It really is going to be this powerful enabling factor to, like, bring in, you know, the next 100,000,000 people to be able to learn to code. So I I'm, like, very, very bullish on that. I wake up every day, and it's the fight of our life to make sure that we keep making the product better.

I'm joined today by Logan Kilpatrick from Google DeepMind. He shares the tools we'll need to reach 100,000,000 developers, how Google's AI comeback story happened, and why the next six months are gonna be more exciting than ever.

People are finally figuring out how to build AI products that people actually like. We sit inside of DeepMind and we benefit from, like, directly being next to the researchers that train the models. It ends up being, like, a really human problem. All of these things had to happen to like put us in the place where we can be successful. I'm very optimistic for the future.

You've spoken offline that you want to get to a 100,000,000 developers using Google's products.

Yeah. A long way to go to get there.

Yeah. How how the hell do you even think about trying to go after such a kind of ambitious goal?

Yeah. That it's a good question. I think today where we are and and I mean, for what it's worth, it won't be just us who are who are doing this. I think, like, this is an ecosystem wide thing that's going to happen, but I think there's, like depending on who you ask and which which charts and data you're looking at, I think there's something like 30,000,000 professional developers today who are actually writing code. And I think there's this you know, I think that the Twitter narrative or the X narrative is perhaps, like, split in two.

There's, like, half half of people, and I don't know if this is actually true, but from a percentage perspective, but, like, it seems like half of the world thinks you shouldn't learn to code anymore. Just, like, go prompt and do all this stuff. And the other half is, like, you still have to learn to code, etcetera, etcetera. Things will change, but you should still learn. And I'm in the you should still learn to code camp.

Not that I yeah. I I think that will things will change for sure, but vibe coding and AI and everything that's happening feels like it's this perfect storm of actually taking a lot of the drudgery out of building software and and specifically actually learning to code and building software. My mental model has always been, and this is, like, my firsthand lived experience, so I don't know if this tracks across other people. But part of the challenge of learning to code is just, like, the really interesting things to be built are on the, like, tail end of the coding experience. Like, when you first learn like, I started to learn to code in c plus plus and I was like, this is as boring as could possibly be.

Like, this is not interesting at all. Like, partially just because of, like, what what we were building in those classes was not cool. I think if I was, like, c plus plus programming, like, you know, games in Unity or something like that, which is probably really, really difficult, it might have been cooler. But, you know, the entry level stuff is just not fun. And I think vibe coding and all this AI accelerated coding is just, like, such an interesting way to help, like, bring the excitement of building stuff.

Like, get to that magic moment with coding way faster than you would have traditionally been able to do and get people through the, like, things that are hard about coding much faster than they could have before. Like, I remember when I was learning to code, I I started undergrad in community college, and I was I was I would go to the professors, and you would have to, like, literally find a professor or you would have to find a tutor or you would have to, like, go online and, like, you know, maybe get lucky that someone else had this problem before and asked it on Stack Overflow and, you know, has an answer that you could actually understand. Or you'd maybe take the stab of, like, asking a question on Stack Overflow, and then everyone would yell at you for asking your question in the wrong way or whatever it was. And I lived through that many times. And and now you just have, like, a personal tutor through AI that can just, like, help you on demand twenty four hours a day sort of reformulate whatever the problem is.

I think that's just so cool and, like, I think it it really is going to be this powerful enabling factor to, like, bring in, you know, the next 100,000,000 people to be able to learn to code. So I I'm, like, very, very bullish on that. And I think we'll we'll do our best as well through the Google Developer products to, like, help play a part in that.

Yeah. And I guess like if you're reaching these people that, you know, I guess what we're talking about is, yeah yeah, the vibe coders and it's definitely a different segment. How are you kind of trying to reach these people as well as, you know, called classic classic developers.

Yeah. It's interesting. It is the it's a good it's a question and something that our team has been thinking a lot about because our our core product today, AI Studio and the API, have historically been very developed, like, traditional developer centric. And now we're building a vibe coding product inside of, like, many people are doing inside of AI Studio specifically to help try to, like, bring in that next group of people to get them building products and stuff like that. I think the missing link of the experience right now is that most vibe coding products don't they they don't do the thing that is actually useful to the learner or to the builder through this outcome of, like, helping them learn.

Like, so many if you look at, like, what is the outcome that all of the current vibe coding era products are trying to get you to do, the outcome of what they want to do is, like, build software for you, ideally help you deploy that software, and then hopefully, like, monetize and, like, find a way to make money off of doing that process and, like, having that flow. And I think that's great. And, like, I think all of those are required attributes of what a vibe coding product or what a software development product could be. But they miss the important factor, which is how can you actually bring people along with you as they're building these complex systems and writing all this code? And how can you do that in, like, a really incremental way?

And and, you know, here's my personal experience. Like, I I'm a I use vZero all the time. I love vZero. I think it's a great product. You know?

So shout out to all the folks at Vercel and and the vZero team. But a good example of this is when I'm using vZero, there's, like, code flowing across the screen. And I'm like, it's entirely useless. It's not helpful to me at all. Like, even as a developer, as someone who studied computer science, as someone who's, you know there's an ambulance coming by.

As somebody who's who's, you know, deployed software into ChatGPT and, you know, at Apple and now at Google and all these other things, like, is just not useful to see code flying across the screen. And there's so many parts of this, like, software development process, which I think or the, like, vibe coding software creation process, which I think could be optimized to, like, actually be part of that story of how we're onboarding these users to learn how to to learn how to code and understand what's happening. And it's just like a it's a problem space of, like, how do you show the right level of detail and hide the right hide the wrong things and all this other stuff. It's like it is very nuanced, but I'm super interested in this problem space. I think there's there's so much to be done here.

And, actually, like, there's a a similar thing for developer. Like, even, like, traditional developers will really benefit from this, which I'm excited about. So it it it'll be very interesting to see how the next the next few months goes.

You're making me think of like, if you just get billions and billions of logs and like, it means nothing to you. It's just like basically the same thing. But like, you wanna be able to just surface like the right amount of abstraction or like, rather than just giving them everything when they don't understand like these, you know, thousands of lines of code that have just been generated. But they may need to understand some specific part of it so that they can debug what went wrong. But knowing that they may not even really know how to code sometimes and yeah.

Yeah. There's there's something here. And and I I've spent a bunch of time with Amar, who's on my team, who who leads design and product, chatting about this. The other there there's some really interesting magical experiences to be unlocked, so I'm I'm excited for us to to share more about this.

Scaling DevTools is sponsored by WorkOS. If things start going well, some of your customers are gonna start asking for enterprise features. Things like audit trails, SSO, SCIM provisioning, role based access control. These things are hard to build, and you could get stuck spending all your time doing that instead of actually making a great dev tool. That's why WorkOS exists.

They help you with all of those enterprise features, and they're trusted by OpenAI, Vercel, and Perplexity. And if you use them for user management, you get your first million, yes, million, monthly active users for free. I honestly don't know any dev tools that have a million monthly active users apart from GitHub maybe. So that'll get you pretty far. Here's what Kyle from Depot has to say about WorkOS.

We use WorkOS to effectively add all of the SSO and SCIM to Depot. It's single handedly like one of the best developer experiences I've ever seen for what is, like, a super painful problem if you were to go and try to roll that yourself. So for us, we can effectively offer SSO and SCIM, and it's, like, two clicks of a button, and we don't ever have to think about it. It's, like, one of the best features that we can add to Depot. It's super affordable, which effectively allows us to like break the SSO tax joke.

Essentially say like you can have SSO and SCIM as like an add on onto your monthly plan. Like, it's no problem. So it really allows smaller startups to essentially offer like that enterprise feature without a huge engineering investment behind it. Like, it's literally we can just use a tool behind the scenes and our life is exponentially easier.

You recently launched Gemini CLI, which I think is, you know, probably like people at the moment maybe know Claude code a bit more at the moment. But it's like it it's like kind of a there's someone at Google's just like, I mean, I've mentioned Claude Cove. And the

the lightning strike happens.

Yeah. And also, I know there's like the windsurf. A lot of windsurf team just joined just joined Google. How are you thinking about those kind of tools as well?

Yeah. It it's super interesting. And like, we we've also been spending a bunch of time thinking like, what is the what what are the core problems that you're trying to solve with, like, a vibe coding product, for example, which is different than a developer IDE, which is different than, like, a CLI. And it it actually feels like there's a different user persona for all of them. It it has been just, like, to reflect on the success of some of these, like, CLI tools.

It has been really interesting to see how successful they are. My 2¢ as a developer, and I think maybe this is, like, maybe not like a generational thing, but it just depends on, like, how you started to code. I was never, like, a VIM or e Emacs person or whatever. Like, I use the terminal when I have to to run commands, like, out of functional use, but, like, personally find the terminal interfaces to be, like, a very bad way in order for me to personally develop code. So I've never been, like, personally, like, excited about any of those of those CLI tools.

It just doesn't feel natural to me. It feels like I have way more creative power in an IDE than I would in a in a CLI. And and, obviously, different people feel differently about this, so it feels like it's almost not even worth yeah. I I've I've kind of just written off that we'll have both of them because some people like them and some people don't. But, yeah, it is really interesting that so we have Gemini CLI, which is crushing it and doing really, really well.

A bunch of folks from the Windserve team are now part of DeepMind, is exciting. So it'll be cool to see what they what they come up with and and and cook up. And I think also my other reflection was just around, like, how much like, even, like, two or three years ago, it felt like, you know, GitHub Copilot had already just, like, won the market. There's no reason to do anything else in AI developer tools. You know, they the market basically didn't exist, or it had totally been won by GitHub.

And then you sort of fast forward to, like, maybe, like, eight, nine months ago, and I think the narrative was like, oh, Cursor's won the whole market, and then there's no point in doing anything else. And then now you have all this activity with Windsurf and Cognition, and and Cursor is still crushing it. And there's, like, tons of, like, AMP code from Sourcegraph, and there's, like, 50 other tools, Klein and, you know, etcetera, etcetera. There's, like, all of these tools that are doing really, really well. And I think it's actually solidifying in my mind just how big the developer market actually is and how different the needs of developers are.

Like, developers are very particular about the tools they use. I think there's a huge market for for a lot of these companies to go and build, like, really useful, successful products. So as as someone who loves building developer tools, it's cool to see DeepMind getting as as excited as I am about solving these problems for developers.

Yeah. I was I was just thinking, Logan, because I think I I don't think I met you at the time, but I think you were you were at the first AI engineers conference. Right? You gave a Yeah. Presentation.

And I think that that was also I don't know if you were there, but that was the copilot presentation. And it was like I don't know if they were supposed to, but they they kind of talked about how they got to like 40,000,000 revenue in like six months or something crazy. And it kind of felt like it was like a almost like a lap of honor and like it was like good good game everyone else. Like this is the best, you know, this is the greatest business ever built and like they've won. And yeah, it is kinda crazy now when you think of how many things there are.

So

yeah. And also Copilot's doing well too, which is which which is what's remarkable to me. Like Copilot is doing well at the same time that like Versus Code has built a bunch of AI features into it natively at the same time that, like, lots of people are using Cursor and and other tools. So I I think it's, like, generally, like, a really, really cool moment for the ecosystem to see that, like, level of progress. And, And like, obviously, all of us as developers are, like, the beneficiaries of it.

So I'm like, I love it. Like, the competition is great. Hopefully, it keeps being you know, there's lots of stuff being built because it means I get better tools as a as a developer, which is awesome.

Yeah. And and actually, kind of speaking of Copilot, I know, like, I know a lot of people that use Copilot because it's the only mandated option at their company. And I know, like, once you get into like the kind of bigger companies, they're much less like open typically to like trying out the new tools like they may not have cursor even. How much of that like do you think about with like Google obviously has its own like suites of like products and stuff. You already have probably you're probably sell sold into, like, 99% of companies already.

Like, how do you think about, like, leveraging that?

Yeah. It's an interesting question. I think some of this is, like, what what do people come to Google for? And I think, obviously, Google Cloud is a huge business, and, like, we we we do sell sort of enterprise infrastructure to lots of people. I do think, like or and, yeah, there's, like, tons of very successful products.

Like, BigQuery is a great example of this, like, very ubiquitously used across the ecosystem. Lots of lots of companies and developers are using it. It does feel like this AI moment is where is is, like, so competitive that everyone it's like reset all of the priors for everything that's happening. And I don't know if you're laughing at the lightning background or the thunder.

Oh, no. I was well laughing just that like, yeah. It's I I felt like yeah. It was very humble. I'm sure it's still I'm sure people still probably feel a little bit safer buying from Google than from a new startup.

No. That that's true. But it is it is, there's definitely a customer segment where that's where that's the case. But it's actually I'm I'm glad I got your reaction to that because, like, I think yeah. There's there's definitely customers where that's the case.

But I think more and more, it's become like AI is this frontier problem where, everyone is, like, kind of willing to or it seems like the the mentality has shifted over the last eighteen months to be like, everyone is, like, kind of willing to pay a little bit of a tax to and and, like, take on a little bit more risk of of being on the frontier at the cost of, like, sort of disrupting how maybe they procured software in the past or for example. So I think it's, like, disrupted a lot of the conventional wisdom about, like, you know, just because you're Google and you have lots of distribution doesn't mean that you just get to, like, win whatever the market is. Like, it is super it's a super competitive market across, like, all dimensions of AI from the models to the products to the developer tooling, etcetera. I'm reminded of this all the time. Like, we and I when I talk to our team, like, we need to win across every dimension of of the product experience from how we engage users and support to, you know, how we're showing up in the ecosystem to, like, what are the error messages that are being returned in the API to Mhmm.

You know, how are we making it easy to adopt Gemini models when, you know, people are are testing different options in the market. So all of these are dimensions you have to win across. And if you're not laser focused on that, like, it's very easy to, like, get lost in how much stuff is happening in the ecosystem. But I don't wake up any day of the week thinking, like, oh, it's you know, we just get to sit back and and sell Gemini to everyone because we're Google. Like, it is I wake up every day, and it's the it's the fight of our life to make sure that we keep I mean, not from, like, like, a purely competitive standpoint, but just, like, to make sure that we keep making the product better.

And, like, that's the north star. And, like, at the end of the day, that will translate into all the things that we care about. But I had an interesting conversation where there's, like, so much so much of what people are doing in AI right now is, like, reactionary to what other people are doing. And that adds this, like, added level of, like, FOMO difficulty where, like, you need to, like, oh, our competitor shipped this thing. You know, we have our own road map of what we think me makes us successful, but, like, we also need to do what they did to make sure that we don't lose the customers who care about that specific thing.

And it is this, like, really difficult virtuous cycle to break. Yeah. It it makes it tough.

You gotta do all the things all the time and

And you can't. And that's the problem. And then they're you're, you know, fundamentally, is a it is a difficult problem to solve.

Yeah. So how are you doing things at Google? I think last time I interviewed you, you were at OpenAI. I think when you left OpenAI to join Google, I think there was like people were surprised. I think it's fair to say that there was a lot of surprise because OpenAI was like absolute I mean, still is absolutely crushing it.

But I don't think there was that feeling about Google and AI and a lot of people were kind of, you know, I don't know how to say that negative, but feeling like Google wasn't taking advantage of the the advantages that they had. And if we fast forward to it's like today, so I I can say like I use Gemini 2.5 Pro a lot. It's the only one that I use that is able to like process like, for instance, like show notes and stuff because of the context is amazing for like whole transcripts of an hour long. And I'll use it for this episode. For Flash is like amazing for like the fast like two point o Flash for like real fast like processing.

It's like absolutely brilliant for like voice stuff. You know, v o three is just like unbelievable. It's the best video model. It's like Google has it seems really stepped things up over the last like eighteen months, which I don't know if it's just coincidence that that's when you were there. But what what do you think has, like, changed, if anything has changed?

That is a great question. There's definitely a a a large degree of coincidence. I am hopeful that I've that I've helped in in some way and and helped push the rock up the hill, but I I think there's a couple dimensions to this. One of them is and I think this is kind of like an under underappreciated reality for Google and maybe just for big companies in general, which is, like, organizationally set up to to be put in the place to be successful is extremely important. And I think if you look at just, like, how Google was oriented pre LLM era, like, Google was doing a lot of things and was, like, largely very successful at most of the things that it was doing across search, across, you know, Google Docs, making, you know, a bunch of long term speculative bets, which have already started to pay off with things like Waymo and others.

And, you know, YouTube is now, like, one of the fastest growing businesses in the world, and Google Cloud's the eighth largest enterprise or, I think, now the sixth largest enterprise business in the world. So, like, for all intents and purposes, has done really great work. And and and further, actually, like, AI has been a piece that has underlied a lot of that. So, like, since, you know, Sundar took over as as CEO, like, Google's been oriented as an AI company. And, like, Transformers as a good example have been, like, powering the Google search experience.

And, like, now today, Gemini is deployed to, like, over 2,000,000 2,000,000,000 users across Google search, which is just, like, crazy to think about. It's the most widely distributed model and the most widely used AI product in the world. And it's actually principally, which I think is really important to think about, and there's obviously lots of, like, critics of the Google search AI experience, but it's oftentimes the first entry point that, like, hundreds of millions of people have to, like, what this, like, generative AI product experience can be like. Like, despite ChatGPT and other products being, like, very successful, there's still, you know, hundreds of millions of people have never used those products. And so search acts as, like, an entry way to sort of get folks familiar.

But I do think there's things that, like, Google was was, like, you know, slightly behind the curve on. And I think one of those was, like, we just organizationally weren't we had, like, many different teams doing a lot of different stuff. And I think this was back in 2023. Google made the decision to, like, bring actually bring those teams together, and it made sense, like, as it as, like, the LLM, you know, area started to materialize in, like, a really real way. It was like, okay.

Well, we have lots of different team doing LLMs. Let's bring them all together and, like, build a single consolidated team to go and work together and push this up the hill. And that was DeepMind. So historically, we had Google Research, which is doing a lot of this. Google Brain was doing a lot of this, which is where the transformers came out of.

DeepMind was, like, had the original mission of creating AGI and had done a ton of, like, the original RL work and other things like that. And I think bringing all of these teams together has actually been what has made it so that we've been able to make this level of progress, from a model perspective. So it's been, yeah, it's been awesome to see. And then, like, to further that that sort of, trend, like, DeepMind has now also gone from doing just foundational research to building the models and deploying them across Google to now actually also, like, taking on ownership of some of those product experiences. So our team from a developer perspective, the Gemini API and AI Studio, we sit inside of DeepMind, and we benefit from, like, directly being next to the researchers that train the models.

And the Gemini app is also part of that part of our organization. So it's like the the universal assistant to, like, help everyone accelerate their day connected to all of your Google stuff, which I think is, like, a really exciting feature for the Gemini app, also sits inside of D MIND. So they benefit from, like, the teams that directly train the model. So I think all of these things had to happen to, like, put us in the place where we can be successful. And it just takes it takes time to, like, get, you know, many thousands of people, thousands of humans on the same page.

Like, it is very much like a not actually a technology problem. It is like a human a human problem at the core, which is like so fascinating to think about in this like age where everything feels like a technology problem. It's like, actually more often than not, it ends up being like a a really human problem.

Yeah. I've heard you say that actually, and I know you're a you're a fan of humans. I

I'm a fan of humans. I think there's just there's there's no there's so many good reasons to bet on all the like, I think there's a lot of existential worry about the future and AI and stuff like that. And I think I think there is, like, some amount of it which is warranted and, like, it's good to be thoughtful about what the future looks like. But, yeah, there's just so many inherent things about humans that, like, humans want to interact with and be around other humans. There's no replacement for that.

And I'm I'm really I'm I'm very optimistic for the future. Yeah. There's just, like, so many of these, like, really incredible leaders across Google who are just, like, very down to earth, awesome people to be around and, like, are having a like, actually having fun, which I think in this AI moment, which I think is just, like, really easy because there's just, like, so much it's almost this, like, exhausting narrative of, like, how competitive everything is. And really, at the end of the day, like, this is it's also supposed to be a little fun. So it's it's important to, like

I I'm always reminding myself of that. I feel like your job sometimes must be, like, quite hard to be fun because that you're you're, like, often like the very like point of contact. You're like kind of almost like if someone's got an issue, they're gonna like tag you at a like kind of like

That stuff

do you

Yeah. That that stuff is is just not that difficult. Like, part of my I I think it's just, like, so innately in who I am now that, like, I and the historical context is, like, I during undergrad, when I started undergrad, I actually started working at the Apple Store. So, like, I spent my I was, like, twenty five hours a week in addition to going school full time, was working at the Apple Store and just, like, helping people solve their problems. And it the Apple Store is such a unique and special place, And you you just would and and I was actually working at Apple Store in California, so it would be, like, this this huge range of people who would come in and have questions.

And I started literally, like, selling people iPhones and eventually became, like, a technical expert in, like, helping troubleshoot people's problems. And on that side of things, you would have people everyone from, like, someone who literally just, like, fundamentally didn't actually understand what their phone was doing. It was like this, like, foreign magic to them, and they were just, like, truly lost in the technology to, like, literally engineers who were, like, on the core teams at Apple building whatever the service was who would come in. And, like, you would need to be able to, like, talk to both of these audiences, which I think was, like, such a profound experience for me to, like just, like, get many, many thousands of iterations on on goal of, like, doing those types of interactions. And then now it's just, like, the reward model in my brain is just, like, trained to actually, to my detriment in some ways because, like, a bet like, a lot of this stuff is not super scalable.

I think, like, acts actually and acts or Twitter, like, does make it kind of scalable. But, like, sometimes it ends up being, like, a bad crutch where, like, no. Actually, like, it's probably not the right thing that I'm spending a bunch of time answering people's problems on on Twitter and x. Like, it's usually the symptom of we have some other problem. And so I do a lot of that reflection as well.

It's like, how do we actually solve the thing that our customers have an issue with and not just, like, Band Aid fix a bunch of problems that people are complaining about. So that that's the, like, challenge that I have now of, like, finding that balance of, like, actually answering the person's question versus, like, how do you take a couple steps back and couple steps higher up of, like, broadly solving that problem space and making it easier for customers.

Yeah. But I think it's kind of wild that you've managed to get to this point doing that.

Yeah. It is. It's fun.

Very impressive. Yeah. Mitchell Mitchell Hashimoto, the Hashikov founder, I think also worked at the Apple store when he was Interesting. Young. It's maybe It's a

great place. Lots of lots of great people.

Yeah. Okay. So finally, the stuff that you want I'm sure you are most excited to talk about. What kind of well, you tweeted like I think a couple weeks ago saying that you think the next six months is gonna be the most exciting in AI so far. Do you wanna give us a bit of a tease of what you think is coming along, coming down the line?

Yeah. It it's been there's so many dimensions to this. I think just like generally, this this will be true, and there was lots of people who are like, oh, you could say that every six months. And, like, I think you're right. I I think everyone is right about that who was who was commenting and replying that.

I think it's just, like, worthwhile to, like, constantly remind yourself because I do think there's, like you know, it feels like it ebbs and it flows, and, like, this isn't it's not, like, this clear it's not super clear all the time across every domain, but, there's a bunch of things getting me excited. One, I think, like, people are finally finding out, like, figuring out how to build AI products that people actually like. And I think if you look at, like, the circa, you know, 2023 era building AI products, like, the reason and there's a couple of, like, maybe notable exceptions to this. Like, the reason no one built really great AI products during that time was because they just, like, didn't know what to do. The models were changing so quickly.

People were still forming mental models of, like, what would a good customer experience around these models look like, etcetera, etcetera. So as a user of these products, like, we're it, like, takes a long time to build a great product, and I'm excited that, like, there'll be more and more cool products in the next six months that I'll get to benefit from. And, like, I think the models are already so good today that you could just you you could probably build the thing you want. It's just like, do you have the right context in order to build it? And, like, yeah, the models will also get a lot better.

I think on the model side, like, there's so much We're we're we we continue to get these, like, new dimensions of scaling, which is exciting. Like, historically, you like, the only the the way the models would scale was pretraining, and then, you know, everyone was asking, like, is the are the scaling laws over? And, like, not now then you could, like, we saw probably, like, a year and a half of, like, post training scaling getting better and, like, lots of stuff happening there. Now now you have all the RL stuff that's scaling up, which is really exciting, and, like, x AI is talking about how they've just, like, invested everything into RL, that RL stage and, like, actually haven't been doing pre training scaling, which is really interesting. And then, actually, there's, like, a new so I think the reasoning piece had been, like, the most recent scaling dimension.

I think now we're actually seeing this, like, actually with tool use as well, which, like, the models are becoming systems themselves and, like, getting access to more and more hosted tools and different, like, tooling tool tool chains, etcetera, etcetera, which is, like, this new dimension of scaling because you can there's oftentimes, like, a lot of headroom from the model side as far as, like, how you train the models to be better at those, like, complex tool interactions. So I think we'll see that as, like, there's, like, unlimited low hanging fruit there, which is really exciting. So, yeah, there there's just all of these really unique things that are happening. And, like, from a building developer, like, building on these things or, like, building developer products, like, there's so much excitement about the developer ecosystem. Like, AI coding tools just continue to work.

Like, you know, there's millions and millions of new people coming in using these five coding products. Like, people are actually able to build businesses around those. Like, you know, Lovable's announcing the fastest path to a 100,000,000 ARR, which is awesome. And Bolt's been super successful and all these other so it's just like there's so much to be optimistic about, and it feels like all that acceleration is, like, incredibly positive and, like, very positive some for the ecosystem, which is just yeah. It's it's really cool to see.

Yeah. And and are there any developer tools that you're really excited about outside of, like, the kind of AI coding tools?

Yeah. That's a good question. I think and maybe this is this is too self serving, but I'm I'm very excited about the live API that we've been working on for a while. I think there's just something, like, really, really unique here where we haven't really sort of cracked the nut on what like, how that product experience is gonna manifest across all these new products that folks are building, but I really do think there's something special here. For folks who haven't tried this out, we released it back in December.

We've continued to make all this progress on, better models, better function calling, but essentially, like, the model can you can talk to the models. The models can see what's on your screen. They can hear what saying, and then they can actually do all this, like, very interesting proactive proactive audio, proactive video, and, like, take action based on what they see based on, like, your system instructions and all these, like, really, really interesting use cases. And some of the, like, early traction that we're seeing with this is around, like, sort of, I I for a lack of a better word, like, copilots and or, like, helpers that, like, help you actually use complex products. So you can imagine, like, as you go into, like, you know, Photoshop as an example.

There's, a thousand little buttons everywhere that, like, I have no idea what these things do. You can literally just click a button, and this this experience doesn't exist in Photoshop. It exists in other products. But you can click a button and just, like, talk to the model and explain what you're trying to do. And the model also has, like, vision access, and it can, like, do things like draw bounding boxes around different buttons or different workflows and, like, show like, actually help you use different stuff.

And I think there's so much there's so many interesting ways to do that, especially for people who are, like, not deeply, technology native. Like, I think about this all the time with, like, you know, helping my grandma use, you know, different websites. And, like, she has a really hard time figuring out, like, how to navigate through all these different websites. And, like, now she can literally just click a button and be like, here's what I'm trying to do. Like, help me accomplish this this task.

And it's such a cool yeah. That's obviously, like, a very basic use case, but there's so many advanced use cases of this as well where I think we'll start to feel this, like, AI co presence as, you know, as the live API gets more adoption. So I'm excited to see what people what people build with it.

That that is actually really cool. Yeah. I I I can see that being really helpful, especially like with all the time you just don't like, you're trying to follow a tutorial and there's stuff that happens outside of that context of like, that that would be amazing.

I I want this for pair programming as well. I think there's something just so magical where like you go back to like, how do we bring in the next 100,000,000 developers? Like, every IDE just has a button, and you click the button, and then you have an on demand twenty four hour a day pair programmer who can see exactly what's on your screen with your permission, and you can talk to the model, and it can it can run commands for you. You can do all these things. Like, that is not some futuristic, like, version of the that is literally the version of the world that we're living in right now.

It's like people just need to build that product experience. So I'm, yeah, I'm I'm super excited. There's there's so many interesting things to do.

That would be amazing. And that's a good way to end it as well, think, Logan, to wrap it up on reaching the next 100,000,000 developers.

Yeah. I'm glad we did I'm I'm glad we got to do this. Hopefully hopefully, it'll be sooner than eighteen months that we get to chat again. But I appreciate you having me back on, and thanks for everything you do, and this was a fun conversation.

Thank you. And where should do do you wanna send anyone to anywhere? Gemini, I guess?

Go to Gemini. Check out AI Studio. You can go to AI dot studio or ai.dev. And, yeah, if you have questions, if you need anything from from our team, please email me, DM me on Twitter, whatever it is. I'm happy to happy to help answer questions or help people get higher rate limits or scale up or whatever it might be.

Hell, yeah. And Logan has a podcast as well.

I do

have a podcast, but

but Jack's podcast is probably you should probably watch this and not my podcast.

Go search it. Thanks, Logan. Thanks, everyone, for listening. I hope you enjoyed that episode. Logan is an incredibly talented guy.

If you enjoyed it, give it a share, give it a rating or just drop us a message and let us know that you enjoyed it. Thanks for listening.