and tell me about this bureaucracy.” There’s something about Sundar saying to me at I/O, “I made DeepMind the center of the company and then everything else is productizing their development,” and then all the people leaving and SemiAnalysis saying they’ll never be at the frontier again because their culture is so bureaucratic, that is uniquely Google to me. It feels very familiar.
I know you’ve talked to people at Google since all these announcements. Is there a sense that Google’s culture needs to change even more to go and compete?
There’s a reason that so many people are flocking to Anthropic right now. Anthropic is not perfect by any means. They have a lot of red flags as well. But I’m seeing a lot of people from OpenAI, from Google, from Meta flock there because, as we’ve reported a bunch, people in this space are a little bit post-money. A lot of times they don’t care just about the salary or what they’re making. They care about working for a place that aligns with their values that they believe in. Google has really rubbed a lot of people the wrong way when it comes to that.
I could see a lot of people fleeing in the next few months because Jeff Dean seemed to be, according to my sources that I spoke with there, the person keeping a lot of people at Google. I could see a big exodus happening in the next few months.
This is one of the bigger questions. Who was the keeper of the morals and values of this team inside of a Google that seems to be more willing to play ball with the Trump administration? With the Department of War, and with the uses of AI that some people are very, very skeptical or very, very skittish about?
There’s been a lot of reporting about the fact that when Google acquired DeepMind and Demis, they signed a pledge to never use DeepMind technology for military purposes. That has been watered down. In fact, Jeff Dean recently said, “We should not use AI for these purposes.” There are people inside of Google who’ve sent a letter. There seems to be a lot of big-name talents who are saying, “We don’t want to commercialize this stuff in the way it’s being commercialized right away.”
The problem is that it’s the governments and the militaries who are spending all the money right now, and Google is good at making money. It wants to fund AI development with actual revenue instead of raising endless amounts of debt like the startup frontier labs are doing. Is that conflict navigable for Sundar and whoever’s running AI now, or is it just the reality of being a huge corporation that funds its efforts with revenue?
It’s unfortunately not surprising to me, for a huge corporation in the US where you have to maximize shareholder value. They’re subsidizing a lot of their AI efforts right now, it seems like, to me. For example, the other day I vibe coded a new website for myself and Gemini was the only tool that I could use the free version of for nine hours. I was trying to do it as an everyman, so I wasn’t signing into any of them. I just said, “Which one will let me go with the free version and do a ton of stuff?” It was only Gemini.
They’re really trying to get people in their corner here and trying to change developers’ loyalty, change engineers’ loyalty over to them. So they need money. And to be fair, I do think they could do this a lot differently if they wanted. They’re one of the largest corporations in the world and they’re making a lot of money in other areas. They don’t have to be doing this, but that doesn’t mean they won’t. And I think they will continue.
But I also think, like I mentioned, that’s going to lead to a lot of brain drain. One of the current Google employees I spoke with said that a lot of people around him felt like losing Jeff Dean was a big blow for morality at Google, and that it also continues a brain drain they’ve been seeing and that mediocrity is the likely result. So like what you were saying, they may be always catching up from now on and never forging a path ahead like they had been six months ago.
I think this brings me back to that framework I mentioned at the very beginning that I wanted to put a pin in. Does Google want to win at the frontier or can they just sell Google Cloud at enormous high rates to Anthropic and OpenAI while maintaining usable models for its consumer products at high rates?
There’s a world in which Gemini is there for you to search your Gmail and provide AI overviews and be cheap enough to run to let consumers vibe code websites for nine hours at a time. But the real money is in selling TPUs in Google Cloud to Anthropic and Anthropic is going to go collect the government money and Google is just there as a vendor. They’re just selling to Anthropic and OpenAI and whoever else needs data center capacity.
We can see Google Cloud as a business that is exploding. They don’t seem to mind that they’re selling their own capacity to Anthropic instead of Google’s own products, which is one of the challenges of running a company that has frenemies as customers.
How do you see that playing out? Do you see Google ever saying, “Look, we have to win at the frontier. We have to win at AGI. We have to make sure recursive self-improvement of coding occurs so the next model works,” which is a religious belief inside of OpenAI and Anthropic? Or do you see them saying, “Actually, being one or two steps behind the frontier is fine because that’ll make our products work well enough and we’ll just sell capacity to the people chasing the AGI dream”?
The way they are right now, they’re going to keep chasing it because their investors want that. It makes their stock go up to say, “Hey, we’re chasing this. We’re at the frontier. We’re the only large corporation that’s almost or nearly or equally caught up with OpenAI and Anthropic.” They don’t need to go the way of Amazon and, in my opinion, Meta right now.
In a year or two, if they continue the brain drain and they keep falling behind, then yeah, maybe they’re going to do what you said in situation two. But for now, there’s no reason for them not to keep chasing, especially because that’s a great thing to talk about on an earnings call and it makes all your investors happy.
The AGI piece is really interesting to me, especially in the context of Google. Demis Hassabis was the final speaker at Google I/O. After an hour and a half of very practical demonstrations of AI inside of Google products, Demis came out and he talked about protein folding and drug discovery, which are the things that he really cares about.
And then he ended the entire presentation by saying very confidently, “We are at the foothills of the singularity.” Google doesn’t say, “AGI.” It says, “the foothills of the singularity,” which is an enormous claim.
I thought, “Boy, that’s an incredible way to end Google I/O.” And then later I sat down with Sundar and I said, “Do you agree we’re at the foothills of singularity? What does this mean to you?” Here’s what Sundar said to me.
Sundar Pichai: Demis and I have had long, deep conversations on this topic.
Nilay Patel: I figured.
In this context, for him, the advent of AGI is what he thinks of as the singularity and I think—
Do you have a definition of AGI? Have you debated? Do you have an agreement?
We debated a lot. Both Demis and I are very close to how we think about AGI … There is a harder definition of AGI, which is that it has to more comprehensively do the wide range of tasks, including cognitive tasks, in a way that’s comparable. We’ll at some point actually put it out as a company, and we are working on that. But that’s what he’s talking about in this context.
I’ve been thinking about that answer since it happened. Do you know what Sundar’s definition of AGI or the singularity is based on that?
It’s the exact same as everyone else’s. In the OpenAI lawsuit, in the materials, we saw the definition of AGI come out of that. It was published and it sounds like he’s talking about the exact same thing.
In Microsoft and OpenAI’s 2019 contract that was made public as part of the Musk v. Altman trial, it’s a 36-page agreement, but luckily, finally, we’re finding out their definition of AGI, which is, “A highly autonomous system that outperforms humans at most economically valuable work.” It seems similar to what Sundar is saying in terms of cognitive tasks. It’s just a highly autonomous system that can do a ton of economically valuable knowledge work at the same level or surpass the level of humans.
That’s the definition I’ve been working with for the past six years on the beat, and it seems like that tracks with what he’s saying.
He said to me that Google would put out a definition of the singularity or AGI and they’re working on it. As far as I know, they have not yet released this definition. I don’t know if it tracks the OpenAI one. I do know that’s basically what the industry says, right? That it’ll be better than you and me at economically viable tasks. They’ve stopped saying that.
Equal or better.
Either way, we’re losing our jobs, right? They’ve stopped talking about it because that means we’re all losing our jobs. So that’s just faded into the background.
I bring this up because Demis is so focused on science, and in particular health outcomes, and in this announcement about him moving on to be chief scientist, Sundar repeated the phrase, “We’re at the foothills of the singularity.” It feels like Demis wanted to chase AGI, world models, protein folding, and all of the next things.
The thing that makes AI economically viable today is writing software code or automating business processes inside of Fortune 500 companies or agentic shopping or whatever boring thing Google needs to do to make money, and those things are getting ever farther apart. If you want to chase the foothills of the singularity in AGI, maybe that has nothing to do with how Google needs AI to make money today.
That’s a big part of the Google culture clash, right? That’s, “We’re going to chase money and shareholder value instead of holding fast to our ideals.” That’s a big part of the AI industry’s general chaos. How is OpenAI going to make a dollar? Who knows? It’s a code red. We’re right back at the beginning.
How do you see it? Is it that Demis wanted another Nobel Prize, so he led them astray and now he’s getting kicked upstairs and they’re going to focus on money? Or is there real validity to, “We’re going to let him chase the singularity while we work on Gmail Search”?
There’s a big difference between, “We’re going to let him chase the singularity,” and, “We’re going to let him do a lot of research into biology, drug discovery, and the other stuff he’s into,” and that seems to be what he’s going to be doing. It seems like he and Jeff Dean are going to be doing similar things, but in very different places.
It’s a tale as old as time. I’ve been writing about this for so many years, product versus research. The big clashes, the funding, the resources. Companies want products immediately and they don’t want to invest in the long-term research that sometimes leads to those products. They only want the quick, “What are we going to turn this into? What’s the timeline? Let’s speed it up for products and let’s compete with all of our three competitors and not really do these long side projects that may lead to something or may not.”
The tension’s building and building, especially as OpenAI and Anthropic are about to go public. Google has a lot of investors to answer to. This is a building of the tension we’ve been seeing for so many years. And for what it’s worth, I do think the Demis situation was probably mutual.
I’ve heard rumors that maybe he wanted to leave completely, but it would be too dramatic for him and Jeff Dean to leave at the same time. So he said, “Yeah, I’ll go to this other role and then I’ll leave later.” That’s just a rumor. We’re not sure. But either way, it doesn’t seem like he was totally kicked to the curb or kicked upstairs. It’s more probably a mutual parting where he really is not that interested in products and he wants to do the long-term research, and a lot of these guys are like that.
It feels like a real situation where you have to be careful what you wish for, because when Sundar created Google DeepMind, he had to pick a winner, and he picked Demis over Jeff Dean. We were all waiting to see if and when Jeff Dean would leave because he didn’t win, and now they’re both leaving because Demis realized that actually operating inside of Google is a very different job than just doing the research.
One of the questions I have here in general is about DeepMind itself, which is famously headquartered in London and answers to no one. It’s Google’s big, fancy AI research arm with the big fancy CEO. That’s getting pulled down. It’s being led by an SVP who reports to Sundar now. There are some reports that maybe the center of gravity and the authority is going to move back to Mountain View to Sundar himself. How do you see that playing out inside of Google?
I can definitely see the winds might be changing and the Bay Area might be the headquarters again, at least in terms of power. London’s autonomy has been going on for a long time. I’ve visited that office multiple times. It did have a different feeling to it. It felt like there was a lot less urgency in a good way. It was really collaborative. There were a lot of long-term research questions and it was a very different feeling than the tension and the urgency I feel when I visit other AI labs’ offices.
Maybe Sundar got tired of that and said, “You know what? I need the pressure to rise. I need products quickly. The research can only be done if it’s going to lead to something that’s trackable.” I could definitely see the center of gravity moving to the Bay Area.
The reporting from inside of Google that I’m most curious about is how that culture changes without these leaders. As you were saying before, Jeff Dean and Demis Hassabis were moral leaders in addition to technical leaders and research innovators. They kept Google from doing a lot of things. They’re gone now. In some of your own reporting and your story on our site, there are people saying, “Who will we sell out to next?”
Sergey Brin, who’s deeply involved in AI efforts, supports the Trump administration. Sundar obviously stood behind Trump at the inauguration. The questions from the staff about, “What will we sell it next? What values are up for sale next without the culture carriers of a Demis or a Jeff Dean?” seem open.
Are there answers? Who’s taking over DeepMind now and what kind of person do we think they are?
It’s Koray Kavukcuoglu, and these questions are super valid because I don’t think Koray’s going to be speaking up on this stuff. He’s a product guy. He’s focused on speeding up the product. I’m sure he and Sundar have had a bunch of conversations about focusing on the short term.
Over the past few months, early this year, I was chatting with a bunch of AI employees at a ton of different companies, and they all, including employees at Google, felt like the messaging from their companies was, “Shut up and focus on the mission. Keep your head down. Roll stuff out. Don’t worry about what’s going on in the outside world politically.” It seems like that’s going to be even more of a thing now.
One of the sources I was speaking with at Google said, “Speaking out on moral issues like ICE agents and AI surveillance and autonomous lethal weapons, Jeff Dean was one of the last people who was willing to do that.” And it doesn’t seem like Koray’s going to be really eager to take that place. These questions are really valid and I wouldn’t be surprised if a lot more DeepMind people left after all these changes.
Let me ask you the hardest question of all to wrap it up. There’s a lot of noise in the ecosystem right now saying that Google has lost the AI race. But as we said from the beginning, Google has a ton of advantages here.
It’s making money, which most of the other companies aren’t. It’s selling Google Cloud infrastructure to its own best competitors. It has massive distribution advantages in Search, in Gmail, in the fact that Apple appears to have distilled Google’s models for Siri. There’s a lot here where Google just has structural advantages. It’s playing with house money. It’s a trust fund baby, as you’ve said several times here.
Is it possible to lose the AI race from that vantage point? Or is it really just about if Anthropic gets to a point at recursive self-improvement that they can call AGI, everyone else has to lose by default?
It’s not possible for them to fail. It’s possible for them to change up their strategy. Like we were talking about earlier, maybe they become an Apple where they’re running on Google AI and they’re fine with that, but they’re picking a different company. Same with Amazon. They’ve given up a little bit, it seems like. I’m going to get some flack for that. Meta’s still really desperately trying, but they’re behind.
I don’t think Google can fail. It’s more, do they eventually realize it’s not worth their time? Does the brain drain continue? Do they realize, “Look, there’s no amount of money we can offer to get some of the best minds back here. Let’s just cut our losses”? That’s a little ways off. I don’t think that’s going to happen in the next few months or even the next year.
But it’s important for these execs to remember that a lot of the people that are most involved and really pivotal in building this technology and keeping them at the frontier are extremely serious about their morals and what they’re doing. They don’t just do it for the money. It’s going to be tough to resell that dream to some of them if they’re trying to rehire after they lose people. OpenAI is experiencing the same thing. So we’ll see.
It’s worth noting Demis is still there. He’s the chief scientist. Presumably he will still be going to the London office and be charming, and Demis is very charming. You can see all that continuing to happen, but what are we shipping? What are the roadmaps? Who gets the GPU access? That seems like it will change and that will have some downstream effect.
If you had to look for signs, is Google succeeding or failing, is the culture a problem or are they just retrenching to go and be even more aggressive? What kinds of things would you look for right now?
What I typically look for is what they’re really offering every time they have a big release and how pivotal it really is. Are they doing something that their competitors did six months ago? And are they even doing that well? That’s how I think about these things sometimes. There will be a big release from an AI company or a big tech company that’s trying to catch up in the AI race. Sometimes it’s six months behind OpenAI or Anthropic, and sometimes the stuff that they’re really excited about is pretty bad.
That’s what I would be watching from Google. With Gemini 4, what can it really do? Does it have that same effect that Gemini 3 had on the industry, where it’s leading for at least a week or two? Or does it not even have an hour of lead time and it just falls flat and it’s used by the enterprises that they work with and no one’s really excited about it? That’s the type of thing I’d be looking for.
I’d also be looking for the exodus. If a ton of the DeepMind engineers leave, if they’re desperately trying to hire and offering more and more desperate pay packages, that’s another sign. Whenever Demis leaves — and that will eventually happen — that’s a huge sign.
We’ll be looking out for all that. I’m confident that the chaos in the AI industry will continue, and Hayden, you’ll be back on Decoder. Thank you so much for joining me today.
Thanks so much.
Questions or comments? Hit us up at decoder@theverge.com. We really do read every email!
Bluesky CEO Toni Schneider is joined by COO Rose Wang for a Disrupt Stage session on whether social media can start over and where Bluesky fits into that potential reboot.
SkiaSharp 4.0 is stable, co-maintained by Uno Platform and Microsoft, and brings a newer engine, cleaner API, and up to 30% faster rendering for Uno Platform apps — all on a predictable release cadence you can finally plan around.
If you've built a .NET app that draws anything — text, geometry, images, a custom control that needed just a little more than the platform gave you — you've almost certainly touched SkiaSharp. Maybe you knew it. Maybe you didn't. Either way, it's been quietly doing the heavy lifting under some of the most-used UI stacks in the .NET ecosystem for over a decade. And with SkiaSharp 4.0 now stable, it just took its biggest step forward in years. Let's unpack what's new.
Why It Matters
Why SkiaSharp Matters
Cross-Platform UI Rendering
Here's the thing about drawing APIs — even on a single OS, they don't agree with each other. GDI+ on Windows behaves differently than DirectX on the very same machine. Now multiply that inconsistency across iOS, Android, Windows, macOS, Linux, and the web.
SkiaSharp erases that problem. One rendering API, every platform .NET touches — pixel-perfect text, hardware-accelerated geometry, consistent color output, no lowest-common-denominator compromises. If .NET runs there, SkiaSharp can draw there — consistently.
Foundational to Many Stacks
SkiaSharp isn't some bespoke rendering engine dreamed up in isolation — it's a .NET binding over Skia, the same open-source graphics engine that powers Chrome, ChromeOS, and Android. When Flutter needed a rendering engine, it started with Skia too. That's not a coincidence — that's a signal.
It's also why SkiaSharp sits underneath .NET MAUI and WinUI 3, quietly threaded through the framework in ways most developers never see — from custom drawing surfaces down to build-time asset compositing.
SkiaSharp Is Foundational to Uno Platform
For Uno Platform, SkiaSharp isn't an optional backend — it's foundational. Our promise has always been that a single C# and XAML codebase produces pixel-accurate applications across every platform target: Windows, macOS, Linux, iOS, Android, and WebAssembly. SkiaSharp is the rendering foundation that makes that promise real.
That foundation matters more than ever with Uno Platform Studio 3.0, where we're enabling agentic workflows on the web and across any IDE or CLI — go from a prompt to a running cross-platform .NET app, right in your browser, no install required. AI can scaffold the app, iterate the logic, refine the structure. But every pixel the user actually sees, every transition, every rendered surface — that's SkiaSharp. It's the layer that makes "runs everywhere" mean something visually, not just structurally.
Co-Maintenance
Stability & Co-Maintenance
A Major New Version
SkiaSharp 4.148.0 is the first stable release of SkiaSharp v4 — and it's not a minor version bump. It brings the native Skia engine current through milestone m148 (two and a half years and 28 milestones of upstream improvements), a cleaner and more correct API, and rendering that's up to 24% faster on the GPU-accelerated backend. For Uno Platform apps specifically, early testing is showing improvements of up to 30%.
Microsoft & Uno Platform Collaborating Together
Here's the part that changes the story going forward: Uno Platform is now a formal co-maintainer of SkiaSharp, alongside the .NET team at Microsoft. That's not a branding exercise. Uno Platform's engineering team has a direct, ongoing role in SkiaSharp's release process, triage, and direction — helping stabilize APIs, land variable font support, and benchmark performance.
Issues that affect Uno Platform users — WebAssembly rendering behavior, performance regressions, API gaps relevant to cross-platform .NET development — now have a team with deep context actively working on them. We're part of the process that produces fixes, not waiting on them.
This co-maintenance is one concrete output of the broader technology collaboration between Uno Platform and the .NET team at Microsoft, spanning Android bindings, AOT compilation improvements, and .NET 10 contributions. SkiaSharp is just where it's most visible.
A Predictable Cadence Going Forward
One of the most important things v4 ships with is a release model you can actually plan around. Going forward, SkiaSharp ships on a regular cadence in two channels that follow the upstream Skia milestones:
Stable channel — corresponds to Skia milestones in Chrome's Stable and Extended Stable channels
Preview channel — corresponds to the milestone in Chrome's Beta channel
No more guessing when the next meaningful update lands. That predictability matters — especially for teams maintaining production applications.
See It in Action with Uno Platform Studio 3.0
Want proof this all adds up to something real? Uno Platform Studio 3.0 can spin up a full, running .NET app entirely in your browser, from a prompt, with AI. No install required. Every pixel of that app's UI? SkiaSharp. We've gone from "a drawing API that solves a platform gap" to "the rendering layer underneath AI-generated .NET apps running live in a browser tab." That's a decade of evolution in one sentence.
Watch
Explore What's Possible
We sat down with Matthew Leibowitz (Senior .NET Engineer at Microsoft & SkiaSharp maintainer) and Martin Zikmund (Lead Engineer at Uno Platform) at our SkiaSharp 4.0 launch event to talk through exactly what v4 unlocks and where things go from here. Watch the full conversation below:
What's New
What's in SkiaSharp 4.0
Let's Recap What the Video Covered
Talk is one thing — let's ground it in specifics. Here's everything that landed in this release.
A Better Engine — For Free
The native Skia engine is current through milestone m148: years of upstream rendering, codec, performance, and security improvements that benefit every app automatically, with no code changes required on your end. The highlights:
Sharper downscaled images — mipmap sharpening is now on by default
More accurate colors — transfer functions for Rec.709, HLG, and PQ corrected to match industry standards
Security hardened — modern compiler mitigations enabled across all platforms, all bundled native dependencies updated
Security improvements that flow from upstream automatically, without you having to do anything — that's exactly what a well-maintained dependency should do.
New Capabilities
Variable fonts — full OpenType variable font axis control across SkiaSharp and HarfBuzzSharp. Query axes, set positions, create typeface variants for weight, width, slant, or custom axes
Color font palettes — switch between OpenType CPAL palettes for emoji and icon fonts, or override individual glyph colors
Animated WebP encoding — SKWebpEncoder now handles animated WebP output
Zero-copy stream conversion — SKStream.GetData() for efficient stream handling
A Cleaner API
v4 completes a migration that's been in progress for a while: legacy APIs are retired, the surface is clean. Underneath, the object lifecycle was reworked so native singletons are properly reference counted — quietly fixing a whole class of use-after-free crashes that could occur when the garbage collector finalized managed wrappers during in-flight native calls.
The kind of fix you never see. Which is the point.
Faster Performance
A newer engine shouldn't just do more — it should do it faster. On the hardware-accelerated GPU backend, the work that dominates modern app UIs (elevated cards, drop shadows, layered surfaces) renders up to 24% faster on v4 than on the previous stable release. A busy dashboard of shadowed cards rose from 65 to 80 FPS; a scrolling activity feed went from 47 to 58 FPS. Scenes that don't lean on shadows — charts, text, vector maps — were already efficient and carry over unchanged. You get the upside with no downside.
Procedural Perlin-noise shaders also run about 6x faster on the CPU — a nice win for generative textures and effects. For Uno Platform apps specifically, we're seeing rendering improvements of up to 30% in initial testing.
Modernized AI-Powered Infrastructure
SkiaSharp wraps an enormous Google C++ codebase — syncing upstream Skia milestones, auditing for CVEs, generating release notes and API diffs, keeping docs current. A lot of that work is now driven by agentic workflows.
The point isn't automation for its own sake — it's that human engineering time now goes to API design and correctness, exactly where it should be. It's a big part of how the project keeps pace with Chrome's release train. The test suite also moved to xUnit v3, builds run on reproducible Docker images, and device/WebAssembly testing runs through DeviceRunners. The machinery is modern, and it makes the project easier to contribute to and maintain.
Get Started
Try It Yourself
No shortage of ways to jump in and see this for yourself:
Get started fast at the official SkiaSharp site — docs, tutorials, and API reference all in one place.
See it live in the browser — the Uno Platform WebAssembly gallery shows SkiaSharp rendering live, powered by Uno Platform's Wasm renderer.
Play with SkiaFiddle — write and run SkiaSharp code directly in your browser at the SkiaSharp fiddle. No install, no setup, just code and see results instantly.
Try it in Uno Platform Studio — spin up a full cross-platform .NET app from a single prompt, no install required, and watch SkiaSharp render every pixel of it live.
If you want to try the v4 bits directly in an Uno Platform app today, set the SkiaSharpVersion MSBuild property in your .csproj:
That's all it takes to opt in. Give it a try against your application and share your feedback.
SkiaSharp isn't a niche library for developers who need "extra" drawing capability — it's foundational infrastructure that most of the .NET ecosystem already depends on, whether that dependency is visible or not. With v4 stable and Uno Platform co-maintaining alongside the .NET team at Microsoft, the future here is genuinely bright. Faster releases, a predictable cadence, closer collaboration, and a rendering foundation that's only getting stronger under the AI-native, agentic workflows reshaping how .NET apps get built.
You soon won’t have to look at two Copilot apps in your Windows taskbar. | Image: Microsoft
Microsoft is finally beginning to combine its consumer and commercial Copilot AI assistants into a single "super app" interface, starting with the Copilot and Microsoft 365 Copilot apps. Both personal and work accounts will be moved to the new unified app, which recycles the "Microsoft Copilot" name but features an updated app icon. The single app also means there won't be two annoying Copilot icons in the system tray or taskbar anymore.
"Starting with a fresh look, the updated app combines the best of Copilot chat and image creation, with the power of Microsoft 365 for your work," Microsoft said in its update announcement. The updates are …
A typical apartment building generates data on everything from leaking toilet flappers to tenant demographics, yet most of the real estate industry still runs its analysis by hand in spreadsheets. Fundrise co-founder and RealAI CEO Ben Miller argues the missing piece isn’t a better AI model but a proprietary data layer that general-purpose tools can’t replicate. Miller joined Ben Lorica on Generative AI in the Real World to cover how RealAI evolved from a data project into an AI-powered analyst tool, why the AI acts as an orchestrator rather than a financial calculator, which investment trends aren’t yet pricing in AI’s economic impact on real estate markets, and why he believes the trillion-dollar AI CapEx build is temporarily masking what will eventually be a deep trough in white-collar hiring.
About the Generative AI in the Real World podcast: In 2023, ChatGPT put AI on everyone’s agenda. In 2026, the challenge will be turning those agendas into reality. In Generative AI in the Real World, Ben Lorica interviews leaders who are building with AI. Learn from their experience to help put AI to work in your enterprise.
All right. So today we have Ben Miller. He is the co-founder and CEO of Fundrise. And today, we’ll talk about their interesting AI application called RealAI, which you can find at RealAI.com. Their tagline is “Answer any real estate question in seconds.” And we’ll also talk to Ben about broader trends in the AI industry. And with that, Ben, welcome to the podcast.
00.50 So, I definitely want to talk to you about RealAI in more detail, but I think to take a step back. . . so broadly speaking, as I understand it RealAI does the job of a real estate analyst, but I think, Ben, for our audience, they don’t actually know what a real estate analyst does. So maybe if you can briefly describe what does a real estate analyst do?
01.16 Yeah. So our product’s designed for a professional real estate investor, an institutional real estate investor will typically buy large scale properties that are worth tens of millions of dollars. And then analysts, like any financial analyst, would [do due] diligence [of the] property, make a lot of financial pro forma, do different sorts of investment memos and then make a recommendation to buy or sell. And they also then manage the investment after you acquire it. And asset management is actually where most of the activity happens in terms of how you optimize for rents and occupancy and different financing. And so the real estate analyst’s job is to be the financial brains of the owner of the property.
02.07 And typically this person, when they go about doing their job, they have access to different data sources, they pull it together, they run some financial models, and then they write up recommendations, right?
02.24 Yes, exactly. That’s what the white collar worker today is like. . . manual CRUD operation. Right? So they gather data points from different reports and different online services they might have. They usually populate a spreadsheet cell by cell that way, and then they also will write up a memo. And that is all done by hand, so it’s sort of like handcrafted. And all of that work, at least the first draft, can be done by AI. And then the question becomes, “why our AI versus Claude or ChatGPT?”
03.05 As I understand it too, Ben, RealAI actually didn’t start out as an AI project. It started out as a data project because back in 2023, as I recall, you were starting to talk to me about wanting to start a data business of some sort for real estate. Right? So I guess, for our listeners, Ben, why does real estate generate so much data? I think I remember you telling me that actually, most people don’t know a typical apartment building throws off so much data, right?
03.40 So Fundrise is a fintech platform that democratizes investing into private assets. And one of our original investment classes was real estate, where we would have a fintech website and an iOS app, an Android app, and we would have millions of users and they invest through our platform, and we have a sort of Robinhood-type platform. So we have payment processing and a system of record for ownership, and all of the API services you need and microservices for that company. And then over time we started eating the value chain, and so we started heading from the investor towards the asset. And eventually we ended up where when we acquire the asset, we run the asset. We vertically integrated, so we’re the real estate company and the tech platform, and the fund manager. And as we acquired the asset, we discovered the underlying infrastructure for that asset had a lot of data in it. And the way that people in real estate do their work is sort of how we started. It’s all very manual.
Yeah. So there’s a bunch of different kinds of data sets you care about. At the property there’s a lot of activity happening. So, a toilet might be running and that’s the water bill. You have all sorts of leasing activity. You have marketing funnels of ‘where does the tenant come from’? You have lots of activity around the building in terms of like, who lives there, who’s moving there, who’s leaving, what’s the rent of the property across the street, what’s being built? So there’s endless amounts of data that’s happening at any point in the real world. And that real-world data matters to the person who owns real assets in that location.
05.46 And so then, it seems like the traditional approach was to ignore all of this data? Is that right?
05.56 It had to be distilled so that a human being can pretty much only consume a certain amount of tokens per second, and that usually ends up in a spreadsheet. So what happens is, people at the property or property managers, or maybe there’s our market company, CBRE, will produce market reports and they turn all that data into a spreadsheet so that you might get a 12-month reporting or rent roll or a market report. And so they take a lot of data and they distill it or transform or aggregate it into very few metrics. And that was necessary. And then once a month you get emailed six reports or 12 reports, and those reports are in a sort of dashboard for how real estate people make decisions. And I appreciate this and this is so obvious that there’s just you know, those reports are both a distillation and there’s a lot of meaning lost in the transformations, and a lot of data just left behind because the system of record for real estate and for a lot of industries, really the accounting system. . . income statements and, you know, get operational data like a toilet flapper that’s leaking, and that’s not coming into the report. And how could it?
07.36 So basically, it sounds like what you’ve done is the classic digital transformation of this industry. If you look back to the data in 2023 to today, what is the evolution of this data set? What was in the original data set, and what have you added since then?
08.02 So we originally were thinking about how all this data could be really useful for the real estate industry, the original business we called Basis, and we were originally using it just for our own company. And, there’s sort of property data. If you’re going to make decisions in real estate, you care about change over time. So time series, and you care about comparative analysis. How is my property versus another property? How’s my neighborhood versus another neighborhood? My city versus another city? And so you need to have not just your own data, but you need to have the context.
08.42 You need to have the data of other properties that are in other places. And the entities in real estate are places, properties, and then people, which was that we had this sort of insight when we were in the data gathering property data and place data, is that there’s a data model in real estate which essentially is the property management system, which is basic accounting systems. That’s Yardi and Tratta, RealPage. . . And those accounting systems have an implied data model that is a property address and the unit, but inside the unit is a person that’s not in their data model. And a really big gap in their thinking. It’s sort of like digital marketing. Kind of like TV. . . you’re going to do it on CNBC, you can do it at 3 p.m., but you don’t know who is watching. And then the internet showed up, and you can actually then target with Facebook or, you know, Google. You can target people who have eggs for breakfast and like Mozart. So all of a sudden the people data becomes really critical. And the real estate industry didn’t have any people data. It wasn’t even part of their thinking. It’s not currently part of their thinking. So we started realizing, “Hey, we get the kind of granular people data you have in the marketing industry.” We bring that into the real estate industry because you care about who lives in a building and who’s leaving, who’s moving in, who moving in the neighborhood? Who’s moving into the city. Are they rich or are they poor? How old are they? What’s their gender? All this stuff. Do they have iPhones? Do they have Androids? So we started gathering up huge data sets of people data and then putting that in the same databases you have with real estate data and building out, and you can actually see that people matter to real estate and you can actually make really good insights, new kinds of data, correlations and stuff like that for the sector.
10.43 By the way, this is the kind of thing that the hedge funds have been doing for years. Right?
Right. Yeah. I was actually talking to a famous data tech guy, and we were on the phone, they’re like, you know, we sell our data to all these hedge funds, for like 50 big hedge funds in the world, which probably be Renaissance, and Jane Street, and Citadel, we’ve only ever had one real estate company by our data. I said, “Who’s that? Who’s that?” They said, “Blackstone.”
So yeah, high-frequency traders are data hogs and they put their data into these massive data models, and they do a lot of data engineering and data transformation. And real estate people do it with Excel spreadsheets and do it by hand. And so there’s this journey that every industry goes through. . . real estate’s way back in the stone ages, where they still do everything by hand with people.
11.45 So the end result for RealAI. . . the back end data. . . most of it is structured, probably sitting in some sort of data lakehouse or warehouse, right?
We have two different kinds of data. We have a transactional database and we have one for rows and one for columns. Because we have a lot of data and we do a lot of compute. . . And because mostly people when they’re using AI are doing computations on the fly, we had to pre-compute a lot of stuff so that when people are asking questions about what’s happening, how much growth there is in the market or what’s happening with like rents or whatever the things are, a lot of that stuff has to be pre-computed every night. And that gets pre-computed in a Snowflake database. And we also do clustering algorithms, we clustered people by city and by block and by lots of different things so when people want insights it’s all pre-computed. And then that sits in Snowflake migrating to a Databricks-type database. And then there’s a separate database which is the transactional store that’s on the fly, that’s fast when people are asking questions and hitting our API.
13.00 So, let’s set aside the AI model and whatnot. For listeners who haven’t tried it, I recommend you go to RealAI.com. I use it once a month to monitor my properties. But basically then what happens is you have this chat interface and then I presume the AI model acts as some sort of reasoning layer and also obviously parses your intention and what you’re interested in, but then the heavy lifting is done by the data back end, correct?
13.38 Yeah. I mean, I think most people have gotten here where you do deterministic analysis, real and where you’re writing Python and doing SQL. . .
13.49 Yeah. Okay. So in your case, if you hallucinate there’s consequences, right?
Yeah. So the way we structured our application is that where there’s facts, there’s essentially a field that goes and pulls it from the database. So it’s not coming out. It essentially says, “What was the rent for the last 12 months in this property or in this neighborhood?”
14.16 But the AI model is the one that translates that prompt into some sort of SQL query?
Yeah, but I mean, we used to use Claude and we moved to Vercel, but we have basically a software sandbox and it opens it up and does all that sort of. . . This is for real estate, for most people you say, well, you probably know how to do this, but if you’re going to do an analysis about your home or your property, you know, most people can’t write Python, most people aren’t good at doing SQL queries, plus they don’t have a good understanding of the the database schema. And so we built a semantic layer on top of it. And it really democratizes data science. Originally you said we were a data company, and then AI showed up and the AI sits on top of our data, and that made it a lot easier for people because to do the type of insights or analysis that otherwise they would have no idea how to do before.
15.28 But just to clarify, the AI is not doing any kind of financial modeling or calculation, right?
No.
There’s some sort of forecasting or financial model or some sort of financial computation involved, and that’s not being done by the AI?
Right. The AI is like an orchestrator, and has the tools, right? And the different tools it uses and the tools are the web search or the calculator. I think everybody’s gone here, where I think a year ago or something, people were using vector databases and RAG and we moved away from that because you just want to be where it’s deterministic analysis. You want to just use regular software, to put that tool in the hands of the AI.
16.21 So the hard thing here is really maintaining these data sets in many ways, right? So that is the key, because you have to gather or you have to have domain knowledge to understand what data you need and what data really matters. And, now you have a year or so of history of people using the app and you’re understanding, really what’s valuable.
What’s the hard thing? I mean, it just depends on who you are. We didn’t mean to do this, but luckily we have a lot of data. The data is proprietary. The data is really valuable. And so that gives us a reason to exist because what happened is that we built RealAI and we rolled it out, and Claude arguably made it obsolete. And this is a strategy question. I think the hard thing is trying to build the software business.
17.18 How can it be obsolete if it doesn’t have access to your data?
Well, because people have. . . it’s really hard to get them out of their. . . they’re sitting in Claude. Trying to get it. And the reality is that. . .
17.38 There’s no A/B test. They can’t tell that they can get better results if they. . .
So, our product is really good at, let’s say seven of the 15 things a real estate person needs to do. But there’s other things they need to do that they know our products are not designed to do, and so they need to have a general purpose tool that’s doing all their stuff. And so the day before yesterday, we got approved to be in the Claude marketplace. So we built a connector, we’re building a plugin, and now we’ll go to where the customer is, which is Claude. That’s our current strategy. And you can see it’s getting better. Claude is not as good as ChatGPT at making it developer-friendly. But we can build what’s almost the same as RealAI.com in Claude. And that’s where we’re currently headed. And you and I talk about strategy. I think that’s actually a temporary strategy. I don’t think that’s where the market ends up. But, yes.
18.50 But obviously, as users interact with your app, there’s a lot of learnings there that you can leverage to improve the app itself or even maybe the model powering the app. But in other words, there’s a compounding loop. So the question is will Claude cut you out of that at some point?
I think the hard thing is always the people. I think the technology is actually at least. . . You know, I’m not an AI researcher at the frontier. The hard thing is really building a software product that people become addicted to, and the data engineering and data science and stuff, that’s just a grind, this is work. In the short term, we’re sort of conceding. . . I believe we should concede that the customer wants to be in Claude today, and they don’t want to be in a different application. Claude’s like the hot new hotness. But I think they’re going to get Claude fatigue within a year. I think that the real estate people are lagging the rest of the industry. Most people who’ve been using Claude for the last six, 12 months hit a lot of limitations for enterprise. And I think they’re going to want a harness that’s designed for real estate. But, in the meantime, we’re going to sacrifice some of the things we don’t get by having them in our app. And I actually think they’ll end up going back and forth between Claude and our app, because our app is better for real estate than Claude, but Claude is better for everything else. And the way that the Claude plugin and, and I think it ultimately ends up being a RealAI app in Claude. It wants to go back and forth between our app and their app and wants to be seamless to the user. But after that—and this is really a strategy question—after that, do we end up with an open source real estate fine-tuned, trained model instead of Claude in a year or two? I don’t know, but it seems that would be a good expectation.
21.23 At the end of the day, even if you open up that model, that model still needs the data, right?
At the end of the day, that open source model a year from now. . . when we rolled out Sonnet 5, it didn’t change. It didn’t get better for the customer. Like we just don’t need the Fable. . . it’s actually worse for the customer. So I think that once the open source models are comparable to—I’m just gonna say Sonnet 5 or whatever that generation is—the customer is not getting the premium they’re paying for. They just don’t need it.
22.05 By the way, there’s tons of companies now that focus on helping enterprises get to specialized AI. And basically there’s two branches, right? So there’s the post training branch which comes down to fine tuning and reinforcement, fine tuning, tons of startups there that are coming online. And then, there’s even companies that will help you pre-train models from scratch. I think the trend is toward the inevitable simplification of both things. Which will mean that a lot more of the compute will go towards specialized models like yours. And then less of the compute going to the. . .
22.52 I’m not as confident in that. . . I mean, I’m agnostic. I think we could end up in a world where Claude replaces Microsoft and everything’s inside Claude.
Yeah, it’s possible. But then, I think that the general purpose models increasingly are going to be perceived as too overkill for a lot of things.
23.15 Yeah, I think that’s true in theory, but the consumer is so sticky and it’s hard to get them to change their behavior.
Yeah. If it’s a consumer app. But most enterprise work is just back end. . .
23.31 But I’m dealing with an SMB not like a Fortune 500, generally. And the SMB, in a lot of ways, acts more like a consumer than like an enterprise. And so I hear what you’re saying, though.
23.46 Hey, I have a question for you as far as UX. You have an application where someone is asking a question. It could be an important question for the user. How do you communicate uncertainty or how do you say, “Hey, here’s an answer, but there’s a lot of caveats behind this answer” How do you do that in your case?
24.23 That’s very hard to do. I was just getting that feedback yesterday because even facts in the data world are not 100% facts usually.
Yeah. I mean there’s just methodologies that change the information. Is the rent growth a point to point 12 months ago? Is it a median? There’s just a lot of ways you can get to something that’s true, but not the same answer as someone else. But my actual experience with consumers is that they actually don’t care about that. It’s all really about perceived credibility. . .
25.16 So I guess in the UX then, the question is if there’s some sort of uncertainty or disagreement, do you surface that?
25.28 Yeah, we surface data coverage. We have a whole section in the response that’s about a meta analysis about the data and about where we made decisions. . .
25.39 Or do you say, “This property, we think the rent you can charge is this?” But it’s really more likely in this interval. It’s not a specific value. And here’s the reason why there’s an interval instead of a value.
26.00 AI is really good at that kind of caveats and adding context in and again like there’s. . .
26.12 But it’s a UX problem, though. You have to decide to surface it, right?
We definitely surface it. But I find that the user ends up like it’s really a journey. In the first phase, they’re trying to determine if the data is good, and then once they’ve decided the data is good, they don’t really care about the caveats. They’ve sort of internalized to them how to think about the data. And it’s really a problem in the beginning of a user’s use of it. It’s not really a problem once they become a power user.
26.50 So I think that that’s a UX debate that people will have, moving forward. Because I think basically, like I said, in terms of actual studies around decision-making, it does make a material improvement. Since we have you, we’ll close the discussion. And before I ask a question, listeners, this is not we’re not about to give you financial advice; we’re not financial advisors. So there you go. That’s the discussion. First, first thing then, you have access to all this data in real estate, right? So you have all the patterns, but you’re also obviously an investor in AI and AI startups. Right? So, obviously you’re seeing trends in AI that you think aren’t being reflected yet in the real estate market.
27.53 I know I think you’ve mentioned to me in the past, I think so, as an example, to make this concrete, I think you mentioned to me in the past that maybe if you’re buying property with the assumption that there’s a lot of back office workers moving forward and those back office workers are located in, I don’t know where. . . Phoenix, Arizona. . . Maybe you should have second thoughts. Right? So what are some of the AI trends that you think are not yet being priced into real estate investing?
28.28 What’s good about the trends I’m seeing in the data I’m seeing is I think they’re consonant or consistent with what people would intuitively think. So, AI is causing—in the data—a lot more disparity in outcomes. And so in real estate something used to be a truism. I mean, people will say people always need a place to live. And so you would buy workforce housing, affordable housing, things where if you have housing that’s affordable. . .
It’s just like cheaper is better than expensive, right? If you can offer housing that’s less expensive, less expensive to file, it’s less expensive to rent. That’s good business. And that actually I think is breaking down, which is maybe unintuitive but also maybe intuitive. Is that where the marginal dollars are ending up in the hands of people who have a lot of money and the people who don’t have a lot of money, which is basically the bottom half of the country, or even maybe the bottom 90% of the country. . .
Maybe. You can break a lot of different ways. But the point is that, this K-shaped economy where the normal person is doing worse and the best are doing better is. . . AI is accelerating that trend. And that way that affects real estate is if you’re an investor, you want to actually focus on the high end. If you’re going to be building, you can be investing. . . high end essentially is price-insensitive and the low end. . .
30.15 Because the low end people have no disposable income.
The low end is under strain. And then what is happening in real estate is that strain is showing up as a political manifestation of controlling pricing. And so you can’t make money providing affordably priced housing because the government won’t let you. So the government will let you take essentially excess profits from the rich, but it’s becoming more and more challenging to invest in and build for the normal person, because the government’s intervening in all sorts of small ways that people don’t see around how permits are pulled and lots of impact fees and things like that. I mean, in San Francisco and LA and New York, you see obvious examples of that. But it’s happening everywhere. And in real estate, you usually make an investment over a five year period, maybe even ten years. And so the long term trend is the most important thing. And I think then you essentially say. . . this is terrible by the way, from a social point of view, the societal point of view is very negative. I’m not commending this trend, I’m just saying that’s what’s happening on the ground.
31.40 What about this notion that certain hubs are optimized for a certain type of work? Is that something you think about?
31.54 Yeah. That’s the thing you were I talked about last time. San Francisco is obvious, and then you look at Charlotte. But certain types of people move to certain places to get certain types of jobs and work from home drove that middle income white collar worker to work remotely, or they moved to more affordable places like Charlotte, Tampa, and Orlando, all these places where it’s affordable to live, and they do middle office work, back office work, and that job is getting decimated by AI. In particular, a lot of cities where young people would move there. They’d move to an apartment building, they’d move to downtown. They’d have a new job in a middle office of a medium to big size company. Those jobs don’t exist today. And so they’re struggling with where to live and they have roommates or living at home. And so the real estate that used to be where young people moved to is struggling as a result.
33.08 So I took you on your first Waymo ride. And how about things like that? I mean, now I can Waymo from anywhere.
33.19 I tried to invest in Waymo and I was not yet successful okay. I reached out to the CEO I had 37 connections to and I got no love. But yeah, the Waymo thing is such a good example of diffusion, a diffusion problem because the technology is completely mature and you can’t get it rolled out in any of the. . . DC’s fighting it and Boston’s fighting it. And for it to really affect real estate, it needs to become as seamless as Uber and Uber rolled out way faster, way more aggressively. They obviously took more risk on the rollout. So, you know, if you said ten years from now, how does that affect the patterns of living? I think you’re going to see. . . I’ve looked at this a number of times, but I think it’s premature to make these investments. You want to be investing in the wealthy satellite towns of. . . if you’re doing San Francisco, which is a little bit challenging to the Athertons. But in Atlanta, it’s the Greenvilles. There’s a lot of great wealthy exurbs that have many miniature downtowns that are cute. And so I think you’re going to see these satellite cities become very, very successful as a result. But those satellite cities are going to be mostly for the wealthy. So again, it’s all about wealth, not about affordability.
34.50 All right. Closing question. You’re a CEO. I know you’ve talked about the impact of AI and automation on hiring. And you talk to other CEOs. So as best you can tell, set aside the headlines and the news accounts. What are you hearing from CEOs about AI’s impact on the workforce and hiring? Hiring in particular.
35.20 It’s more mixed than the headlines. I think I’d say it’s both true that we have needed less people, haven’t hired as many people and also have needed to expand as a result of AI. And so I think I’m hearing a “both” situation. So it’s really a question of what’s the net impact? And I feel like it’s too soon to call. I think on net it’s still net negative on hiring, but it’s too fuzzy for me to have a good call on it.
35.58 Do you have any sense whether or not all the stories we’re reading about. . . if it’s particularly challenging for new college grads and people looking for that entry level job?
36.17 We don’t hire those people, and we rarely ever did.
Because the stereotype is those jobs I can just use AI.
36.28 I think of it as this cascade effect. Where we hired hundreds of people over the years, and what happened was there’s this elevator dynamic or escalator dynamic, where the person who was an early grad becomes middle and then they make space for the next person.
Right. And that dynamic. . . AI’s changed it so much because you just don’t need. . . the people who are at the upper middle are so effective. . . And actually, this is something everybody learned in 2020-2021—more people means less work gets done because you have to manage people and you’re needing meetings and remote work is really brutal for information transfer. And so it’s actually way more effective just to do it yourself with AI than it is actually to have five people. And I think it’s actually more satisfying. And so I believe I’m net negative on jobs in the white collar work. And we’ve invested in some data centers that are like $50 billion data centers. And when you go there, there are thousands of people on site working. I mean, it looks like we’re in China. . .
So I think what’s happening is that the job market doesn’t seem as bad because of this trillion dollar a year AI CapEx build. But I think that’s a five-year build, not a 50-year build. And then I think that in the meantime, that’s coring out the white collar worker. And so I think on the other side of this, when the bubble finally bursts, it’s going to be a deep trough.
38.22 So then that talent pipeline . . . if you don’t have the entry level jobs who become the middle managers. . . and so on and so forth. . . So you have the entry level developer who gets more knowledgeable. . . then you don’t have that. Right? So you end up with what?
38.42 Your talent pipeline, it’s a thing. I’ve one of the things I’ve said to the team and this is harsh, but I’m going to say it, but it’s a brutal thing. There’s a lot of people in an organization who are good but aren’t going to be the next great people. And I said, “If we don’t have a talent pipeline anymore, you need to turn those people over faster.” Essentially, unless you were bad. . . bad people managed out, but if you’re good, we kept you. But if you’re great, you get promoted. But now if you don’t have this talent pipeline, you have to change what you expect of the good. You have to turn over some of them, and look for people who are good with the prospect of building that. . . So you have to change the talent pipeline or talent escalator because you don’t have as many people coming through as you used to. So it’s more of an up-or-out dynamic than it used to be.
39.42 Oh, I forgot one last question on RealAI, which is, I think you still do seat based pricing, right?
So that’s it then? If I’m a real estate analyst, you’re charging me $50 a month, but I’m hammering this thing, costing you $1,000 a month.
40.04 No. The seat buys a certain number of tokens. So it’s more like it’s more like usage tiers, is how it’s priced today, but I don’t know how to price it. We’re still early in this evolution. I mean, we’re real estate. . . so real estate is going to be lagging other sectors in terms of adoption. And so I don’t know what the right way to price it is because it doesn’t want to be only usage, either, I think that’s not right. But I don’t know the answer.
40.41 But even the tiered pricing, if I’m on a token matter, at some point I’m already at your top tier price and I’m still busting through it.
40.52 Yeah, that’s a great situation though, because tokens aren’t that expensive. I know everybody’s obsessed with them, but I think tokens are the smallest part of our cost.
41.04 But if you’re charging me $200 at the very top tier, I’m actually spending so much compute. . .
41.15 I think that would be a great scenario to have in the future. . . for the few power users, that’s happening for us. Great. I hope they feel like they’re getting a good deal.
Game engines combine strict control loops, complex state transitions, and real-time feedback, making them an ideal environment for exploring agent-based and AI-enabled system design.
This talk presents a Doom-like simulation used as a controlled testbed for designing and evaluating agentic systems. Non-player characters act as autonomous agents, the game world serves as a shared state substrate, and the engine loop functions as a control plane responsible for scheduling, rules, and constraints.
Rather than focusing on graphics or gameplay mechanics, the session examines architecture: how world state is modeled, how agents perceive and act, where determinism is required, and how autonomy is bounded. A key invariant explored is that all agent actions must be observable, attributable, and reproducible through world state changes.
The goal is to show how techniques from game development - simulation loops, event systems, and spatial reasoning - can inform the design of real-world agentic and AI-enabled applications.
Log in
