Google DeepMind's Logan Kilpatrick: Why the Model Eats the Harness

10,850 words · about 43 min to read

0:00 So, we could edit this set so it looks like we're

0:03 Yes. Yeah. Yeah. I I want this where we were talking off camera like we should do that for the intro because I think it just like makes all this stuff more capable. I've seen these examples of like such subtle nuance that like make me appreciate that it's like the world understanding playing out. I was I was giving a talk um and was on stage with with my friend Tulsi who leads the model team. I had mentioned to someone in the crowd to like edit the video and they had literally like took the picture, edit it with Omni in real time and this like dog came on the stage. In the edited version, the other guests sort of like look down, see the dog, they like chuckle a little bit. This is while I'm like opining about whatever AI

0:40 laugh about your jokes.

0:41 Yeah, that it was not my jokes. They they laugh at the dog coming up. It jumps onto my lap. I sort of like acknowledge the dog. I keep talking. I'm like petting it or whatever. and just like there's like so much subtle subtlety in getting that right and the model crushed it and it's just it very interesting and like still trying to like absorb and digest like what that means for you know the way we make content and all these other things.

1:06 That's so interesting. I'm delighted to have Logan on the show.

The agent harness is Google's new product throughline

1:26 Uh Logan runs Google AI Studio and the Gemini API. You spend a lot of your time thinking and building for the next generation of builders.

1:33 Yes.

1:34 So I'm excited to talk to you about everything from Agentic AI to AI coding, world models, and more today. Um and right off the heels of Google IO. So what better timing?

1:43 Yeah, I'm super excited. Thank you for having me.

1:45 Wonderful. Um let's start with Agentic AI. So, Sundar opened IO by calling this the Agentic Gemini era. What does Agentic AI mean for Google?

1:55 Yeah, it's a good question. I think and we were uh we sort of if you if you followed closely, we did sort of mention some of these things back with like Gemini 2.0, which I think was like a little bit early. And so, I think this era, this like Gemini 3.5 era feels like it's actually becoming true now. And we're in the era of agentic coding uh or agentic products and everything agents as far as Gemini goes. I think for us this agentic layer and I think we we

1 Highlight · Anti-Gravity Agent Harness as Google's Product Throughline

2:22 announced this actually at IO um sort of being powered by the anti-gravity agent harness is this like additional through line for Google that sort of connects all of our products that they're sort of like based on now. And so historically like prior to Gemini there actually like wasn't a through line for the you know probably subund number of Google products that we have the 50 Google products we have there wasn't a through line we had Gemini it became this through line everything is now sort of using Gemini in some way that's now becoming true for anti-gravity as sort of all of the products rebase um to become sort of like agentic native products and like actually taking action on behalf of users and helping them get things done. you see this like new through line emerging which I think is actually really really interesting. Um and

3:06 and sorry help me with anti-gravity is the the IDE right or the nonie.

3:11 Yeah, anti-gravity is is a lot of things which I think is is sort of again uh is is an opportunity for us. Um you have sort of a core IDE you have sort of like the agent first experience if you want it on the web. You have a CLI. You have an SDK. But I actually think and I don't know how much we've framed it this way, but like it really is an ecosystem of stuff that we built and it's designed to sort of like meet developers wherever they are. So you could use it through the Gemini API if you want to and you want a managed agent that you don't have to do any of the the sort of infrastructure work for. Um and then the most interesting bit is like it's not just the ecosystem of anti-gravity stuff. It's also powering like literally it's the the same harness is actually powering all the other Google products.

3:49 So anti-gravity will be powering a bunch of agent stuff in search in the Gemini app um across like cloud and and AI

3:56 studio which is really exciting.

3:58 I see. So it used to be the Gemini API. So like the language model was the through line in terms of how AI gets baked into every Google product. Yeah. And now it's not only the API, it's the the coding harness.

4:09 Exactly.

4:09 Um that's that's being used in each of these products and therefore it's the coding agent itself that's driving more agentic properties.

2 Highlight · Specializing the Base Agent Harness for Coding

4:16 Yeah. In the products. Yeah. Fair description. Fair fair description. I think more generically too it's just like it is the agent harness. I think like coding as sort of like a specialized use case of the agent harness I think is is obviously powerful but it is like coding has proved to be the general purpose agent harness in addition to also working really well for coding.

4:33 Are agent harness and coding harness synonymous or not?

4:37 There's definitely nuance. I think there's like optimization that you can squeeze out of like specializing and actually you see this where like the you know technically the agent harness that gets used for the way that AI studio uses it is like a little bit specialized for you know the vibe coding use case and the the way that the Gemini app is using the agent harness is a little bit

4:54 specialized for the sort of consumer always on 24 by7 agent. Um so I think you have that base harness um that like probably has like 80% of the same stuff and then you specialize for for coding or for whatever the use case is.

5:07 Interesting. How do you think about the cannibalization of the existing business especially now that you are you know

Agentic capabilities drive positive-sum ecosystem growth

5:13 going much more aggressively into agentic properties because I could see for example if all you're doing is search or summarization um there's you know not as much of a cannibalization fear whereas if you're actually going through my emails reply replying to them for me like am I even going through my email anymore and so I could imagine that there's actually just fewer human eyeball hours um uh on your products as a result of having more agentic capabilities. Is that fair or how do you think about the count cannibalization?

5:44 Yeah, it's interesting. I think one sort of observation I have is that like at the beginning and I think Sundar's done a great job of of sort of talking through this is at the beginning of the the sort of current AI era like everyone assumed that AI being able to answer questions for you was going to be like negative sum for for search. Um, and actually what's ended up happening is it's been incredibly positive sum for search. Like people are searching more, people are doing more. And so I think

6:11 and agents are searching too.

6:12 Yeah. And agents actually again there's like this whole market that spawned at the same time that agents are doing more at the same time that humans are also searching more. And so I think it will be obviously there's a finite amount of like human time in the world. Um, but from from like my early feelings of how a lot of this is playing out, it does feel like it's it's very positive some from like an ecosystem value creation like how the human behavior aspect of it turns out I think is like somewhat clear in the next 1 to two years much less clear you know 3 to 5 years from now when the technology is improved and the products probably look a little bit different than the way that they do but ultimately like that is the success of product I think like we we have a bunch of conversations with Demis all the time and it's like the point of building the technology is so that it can go and do stuff for you. Like the point like success for Google like probably doesn't look like you know maximizing eyeball time in front of our products. It's like maximizing outcome for customers to like do the thing that they want to do so that they can go and live their life and do what they want. And so I feel like

7:13 you'll you'll probably see us go down the route of like maximizing outcomes for customers and like not maximizing eyeballs.

7:18 Yeah. I have this term stuck in my head agent growth. Like it seems to me so I'm using using coding agents a lot in my personal time and you know I just let the agent make all the infrastructure choices for me. I'm like I don't care which database you tell me.

7:32 Yeah.

7:32 And and so and the reason I ask is you know it's true in coding today. I would imagine it's maybe going to be

7:39 generally true for a lot of things. Let's say shopping down the line. Um how do you think that's going to change how advertising works, how value capture works for for the aggregators? It feels like it's a very similar trend. This isn't perfectly true, but a lot of these things are just like proxies of each other. Like the way that SEO works, I think like is directly correlated with like the way that um like I forgot what the term now it's like GEO is like the generative engine optimization or whatever it's called. Um and so it does feel like there's a lot of correlation between uh between the things. My guess is it looks like much less of a radical shift than than the than I think maybe what we assume right now just because

Mainstream product agent adoption sits at the crawl stage

8:22 these things compound on top of each other.

8:24 If you were to, you know, grade the scale of agenticness in terms of crawl, walk run, where are we in terms of how agentic the Google suite of products is?

8:34 Yeah, that's a that's a great question. It's definitely like crawl right now. Um, and I think some of this is like all of the inherent product tension for Google is like you have what 13 billion plus user products. And so like I actually think we have some more like labs like experiences where you're probably closer to running uh or walking. Um, but I think like most of the product experience today is is definitely closer to crawling. And I think that's just like the stewardship responsibility we have sort of building a product that's being used by lots of people. Like I don't think the long tale of customers are like ready to have AI running and just doing all the things like they probably they want to be in the driver's seat. They're cautiously taking the first step. And I think the the Google team and like search is maybe like the most quintessential example of this. Like I think they have a lot of responsibility to actually do that in a way that it brings people along and doesn't just like change everything of how they interact with the internet and the way they associate with products and stuff like that. So

9:28 yeah. Which products do you think are closest to the walk?

9:31 That's a good question. I think Gemini app is definitely closest to walk and so for for Spark I think having a 247 always on agent like literally going and potentially doing a bunch of actions on your behalf is definitely like one of the frontier use cases and I think you'll see I think like anti-gravity is another one where it's like you could have autonomous coding agents you know rebuilding operating systems and doing you know billions of tokens and spending thousands of dollars on your behalf and I think those are again like more and actually like they're in GDM as well as in angle of this. So I think like GDM is taking like very much like a a frontier look at this where I think like the rest of Google's products I think are like more incrementally getting there which again makes makes reasonable sense to me.

10:12 Yeah. Do you think that Google ends up with one two three product surfaces for if for using AI or thousands?

10:21 It's tough. I think a lot of this is actually baked in just like how humans consume products and my sense is that there's something nice about like having this like compartmentalization and this like specialization of products where like it it becomes if you end up with a product that is like doing everything for you inherently there's more work involved in using that version of the product I think I think would be like the default state I think maybe somebody will spin together like the truly magic experience that doesn't make that true but I think I think the long tale of folks end up having to spend more mental energy and more time to actually like get the general purpose product to do the thing that they actually want to do versus like there's something nice about I click my calendar app, it just shows me my calendar like I don't need to worry and deal with anything else.

11:06 This is my hot take for why slide decks have existed for so long of just like you know the thing the piece of information you want to be exactly in the same place. Um, and I think like we as humans are just actually very used to that as opposed to

11:19 the idea of a generative interface sounds so cool to me. But it's like do our brains really isn't that just more cognitive overhead for us?

11:25 It definitely is in certain cases and I think somebody needs to again there's there's a lot of incredibly smart people in the world and so maybe somebody will find the experience that like makes it feel more natural. But to me right now I'm I'm maybe not 10,000 is the extreme version. I'm guessing it looks more like more products going after sort of like different uh and maybe the other answer is like I don't know what it looks like for Google for the ecosystem it looks like a lot more products I think like and that that's really exciting I think like how Google will end up strategically deciding like do our customers want to deal with us having 10,000 products or would it be better to only have three um will come down to like a strategic decision for us

12:03 that totally makes sense um when I talk to companies in the enterprise they say you know everyone's talking about agentic AI, but the only place they've seen agents really working is coding agents. Do you agree or disagree with that take?

12:17 Yeah, I think it depends what your bar for working is, which I think is a lot of the nuance of this. Like I think if you're if you're truly trying to like offload very complicated tasks for for domains in which like it's the models haven't actually crossed the threshold of quality, then like I think that's definitely true. Like the it's not going to solve the problem. But this is something that I want I wish we could like measure. A good example is like open router for example is like measuring you know the the total token consumption that's happening. And so you can sort of like see these trends play out over time of like how much more intelligence is in the world you know now versus a year ago. In parallel the thing that I'm actually really interested to measure is like how long is the average like thing the average like agent run or the average task actually taking place. And I don't think it's something that they publish, but I feel like they probably have interesting data. And I'm sure there's others because because I I do think you're like seeing these like new model capability lands or new model drop and and it's like spiking up and and maybe the the

13:18 curve is still like very low right now, but like you're seeing those like early signs of it spiking up where to like longunning tasks and all the model labs are talking about like we released this new model and it did, you know, three days of autonomous work or whatever it is. um that that's the extreme but I think in practice you're seeing that like trickling up like pretty pretty quickly which is really interesting. So even if the enterprises haven't felt it outside of coding like they are going to like this year um as as sort of a bunch of those other use cases get get much better as well

13:49 from like a you know from the deep mind perspective do you think long horizon agents is like a KPI that matters is it they is it the KPI that matters

13:57 it definitely it definitely matters um I think for deep mind like we're doing lots of things which which we can talk more about later like there's you know a huge portfolio of of different bets that are taking place longrunning agents obviously matters a And I think also like specifically coding agents in that matters a lot like it clearly is an accelerant of like every other part of your business if you have a great coding model. Um and so making sure we have that I think is is super top of mind.

Internal dogfooding and anti-gravity drive Google's coding push

14:22 Got it. Um I'd love to shift gears a little bit and talk about coding.

14:25 Yeah.

14:26 Um okay I'm gonna ask a hard question. Uh a lot of my developer friends were using Claude for a long time. Openi saw that declared code red codeex is now really good. I would say my friends are maybe split 50/50 now in using Claude and using codecs. I don't hear a ton of them using Gemini, which has always kind of puzzled me. Um, what's going on with that?

14:49 Yeah, it's a great question. I think there's one there's one part of the story that I'll add, which is which which is it makes it even more interesting, which is uh December the narrative was that Google had won. Um, and when we landed Gemini 3, I think it was like such a such a profound improvement from a model capability perspective. I think a lot of the narrative was like Google has taken a huge leap forward um and and made that happen. And I think what was interesting to see sort of as a as an ecosystem participant uh is like how not how quickly that narrative shifted but just like the next wind of the narrative obviously was like all the agentic coding stuff that happened over the holidays and then into January and beyond. Um, and that was that was not that long ago. Um, and so it's a it is a feel

15:34 like we've been warp speed ever since.

15:36 Yeah, for sure. And but it is it's a meta reminder of like just how fast things can can change. Um, I think the observation is is not is not unreasonable. I do think the what's happening behind the scenes for us is like trying to push the frontier as fast as possible on coding. Um, and so I think anti-gravity actually like is an important part of that. I think one of the takeaways is that it's actually

3 Highlight · Why Building a Coding Product Drives Model Progress

15:57 really hard to make a great coding model for this like um for this developer use case of like really long running suite work if you don't actually have a product that does that. And so I think like Google realized that it's why the sort of like wind surf uh deal happened. It's why those folks came over and then ultimately built anti-gravity and sort of we've been using internally actually and Sundar showed this at IO uh just like the graph of growth of token consumption inside of Google. Um so you sort of like you need that engine to spin and sort of the meta comment again is like the engine is spinning. It takes time uh in order to like actually make model progress. Um but I'm super confident. I think the the folks the group of folks who we have working on code is like uh I describe it as like the avengers of AI internally. Um, and so like it it really is like the some of the best people inside of Google trying to push the rock up the hill on this stuff and it taking it super seriously and trying to push and I think three flash um you know notwithstanding like some of the conversation about like the

16:57 price and stuff like that like is sort of a step towards actually starting to bring a lot of these capabilities um and like the fruits of that labor paying off. Like it's a flash model that's better than any pro model we've ever released from a coding standpoint. Um,

17:11 and the pro models were really good before. So, there's another thread of this also, which is like everyone forgets that there's like pre-training windows. And I wonder like somebody should like track this online, which would be interesting to see.

17:24 Meaning like the big run like what clusters have been available and

17:27 exactly the big the big runs are like uh are an interesting threat of this. And so it like look it might look from an external perspective that like oh you're you're super behind in some way and like actually you you miss all the context of like where the big runs are and where the large pre-training runs are. Um so I think that that also like obviously there's pre-training has historically been like a massive strength for Deep Mind like we have some of the best people in the world and so excited to see sort of the fruits of that labor and and everything else that's happened like 3.5 flash was like all post training gains which is really cool. Um, so a huge uh huge testament to the team that the work that that team did to actually like make the level of gains and like surpass the previous pro model um literally just with post training which is awesome.

18:12 How religious are you all about dog fooding internally? Like are for example are deep mind folks still allowed to use other models or is it like you guys are using the Gemini harness now and we have to make this really really good? Yeah, there's I mean I think people it's so healthy to be using other models just because like it's it's so sometimes hard to like actually gro what's happening in the ecosystem if you're not. So like I use all the models, I use all the products. Um I think like you know uh folks across the rest of DeepMind are doing the same thing. You definitely have to use the Gemini models though.

18:41 Um, it's just like great from a from a feedback flywheel perspective. And it's part of how they get better is like DeepMind has and Google more broadly has like 100,000 plus incredible engineers who are using the models and giving feedback and like it should be a competitive advantage for Google because we have that scale of sort of engineering resources and like the depth of the talent and can run you know AB tests and live experiments and all that stuff. So, um I think you have to use

Narrow super intelligence in coding redefines developer ambition

19:06 all the models, but I think for for the majority of folks, it's like Gemini as the daily driver, which is great.

19:12 Do you believe in this narrative around like a like a soft takeoff of like once you have a good enough agentic coding model, then it accelerates the pace of research progress and like it's a self-reinforcing cycle.

19:23 It seems obvious that that's true, but I don't I maybe I'm I'm too I've drinking too much Kool-Aid that that's that's the case.

19:30 Are you seeing the signs of it yet? Yeah, I mean I you definitely see some signs of this. I think the signs that are like still early is doing this from a model perspective. And I think part of the context of that is like the resource allocation for some of these like larger training runs is just like significant. And so like you you definitely still have like a human in the driver's seat at making those decisions because like you're not going to accidentally, you know, take 10,000 TPUs to go kick off some job that like actually doesn't make that much sense. Um, but from a product perspective, you for sure see it. Like I think we're seeing this on our team, like we've built mobile apps using anti-gravity and like we'll launch them to the world like faster than I think any team at Google has ever built a mobile app. Josh's team did this with the Gemini uh Mac OS app and sort of like end to end delivered an app sort of faster than any team had ever delivered a Mac app at Google. Um, and it's because of it's because of a dentic coding and so it's great from a product perspective. I think you've said in the past that if you could have a system that could build anything with code,

20:30 humans can't compete on the same level and that's narrow super intelligence. Do you think we've reached that point?

20:37 It is interesting. I think um this like narrow super intelligence example is interesting to see how obviously it kind of feels that way for coding right now where like coding is like just so good um that it does kind of feel like narrow super intelligence. I don't know. It depends how you actually end up the details of quantifying this. But I think the important thing is like it to your point earlier, it works incredibly well for code. Um, and so it would be great if it did a bunch of other things, but it's actually just like so impactful that it can be great at code. Um, and so I spent a lot of time just like letting that that fact sort of just like wash over me because I think it's e like obviously building AGI is super important and very interesting, but like building AGI if it sort of like takes away from the story of like the current present capability of the technology I think is actually like kind of a bad a bad sort of like trade-off. And so I'm trying to like always hold these two things in my head equal at the same time which is we need to build general

21:37 purpose technology but obviously it's so

4 Highlight · How Coding Agents Reset Developer Ambition

21:39 impactful to have this thing and it feels like it hasn't taken away sort of it's been one of the best positive outcomes is that I feel like it hasn't taken away from like human developers. Um it really does feel like an accelerant of what human devel like I as a human developer feel like I have more agency in the world. I feel like I can tackle, this is my personal experience, I feel like I can tackle more ambitious problems. I feel like I used to kick around ideas and they were like slightly out of reach. Um, and I would just be like, "Ah, wouldn't it be nice?" Um, and now I have the opposite problem, which is I'm I'm kicking around an idea and I'm like, "I could probably make this even more ambitious." and and sort of it does it adds a different layer of sort of um responsibil or like some a different layer of burden actually because I'm like oh I can't just like do the the sort of MVP of this like I actually need to like go 10 steps further because the technology enables me and like resetting my my level of ambition I think is something that I

22:40 I've also spent a bunch of time thinking about but I think that will happen in other these like vertical super intelligence domains um which will be interesting and it feels like we're going to get a bunch of those before we've like solved like it's almost like

22:54 jagged like jagged super intelligence I think is what we'll end up with.

22:58 What verticals do you think we'll get super intelligent at next?

23:01 That's a great question. I do spend a lot of my time too much time probably thinking about coding these days. So I'll think for a second of like the other the other domains. Um I think part of this is like things that have like better verifiability obviously are like the ones where you'll you'll see the gains happen more quickly. Um, so like things with like math and finance actually like science could be a really interesting one. Like it would be fascinating to see like some of these domains where there's some level of verifiability like actually like really start to take off. Um, which would be cool. And I also think like an important thing in this like broader narrative about just like what a what impact AI is having on the world. Like you almost like want that to be the case in the sequencing of like things that work. you want you a lot of these like really really good impactful positive things for the world to happen um as early on as humanly possible so that like folks understand what the potential positive impact of the technology is. So I think science could be a really interesting one. Yeah, obviously there's all the

24:02 stuff happening right now with like math proofs and stuff like that which I'm not a mathematician so it's it's somewhat over my head but um

24:08 I saw a great tweet the other day uh why did Ardos have so many problems?

24:12 Exactly. That's a that's a good one. I like that. That is that's a good like t-shirt. Um, so funny. Okay, I but speaking of Twitter, I went through your Twitter before this, so I'm gonna read back another tweet at you. The good thing about Twitter is there's a public record of all your predictions.

24:26 I need to turn on that auto tweet deleting feature or whatever it is.

Vibe coding games requires proper scaffolding

24:29 Um, last October you tweeted, "Everyone is going to be able to vibe code video games by the end of 2025."

24:36 Yeah.

24:36 Did that end up being true?

24:38 It feels close and I think there's I mean it obviously not AAA games like you're not building, you know, the next Call of Duty or GTA yet. Um, but I think it's it feels closer than it's ever been. Um, and I think a lot actually a lot of the interesting bit about video games is you actually need to end up building a lot of this like other stuff like models. And we were talking off camera before this like 3JS is a great example of this. Like 3JS makes a lot of things possible that weren't before, but there's still all these like rough edges that like just a coding agent doesn't solve. And so you need like, you know, sprite generation and like the models aren't very good at doing that natively.

25:17 And so you need like some orchestration layer and tooling in order to make that happen. There's a bunch of other things like that that like are core to like the gaming video game experience um that need to have a high degree of reliability that I think it feels like it's within reach but actually like requires a lot of like product scaffolding work in order to create experiences that are like reusable and replayable and sort of like have the level of depth and requires a little bit of taste in there. Um

25:45 do you see people making a lot of video games inside AI Studio and the other developer surfaces that you have?

25:50 Yeah. And so this was actually based on like us looking at the early data and there was something like in AI Studio at the time it was like 20% of all apps that folks were making were actually games like people trying to build games. A lot of

26:01 Is that the most popular category?

26:02 It's not the most popular category anymore. Um just because I think like the the ecosystem has shifted and like the user base have sh has shifted but there's a lot of a lot of games. Um

26:12 what is the most popular category?

26:13 I think it was like it's like 20% like finance related stuff. 20%

26:17 Wow. People like counting their money that much. people like I think it's mo it's something around crypto actually I think is what people are doing a lot of stuff with uh with finance a lot of like personal productivity things and a lot of gen media stuff actually because obviously the Google suite of gen media stuff is amazing yeah has done a great job um

26:36 but I also think GDM has sort of like a obviously Demis cares a ton about games and sort of like started his career in doing AI stuff because of games um and so I think we'll we'll have some interesting swings at this and Um our team actually in in Kaggle which is sort of bunch of the AI benchmarking stuff we do in GDM sort of works with GDM to build this uh game arena which is uh sort of our way of sort of like testing progress towards AGI like using games as a proxy which again is like very deeply rooted in in GDM's uh history. So

27:07 how close do you think we are to you know rando off the street with a good idea can vibe code a really fun playable game? I want to say this year I actually I think it's I think the model capability makes it possible. I think this is where like I've gotten excited on the product side and you know we were again we were also talking off camerara about sort of like the startups in this ecosystem because um it feels like it's possible. It doesn't feel like there's a gap in model quality. It feels like there's a gap in like you someone who knows what it takes to build a great game actually like putting the scaffolding together in the right way to make that possible. I think there are folks who are doing this right now. And so, um, some of it is like a discoverability and awareness thing that like people just don't even know that they can do that. Um, and some of it is just like maybe certain categories of model capabilities are just like slightly off and we're like, you know, weeks or months away from like that chasm being crossed and then it just like working for most people. And so this is a good segue into I want to ask you about role models next. But do you think vibe coded video games is more likely um going to be you know game

28:11 engine plus coding agents based or do you think it's more likely to be world model based?

5 Highlight · Why world models need proper scaffolding

28:17 Yeah, I think the what will end up happening is the definition of world models will blur which we should which we should talk about with Omni. Um, and it will still I think the like coding agent will look like some sort of world model type system. Um, but you actually do need to make world models useful for like real things. You need like scaffolding. Um, and so I think there's again there's actually a bunch of interesting startups like doing work like figuring out what is the scaffolding for world models so that you can take them from these like very open-ended inherent design of world models, very open-ended spaces and like do it in a tangible way so that it's like grounded in a use case that like you could use in a reoccurring way. That could be somebody maybe will figure out the scaffolding for world models to make games possible. But like the inherent nature of world models right now I think make it so that it's like actually not well suited for like games in the current form. But the progress has been

29:17 crazy. So who knows maybe in like two years the the versions will be able to. But at least in the short term it's like coding agent plus some sort of game engine I think is like where you'll see way more alpha from a a game's perspective.

29:28 That makes sense. Okay. So you said the definitions of world models are blurry.

Omni models merge modalities into one system

29:32 Can we unpack that?

29:33 Yeah. I mean I think like Omni is a is an example of this. you know, we launched this at IO. You can sort of take in any input, create any output. Um, and I think Demis sort of like framed it to the world, uh, rightfully so, as as a world model because of just like the level of understanding that it has of the world. I think that like technically looks different than, um, and I'm I'm not an architecture expert on like the way that they we've done world models before, but it is different from an architectural standpoint than what's happened in the past. Um, which I think is positive because it's getting closer to like some of the ways in which it might actually be more scalable. Um, and historically like it's been like super not scalable. It's like very very expensive to run

30:15 traditional like online world models.

30:17 Yeah. Like Genie being Yeah. Okay. So if you think of traditional world models as being like an action conditioned video model almost

30:25 then like

30:26 right now what we're when we when we say world model what we actually mean is a model that has some understanding of the world as opposed to being strictly technically a action condition video model.

30:36 Yeah. And and so the interesting thing though is like it has understanding of the world but then it also has that like really great and that's where like the line is blurry to me where it's like it can do a lot of those same use case. It's not real time right now, but like it can do a lot of those same use cases that you would describe or like visually could create with that same exact world model, which I think is what's most interesting to me. So, I do feel like this like world model video model thing is going to is going to change and play out in a in a different way than was obvious before.

31:10 And how does it work under the hood? Like whatever you're able to share like is it Gemini plus video models? Is it something different entirely? It is it is a single model. Uh which I think is the important part like this was actually part of the original desire was like you were training like eight different models to do all of those things. Historically it's like you have a text model with the baseline Gemini model. You have audio. You have music models with LIA. You have nano banana.

31:37 You have vo video models. You have a we have a whole suite of audio models. And like it would be great for us, our customers, um, if you just had a single model to do all those things. So it is like a a new setup that sort of makes that possible. Um, it's not like routing to a bunch of different models, which like we you could have imagined. We could have done something like that actually before and done like a Gemini omni model, but this is like a true omni model. Um, and it's starting with like the the use case that works the best right now, which is the why it's the one that's available. um is this like video editing capability. Um the technically it's like functional with the other things. It's just like the quality isn't isn't like perfect um and is not state-of-the-art. So we we haven't rolled that out yet. Um it's also just like the first crank of the model turn on Omni. It's the Omni Flash model, the first iteration. Um and so we'll have like much much more capable uh powerful versions which will be which will be exciting to see.

32:35 So we could edit this set so it looks like we're Yes. Yeah. Yeah. I I want this. We again we were talking off camera like we should do that for the intro because I think it just like makes all this stuff more capable and I've seen these examples of like

32:48 such subtle nuance that like make me appreciate that it's like the world understanding playing out. I was I was giving a talk um

32:56 and was on stage with with my friend Tulsi who leads the model team who I don't know if you've ever had on before but she's amazing. I love Tulsi. Um, and in I had mentioned to someone in the crowd to like edit the video and they had literally like took the picture, edit it with Omni in real time. And this like dog came on the stage. Uh, and like the other in the edited version, the other guests sort of like look down, see the dog, they like chuckle a little bit. This is while I'm like opining about whatever AI nonsense about your jokes.

33:24 Yeah, that it was not my jokes. They they laugh at the dog coming up. It jumps onto my lap. I sort of like acknowledge the dog. I keep talking. I'm like petting it or whatever. and just like there's like so much subtle subtlety in getting that right and the model crushed it and it's just it it very interesting and like still trying to like absorb and digest like what that means for you know the way we make content and all these other things.

33:49 That's so interesting.

33:50 Yeah.

33:50 I'm I'm the biggest bull on generative media and what it means and

33:55 I mean one of the things we've thought about for our podcast is the visuals matter as much as the content for sure. Um, that's how you catch people's attention in the first place, right? And and so, okay, I'm excited to I'm excited to play with Omni.

34:06 I'm excited, too. And I think the and and I think you probably feel this way as somebody who makes content, but I've historically like been like very for myself personally, like I don't use AI to make any content that I produce. Like, it's all my words. It's always my voice. It's always my image and picture showing up. Like, I just I I feel like there's just like so much alpha and authenticity. And so like I would much rather it be me than some AI version of me. What I like so much about Omni is that it's like not changing me. Um it is like changing a bunch of these other bits which are not me. Like I didn't choose any of the like set around us or the the coffee table. It's like so our our words can stay the same and like you can change these bits that are like not personal um and do something more interesting with them which I think is really really cool and feels it feels like the version of what I want sort of like Gen Media to be which is like not a bunch of like AI avatars. Uh it's like

35:02 no Fruit Island videos.

35:03 Exactly. Truly like it really is like it's the original content. It's the person. It's like the personhood is there. It's just

35:10 different and amplified. Super interesting. Okay. I'm excited to play with it.

35:14 Yeah, we should we should send some

AI Studio unlocks rapid application prototyping

35:16 prompts right after this and trust.

35:18 I don't mind the fruit videos though. I'm I'm I'm happy for a world of of both. Um on the coding side, you launched the ability in AI Studio for people to vibe code Android apps.

35:27 Yeah. Yeah.

35:28 Um I'd love to, you know, hear how that's going so far and and where you plan to take that.

35:33 Yeah, it's super exciting. I think one of the strategic things for AI studio and actually this is based on like a lot of the feedback from the ecosystem and actually from developers from others is like so many Google products um there's so many different like ways in which you like touch Google through all these different journeys of building a startup or bringing idea to your life and so um we have this like first class principle of like how do we bring things into AI studio that make it so that you are exposed to other parts of the Google ecosystem without having to like go through nine different UIs across Google. Um, and so Androids are like a great example um, not only of that, but also of enabling people who wouldn't have otherwise built an Android app. And so I literally built my first Android app in AI Studio. Um, very cool to see.

36:19 It's uh,

36:20 what is it?

36:21 Yeah, I I just did like a not a crypto app, just a a plant one. I was planting trees in my backyard. Oh,

36:26 like a gardening app. That's good.

36:27 Yeah. And so it was just like playing around with a a gardening app as I as I was kicking the tires. Um, I haven't had my like breakthrough idea yet of what I want for a mobile app, but I'm gonna I'm gonna come up with something and see go compete on the app store.

36:40 Have you seen anything vibe coded like really fly in the app store yet?

36:43 That's a good qu. It actually be interesting to like see some analysis. I don't know. I'm sure it's like accelerating a lot of things on the app store, but I don't know how much like I don't know anyone like personally who's who's done that. It is interesting and I was going to make the observation too that I think the last time I checked the numbers we were reviewing it this morning it was like 350,000 Android apps built in AI Studio since last week which is crazy. Um and like excitingly it's like 350,000 apps that like probably no one was going to build before. A lot of these are personal too.

37:14 And so this is where I think this like maybe Genui is like farther out there, but I think like the idea of you building software to solve your personal problem is like very real right now. And like people are doing that. It's like one of the most common use cases of a lot of these products. Um, and being able to like unlock a bunch of the native capabilities of the phone, I think is also really interesting because you just have so much context that's like in different places. Um, so I'm I'm getting very excited about sort of that opportunity and uh Android feels like it's becoming the the platform for builders.

37:48 Does it matter that something is an app versus just like the web is so powerful now? The

37:52 Yeah, that is it's also very interesting to see that play out. Web is definitely powerful. There are certain things that the operating systems have that like you just can't unlock. um like lots of like native richness that actually like make experiences feel so much richer. I think about this for like text messaging actually that like the text messaging experience in all of the all the main operating systems feel way richer to me than like any AI chat app that I've ever used. Like if I could just talk to AI and whatever texting app I use, like I would be way happier than having to go to some other app. Um because I think

6 Highlight · The model eats the harness and scaffolding

38:28 we're also just like conditioned on like the operating systems. So

38:31 yeah, makes sense. Okay, I want to ask about the model eats the harness or the model eats the scaffolding.

38:37 What are your thoughts?

38:38 Yeah, I think it's true. And I think part of this is like what we have historically thought of as the model is not the model anymore. Like when I think like two years ago when LLMs were popular, it was like the model was like actually just a set of weights. um it was a set of weights and it was like really like how can you like as simple as possible send tokens in and get tokens out and I think we've just like progressively step by step by step we still call it the model we still call it you know Gemini 3.5 you still call it GPT whatever and and claude whatever but like it's actually not just the weights anymore it's like an entire exp expanding sprawling system that's built around the weights um that sort of like enable a lot of these like next generation experiences from agentic tool calling to tool you know like all these hosted tools search code execution etc.

39:28 Um you know the models are now being spun up in containers and sort of have an agent harness and all that stuff. So the scaffolding is like oftent times a couple of steps ahead of like where the act what is like baked directly into the model and then what ends up happening is like the model eats that scaffolding and it becomes part of like the native model system and there's still value in having sort of the external scaffolding in certain cases and like search maybe is an example of this like there's lots of folks who use different search providers and there's different like use cases that you want and so like sure maybe the model can natively use search but you also want something code execution, another example of that.

40:06 Um, but it does feel like like maybe the agent harness is like the quintessential example of this right now where like everyone's like ah we got to go build a harness and like the harness is where the alpha is and like I think that perhaps won't be true at least in the way that we think of the harness today in 12 months. I think the models will have sort of just like digested a bunch of that. It'll be upstreamed into the model um and the alpha will be somewhere else now. it won't be in sort of trying to spin your own harness because the model just like does it natively

40:35 but I thought that the part of the reason why people are building their own

40:38 harnesses is because if you use a harness from any given model provider you're locked in right so a lot of the application companies want flexibility which is why they're building their own harnesses

40:45 yeah and I think that's part of the scaffolding story is like that starts out perhaps true but then as the model capability improves like it it becomes less true over time actually I think the model the like you you don't have a generalized model if it can't use another harness. And so it is it is important and I I mentioned this uh in another conversation with someone a few weeks ago, but we need something like harness bench, which is like actually measuring like how good are all these different models at adapting to all the different harnesses. I feel like that seems like a reasonable thing we should we should measure as an ecosystem. Um and I'd be curious to see like what models are actually best, but I think over time you expect they they'd be able to use every harness. um unless you're like completely out of distribution which in that case like you're still going to be completely out of distribution even if you're using your

Startups win through focus despite advancing models

41:33 own harness. So not sure it matters much.

41:36 Fair enough. What about the application layer? How do you think about where independent companies uh can you know have a hope of surviving when the model eats the harness and eats you know the stuff around it.

41:48 Yeah. It feels like there's so yeah it's it's an interesting story that like both of these things feel true. both on one hand I everywhere I look I'm like there's never been more opportunity to go and build something at the same time obviously the models are doing more than they've ever done before um I think there's like you know there's that thread of capability overhang which I think there's a huge amount of alpha in there's the thread of the model companies are like going after these like very general problems um and there's just like so much value in these like verticalized domains if you have expertise in that domain you sort of like know the customers you know the ecosystem like it's you can really like run laps around even the best model labs because like focus is the like superpower of startups. Like if you can focus you can do anything. Um and if you look at all of the companies that are big or doing lots of stuff like

42:37 there's just not a lot of focus. And for some for some reasons like rightfully so because you know maybe I'm I'm overly justifying you know Google strategy but like we just have a lot of products. We have a lot of users. We have a lot of different things going on. And so like we actually can't focus in one domain. we have an obligation to do a bunch of things as a big company. Um I think that's not true for startups. And so I think like 24 months ago we were all asking ourselves like wow it seems like it seems like the opportunity space is shifting. Um and maybe it's it's possible one of the outcomes is there's less opportunity for startups in the future. That feels like so far in a way not what has ended up playing out which is really positive. If anything, it feels like there's just even more opportunity than there was. Like now coding has helped you like close the gap on like larger companies that have like established code bases and all this other stuff because you can just like run way faster and write software quicker. Um the agentic like primitive is like a new category that you can sort of build products around that like actually in a lot of cases to the conversation about like the risks

43:39 involved with building like there's risk involved and so like what's your like the risk appetite of different companies is different and so if you're willing to

DeepMind combines scientific mission with applied scale

43:46 take more risk in some domains like you can win a user cohort who's like interested in also taking risk. There's so much opportunity.

43:53 Awesome. I'd love to talk about Google Deep Minds culture and I'm curious what does it feel like to be inside GDM right now? You know, we had Demis at AI sense. He was so inspiring. I've heard Sergey's back. You guys have Nom Shazir back. Like, walk me through what it's like to be at GDM right now.

44:12 It's incredible. I do try to take it all in because it is like a it's like a moment to I try to reflect as much as possible in in the chaos of all the things that are happening just because there's like so much cool stuff going

7 Highlight · Navigating multi-domain focus inside Google DeepMind

44:24 on. Um GDM's culture is interesting and like maybe three observations. one back to this thread of like focus. We're doing a lot of things and so I think you see sort of uh I think about this a lot like from a portfolio perspective I think we have like one of the strongest portfolios which is really exciting but you do see these moments where like another lab or another company whatever it is we'll like pull ahead in a certain area where like we've underinvested just like hadn't been focused enough in that domain. Um, and it's cool to see like the the way we go about trying to like close that gap. I very much I very much appreciate it. Um, I think I've I've watched the Demis thinking game documentary a few times. Um, and like you see sort of like a lot of like details of that like original culture and just like the way that strikes work and all this stuff which is actually really similar today is like you just get a bunch of smart people together and like go solve the problem. Um, and I I love that and it's like very cool um to

45:25 to be a part of. Another one is this I think you see the culture permeate from like who the leaders are. Um, and as I maybe uh this isn't like a a perfect characterization of the ecosystem, but

45:39 like Demis is a Nobel Prize scientist um and like the sort of OG of a lot of this stuff and sort of you feel that in the deep mind culture. I think like Sam is like the, you know, maybe one of the world's best businessmen ever and like you sort of see that in the open AI culture and the way that they go about the world. I don't have a strong sense of who Daario is. Um, but like I think Anthropic is a very interesting place and you sort of at least as an external observer like they're he seems like an interesting guy and so uh somewhat esoteric and so that seems like they're sort of like that in the DNA and the culture of the company. You know, the other labs are interesting. Um, but I like this like very scientific approach to the world. Um, and the way that like Demis looks at like the reason he's doing this and the reason they started this mission was like literally to like solve disease and all these things. And it's like so easy to get and again I'm always trying to pull myself out of the moment, but like it's so easy to get lost in this like competitive race of who's pushing a number higher on SWE or

46:41 whatever it is. It's very easy to lose sight lose sight of like the reason we're doing that is so that like we're can solve problems that humans actually have. And there's a uh my my favorite quote from all of Silicon Valley is something like you know we can't let other people make the world a better place more than we can which is like what this moment feels like. And I

47:00 Gavin Bellson quote

47:01 the Gavin Bellson quote and I think about that all the time and it's like we're all fighting over who can make the world better more than the other person which just like when you frame it like that seems really goofy to me. Um, and so it's very much not zero sum. Um, and I think that's like a a way of looking at the world. I think the last thing about Deep Mind's culture is like we're very it's sort of the engine room of Google. Uh, which I think is like literally the Twitter bio now of the the Deep Mind Twitter account, which I love.

47:29 Um,

47:30 you man the Deep Mind Twitter?

47:31 I don't I I don't want any responsibility manning other people's accounts online. Too much uh too much responsibility to do that. But it does feel like that too. So it's like on one hand you have sort of like the deeprooted lab culture and the other hand you have sort of like all of these partners across the Google ecosystem that we're collaborating with everybody from Android that we talked about earlier to Google cloud to you know Gmail to workspace etc etc and so it's an interesting blend of like I think there's lots of research work happening but like there's tons of applied work that's happening to like actually like work with some of like the forefront customers like deploying Gemini to billion user products is a problem that like only two companies in the world have. Um, and we have 13 of those products and like we that you know Google goes through this all the time now. And it's such an interesting place to like see that happen and see the innovation that takes place in order to make that actually possible. And I feel like it's you can only do that inside of inside of Google which is really cool.

48:30 Beautifully said. Did they uh did it give them a lot of heartburn when you joined and were tweeting a lot?

48:36 That's a good question. to get sign off from from coms.

48:40 I'm very uh one of the the silver linings to my Google experience has been just like how great that group of like folks across marketing comms are to to work with. And I think like I you know their job is protect Google, make sure we tell the right story, make sure a bunch of bad things don't happen. And so I have a ton of appreciation and partnership with them. But it's been an incredible experience to like be able to go try to tell the story that resonates with developers um in a way that feels authentic uh and not have a huge amount of you know

49:14 I you know I don't have to get my tweets approved all the time and all this stuff like is very very positive culture and I think hopefully I I'm always trying to walk the line of uh not not burning uh the the trust and goodwill that that I've accumulated with those folks. But it's been super positive because ultimately I think it's like it's really hard for Google to tell this like authentic story. It's just there's just like it's a big company. There's a lot of people. There's a lot of opinions. And so you take the like magic of Google and you water it down through like a lot of people and a lot of process and you actually yeah

49:46 you you miss the beautiful story which is like Google's doing the most interesting technology in the world and like helping our users with some of the hardest problems in the world. And um it feels it's a privilege to like get to help tell that story. So it's it's a lot of fun. I enjoy it.

50:00 I love what you're doing. I love what Josh is doing. I think you guys have put a really kind of sincere human touch on um as you put it the most important problem of our time. So

50:09 thank you.

50:10 Uh well wonderful Logan. Thank you so much for joining me today. This is a very far-ranging conversation. Everything from agents and coding to world models and harnesses and GDM culture and you lots lots of nuggets here. Uh thank you for for joining me today.

50:24 This is a ton of fun. Thank you for having me and I'm excited to see what the uh folks cook up of where we've been sitting this whole time. Maybe in front of us.

50:31 Maybe there'll be a dog.

50:32 A dog.

50:33 Make my dog dreams come true.

50:35 I love it.

50:36 Awesome. Thanks, Logan.

50:37 Of course.