Google Cloud Next '24 | Gabe Monroy | Google Cloud

Clips
News
More from Google Cloud Next '24

Gabe Monroy

VP, Developer Experience

Google

Gemini Code Assist draws developer interest in Google Cloud’s expanding AI role

One of the key takeaways from Google Cloud Next in Las Vegas last week was that developers will have a major say in which companies get to shape the future direction of artificial intelligence.This was noted by SiliconANGLE co-founder and research analyst John Furrier on theCUBE Pod following last year’s Google Cloud Next gathering, and this year’s event served to reinforce developers’ influential role. Among the many announcements during the week was the news that Google’s Duet AI for Developers has now evolved into Gemini Code Assist, which offers enterprise-grade coding support powered by the Gemini 1.5 AI model and context window tokens for processing data

play_circle_outline Unveiling Google Cloud Next: A Spirited Journey of Excitement, Success, and Lighthearted Banter

play_circle_outline Future of connected agents and the acceleration of software development.

play_circle_outline Gen AI applied to productivity and innovation for end customer experiences.

play_circle_outline Desire for customers to accomplish the formerly impossible with Google Gemini.

play_circle_outline Revitalizing and Transforming Legacy Brownfield Environments: Overcoming Challenges to Thrive in the Modern Enterprise

Info

Gabe Monroy | Google Cloud

Gabe Monroy

VP, Developer Experience Google

search

(gentle music)

>> Good afternoon Cloud Community, and welcome back to beautiful Las Vegas, Nevada. We're here in the afternoon of day three of Google Cloud Next. It has been an insanely exciting week. My name is Savannah Peterson, joined by analysts and brilliant humans all around, Dustin, John, thank you so much for being here. Dustin, we haven't gotten to do a segment yet to this show,

>> No, how did that happen? And it's like day three, and this is the first time we're on the desk together.

>> Criminal, criminal. >> I know.

>> We'll have to talk to someone about that.

>> Yeah. >> I don't know who. (panel laughs)

>> That's me. We'll fix that. Have that fixed, we'll take care of it. Make a note of that. AI, where's my AI assistant?

>> We also got the VP of developer experience at Google with us, Gabe, thank you so much for being here.

>> Thank you for having me.

>> You are a total alum. I think you've been on the show how many times, five, six?

>> I lost count.

>> He's up on the leaderboard definitely, he's good.

>> He's definitely on there.

>> On KubeCon, he's up on the top 10 easy.

>> We need the "Saturday Night Live" jacket, you know? The jacket you get, the five timers club, right?

>> I like that, I like that, sign me up.

>> I like that. >> He's a VIP for sure. You won't stand in line for our party. (panel laughs)

>> Exactly. It's been a huge week for Google. How has this been for you? Give us some highlights.

>> It is been really exciting. You know, for me, I always think about all the engineers and all the teams that have been working so hard over, you know, the past few months to really land all the product impact and products at a show like this. And I'd say I'm really impressed with how this is all being received, right? You know, folks have been trying to figure out what does this AI journey mean for me? And I think what you're starting to see on both the main stage keynotes, the developer keynotes, all in these breakout sessions, is we got the goods, we brought it, and I think that's super exciting

>> Google really did bring the goods, ton of announcements this week. Any favorites for you?

>> I'm biased to my own stuff, of course. But I got to say the full code base awareness that we have in Gemini Code Assist, the ability to move from analyzing just, you know, a single file or a handful files in your repo to getting AI assistance across the entire code base, it's unlocking some really amazing stuff. And by the way, this capability has only been available from our friends in Google DeepMind for a few months now, right? It hasn't even been out there that long. So what we're showing you today is just a taste of what we're going to be able to deliver over the next few months.

>> So as VP of developer experience, what is the scope? Take a minute to explain what are you overseeing, what are your products? What are your areas?

>> How much time do you have for this segment?

>> You got 10 seconds, go. (panel laughs)

>> Look after the console experiences at GCP, how people sign up and onboard, the, you know, CLI, the SDKs, you know, how developers interact, you know, with it at a low level with cloud, terraform bindings, things like that. Look after our developer tools and developer operations business, our cloud observability business, internal platforms inside of Google, so the way developers in Google build stuff on Google, as well as our developer relations teams and technical writer teams. A lot of the folks who are staffing the booths here at the conference today.

>> Well, that's a great start, quick hit. So my next question should obviously be, obviously, we know you from theCUBE at KubeCon, CloudNativeCon, and one of the themes this year, you guys did bring the full package, love the success. Ecosystem is really showing real good proof points, big booths, good after hour parties, lot of business being discussed, but ecosystem has to be successful. The developer and the partner ecosystem has to be a success this year, how do you guys plan to do that? What's the secret sauce? Is it more developer experience? Give us your division because it's looking really good right now with the partnerships.

>> So it's a fantastic question, John, and let me make this really, really concrete. In our keynotes, in the developer keynote, you saw Guillermo, CEO of Vercel, right? So charity majors of Honeycomb. Josh, you know, from Spring, like we are bringing, in my session, at the spotlight session, we had Danny, the CTO at Snyk. Like we are taking ecosystem development incredibly seriously, because our belief at Google is we have to meet developers where they are, right? We're not trying to enforce a Google centric tool chain on our audience, right? We know people are going to use GitHub, we know people are going to use, you know, GitLab, other ecosystem tools. We want to make sure we're bringing those experiences and those products and making them feel natural on Google Cloud.

>> Yeah, and the other thing that we were talking about, Dustin, was two things. Your show didn't need Jensen from NVIDIA to show up, which is a good indicator you brought the goods.

>> The show's not over yet, it could still happen.

>> Where's the leather jacket? We did talk about that in the analyst segment. We will put that on the side. Dustin and I were talking about the Kubernetes' 10th anniversary. Big part of your role now, you know, serverless, you know, you got containers, you got the orchestration with Kubernetes, big part of the stack, big bet paid off big time, how is that going to play into this next year and certainly in the developer experience?

>> You know, as someone who had a long history with Kubernetes, I have to say I was always envious of Google's position as, you know, one of the creators of containers, right? Which are one of the key constructs inside of Kubernetes. And, you know, the ability to now be working on in creating developer experiences that are going to drive the end-to-end software development lifecycle for the best cloud of containers, that's really exciting for me as someone of my background, I have to say, yeah.

>> If you can't beat 'em, join 'em, was that what happened?

>> Pretty much, right?

>> Just kidding, no, yeah, great run at Microsoft and DO, of course. You talked about your scope a little bit, outbound product management, which I'm always fascinated by Google's approach to getting out in front of developers, in front of technologists. Certainly, that gives you some insight into what that field is asking for out of AI, out of the developer experience from Google. What's the feedback so far? Both, you know, where is it already helping and, you know, what are they asking for next?

>> It's a great question and I'll say there's some real clear signal that I've gotten from the customer meetings and individual developer meetings I've had. A few things I'll say, first is a lot of folks have been trialing out other coding assistants on the market, including ones from Microsoft. And there's a moment of, "Wait a minute, we got to pay attention to what's going on at Google." What's happening with Gemini Code Assist, what's happening with large context window, with some of the deep model innovations powered by our integrated stack of infrastructure. It's really capturing people's attention and causing folks to really take a deep look at what we're doing, so that's really exciting. The second thing is, I think there's a lot of questions around how do we measure the impact of these AI coding systems? How do we know that they're doing a good job?

>> That's a huge question right now. I'm glad you brought that up.

>> And on the one hand it's interesting because, you know, given how these things are priced, you know, our list price for this is $19 a month per developer. When you think about, you know, how much we pay developers typically, doesn't cost a lot to get an ROI on 19 bucks a month, right? But people still want to know, like, how much productivity are we really getting? And unfortunately, the answer kind of comes back to what are the standard ways that you measure software engineering efficiency, right? There is no magical answer to this that we've had in the industry. If you're aware of one, I'd love to know it, right?

>> DORA Metrics is the closest I think we've ever seen, you know?

>> And that is exactly what we're hearing from customers. You know, Mark Quigley, director of engineering enablement at Wayfair, you know, I had him up on stage before talking about, you know, how they're looking at developer productivity with Gemini Code Assist. DORA metrics, that's what they're using over there, yeah.

>> Open source community, obviously very important to you, it's part of what you do. What's the feedback loop like between you and the community as you're developing new products?

>> That's a great question, and you know, what I'll say is it's changed a bit in the era of AI, right? Things are moving so fast, so quickly.

>> I was curious about this, yeah.

>> Yeah, and there isn't as much open source innovation happening yet, right? You know, my sense is as someone who's been in that space for a while, the open source starts to come in more after things have settled down and begun to commoditize a little bit. And so, I'm looking forward to the moment where we start to get pieces of that around. You know, for example, the stack around how you build RAG applications and things like that. But what I will say is, you know, we announced LangChain support in lots of different products. LODB, Cloud Run, Serverless Framework, all supporting LangChain and the ability to build a RAG application, you know? A gen AI app that uses your own operational data to provide meaningful insights to customers. You do that on like 10 lines of code now and developers are really, really impressed with that stuff and the feedback's been very positive.

>> I love that, I love the BigQuery having vector embedded into it, so it going to make it so much easier to do that. How's the developer interface into that, one question, and the other one is, Gemini Pro integrated into Code Assist, is that 1.0, 1.5? What's the plans for the new upgrade going into Code Assist?

>> So it's actually very interesting, the technology that we use behind the scenes in Gemini Code Assist use a variety of models actually. And, you know, one thing that's interesting is we do in fact use Gemini 1.5 Pro with the million token context window,

but we don't use for- >> In Code Assist now?

>> In Code Assist, yeah, in private preview now. But we don't use that for everything, right? Because, you know, we want to make sure we're optimizing for latency. The 1.5 Pro model could take a while to get back, right? Some of the demos, we saw 45 seconds to get a response. When you're typing code in your IDE, the experience is not great if you're waiting 45 seconds for everything. So we use different models tuned at different levels to provide a really seamless experience as folks are authoring. But if you want to do something like add an entire feature to my repository, that's a lot of value. And you might be willing to wait 45, 60 seconds or whatever the time may be. Which by the way, those are the times today. This is the worst it's ever going to be.

>> So who tunes that latency? Is that automatically done by Google or developer has knobs for that?

>> We try and build in default gestures that are mapped to different models to really optimize for that latency. But really the key point is those numbers are going to come down over time as, you know, certainly with Google's prowess and infrastructure, we're going to figure out how to get that time down and developer experience is going to be the best.

>> I saw the HashiCorp guys there, they were having a good time last night. They're chatting away. What other partners can you point to that are on this train that you're riding? Who can you point to, notable names in the ecosystem that you could share?

>> Sure, yeah, so we're with the Hashi folks, as you said. We're working with Snyk deeply. Working with LangChain, we're working with Pinecone, working with MongoDB, I mean, it's just a host of partners. In fact, you know, we got a announcement that we made in December around all of the different ecosystem partners that we're working on enabling in Gemini Code Assist. Some of them it's for documentation and knowledge sources, but others were actually fine tuning Gemini Code Assist to understand the languages and frameworks and be essentially context aware about how to write a query, you know, for a document store in MongoDB as an example.

>> ISVs are really digging this too, aren't they? Really getting into the partners are kicking in.

>> Partners are digging in. Yeah, I got to say, a lot of my conversations this week have been with the partner ecosystem. And you know, I think folks are really recognizing that Google is a extremely partner friendly organization. We are out there with open arms, welcoming folks onto big stages, and you know, lots of breakout sessions that are very partner heavy. Expo hall here is filled with them as well.

>> We've talked about it all week.

>> Yeah, I got to say, we were just talking, eight months ago, it wasn't even a year, the last Google Next, so compliments to the team of getting the content together with the backdrop of the landscape is moving so fast, new papers dropping, new code, I mean, you had to kind of keep the eye on the ball, but you guys delivered a lot from Google Next. What are some of the highlights for people who aren't inside the ropes? What's the big accomplishment from things that you delivered on from Next eight months ago and new things are now on the table? How would you put those out there?

>> Well, I would say in general, you know, what we had at at the last Next was a lot of painting a vision for where we thought Google Cloud was going to go. And at this conference, there's been a lot of real honest to god products being shipped across the developer space, across a new area, Gemini Cloud Assist, where we're starting to provide support across the entire SCLC inside a big query. You're starting to see Gemini show up. And you know, you're starting to see just a whole host of innovation that was formerly in preview, now getting into GA territory.

>> I love the comment a couple of minutes ago about this is the worst it's ever going to be, it's just getting better from here. What can you say about dogfooding at Google? You know, I'm an ex-Googler and boy, we love dogfooding things, except that Gemini didn't exist when I was at Google five years ago. What's it like to be a developer at Google today?

>> I got to say it's really exciting because the tools and the quality of, you know, tooling inside of Google, I think, you know, renowned at Google for having some of the best tooling in the industry. But now with Gemini added on as sort of an addendum to this tooling, everything just became an order of magnitude more powerful. So I can just tell you personally-

>> Super coders. (John laughs) But you know, a lot of my job, I do still write code and I try and prioritize as much of that as I can. You see some of my repos up, but I spend a lot of time in Workspace. And I got to say Gemini in Workspace has really served to accelerate the doc writing, reviewing documents, building slides, and a whole host of other things. And I think people underestimate how much of that collaboration experience is a part of the overall software development experience. And that piece has really impacted, I think, developers and, you know, management executives, and a whole host of folks inside of Google.

>> I think that's a really great point you just brought up. It's sometimes the little pieces are incredibly important in terms of that user experience, whatever that might be. I'm curious, so we've talked a lot about good things. What are some of the mistakes that you see people making when they're approaching their strategy for their developer experience adopting AI?

>> That's a really fantastic and insightful question. I've had some customer conversations this week that are actually organizational and cultural conversations masquerading as product and technology conversations, right? You know, how can you buy, how can you sell me this product that is going to fix, you know, this? And it's like you ask five whys and it's like, wait a minute, that's actually a structural challenge inside of your organization. And so, you know, for me, the mistake I see is people trying to go from zero to a hundred miles an hour with technology like AI assistance before they figured out the fundamentals, fundamentals in software specifically. You don't have CICD pipelines in place. If you don't have good monitoring observability, you know, you might not be tall enough to ride the AI coding assistance ride, you know? And so, you know, just making sure that you're sequencing this stuff in the proper order and not going too fast is important. So concretely, I suggest folks run small pilots, you know, have really scoped outcomes and work with a partner if you can who can really coach you and provide some, you know, external validation around your approach.

>> Measurement too, right? I mean, measure before and after.

>> Yes, yeah. As they say, measure twice, cut once, right?

>> What's the coolest thing you're seeing here at the show? Coolest thing?

>> Well, you know, again, I think the coding assistance stuff and the large context window stuff, but I will say as a huge Workspace fan, what we announced with Google Vids, short form video, inside of Workspace. Like we send out newsletters to my team, you know, in order to keep folks up to speed. I'm just imagining like, can we turn that into a short form video and, you know, have some pictures of the team, maybe some cuts of some meetings, and really turn that newsletter into like a TikTok video. You know, that folks can watch. I suspect we'll get a lot more engagement and certainly create a more fun work environment.

>> That'd be great for theCUBE, we'll integrate some of that into our CUBE action. So I got to ask some questions for both of you guys and Savannah too, here. I get this question a lot. It's out there, a little bit out there, so I'll just preface with that.

>> We're here for it John,

let's go. >> Everybody wants to know what the killer app is for AI. Is there a killer app for AI? What does it look like? Or is the killer app every app? So everyone wants to know what the killer app is for AI. Is there a killer app?

>> Yo, I would say that AI is kind of like the internet, right? And it's going to, like what's the killer app for the internet, right? It turns out it's been lots of different things.

>> Chrome. (panel laughs)

>> Chrome, right, yeah. Maybe Netscape before that, right? But you know, there's lots of different use cases, but what I'll say concretely is the vision that Tom has painted during the main keynote of a future of connected agents. You know, that to me is really what this is all about, right? Because pretty soon these models are going to turn from simple prompt and response interactions towards provide some intents. And then, you know, there'll be a plan developed and there'll be a set of steps. And those steps might require a series of prompt and response interactions all trending towards a defined outcome, right? And you're starting to see those capabilities with our large context window and capabilities we have with full code base awareness and Gemini Code Assist, but that to me is really the future. And I think when you think about it, the developer who's surrounded by experts, AI experts, maybe some human experts, and you think about how that can accelerate the amount of software that's created. And let's be honest, there's more software that needs to be created than there are developers to create it, that is really, really exciting. And to me, that's the killer app in the software industry.

>> Well, just to follow up on internet example, I've been using that on theCUBE Pod a lot. You could use the internet by building webpages and put content on there, people can navigate, discover, consume content. The library is no longer needed. I don't need to go to the library anymore, so it's displaced. And then you got search engines, you got bigger, more bigger apps. AI's the same thing. You could use AI and then you could build AI. So the question comes back to when do you use and when do you buy or wait, build your own AI versus just using it? You can't afford like machine learning engineers.

>> That's an important point because one thing we're seeing with customers is gen AI is being applied to productivity use cases and it's being applied to innovation around building AI first experiences for, you know, someone's end customers, right? And in my app dev spotlight session, we actually covered both. We talked about how you can use AI to enable productivity, you know, ideally deliver 10x the amount of software that you're able to deliver. But what are you going to do with that new productivity game? Well, hopefully you're going to put gen AI into a bunch of application workloads and really get AI on both sides of the equation, on the building side as well as on the delivery and creation of end user value.

>> That's the killer app's AI. How do you top that answer?

>> You know, I think, I'm not going to top that answer. I'm going to go a different direction, which is I think it's wonderful that we can ask the AI questions and get back insightful results, and ask it to do things for us, and it goes and does it. I'm actually looking forward to the point at which the prompts are even fewer and farther in between. That it just generally does the right thing for you. And it's less about, you know, in the home automation space, it's less about asking to turn on the lights, and turn off the TV, and open the garage, and it just happens automatically because it's the right thing to do, obviously in this situation, you know? And that's the whole automation side of things. In the coding side of things, it's, you know, instead of having to tell the AI, the code assistant make me a function that queries the database and sort by this and, you know, prompt on that, you more describe what your in-state is and you know, it happens.

>> The other thing about cross modalities is great, the reasoning around multimodal is a big feature that came out of this show. You don't need to worry about plugging the modalities in there, you say, "Hey, I got video, I got the audio, I got DNA, I got text, I got vision."

>> I think the best AI is going to be the invisible AI. When it's at a consumer level and we don't even know that we're interacting with it and it's making my mother's life better, or our family's lives better, or our planet healthier and safer in a lot of different ways, and we aren't even the ones having to prompt that. I'm looking forward to that. We just built on each other quite well there, gentlemen, I just want to say thank you. Final question for you, Gabe, since you are a VIP alum and have probably sat in this chair more times than I have, what do you hope to be able to say the next time you're on the show that you can't say today?

>> Oh, that's a really good one. For me, it's less about what we can say from a product and a technology perspective and more about what we can say from an outcome perspective. What I'd love to have is a set of customers with some really deep challenges who are able to get on a stage with us and say, "Look, Google Gemini helped us accomplish something that we thought was formerly impossible." I'll give you a couple examples of that. A lot of conversations I'm having recently have been around, I got a lot of Brownfield legacy environments. And the thing that's holding me back from building the future is actually these Brownfield environments. And can we apply Google's AI in a way that will allow me to accelerate, remove that weight from the organization and allow us to really build the future and prioritize that?

>> That's something that almost anyone who's been around enterprise software can relate to. If you've got anything that's 5, 10 years old, you can relate to, you know, Brownfield.

>> Are they replacing it or they putting a wrapper around it and having some assistant agent run it?

>> Well, it depends, right? In some cases, they want to wrap it, in other cases, they want to modernize it and turn it into the future, but I think the point is they got to deal with it. And I think a lot of folks have been kind of kicking the can down the road a little bit. And a year from now, I think the time's going to be over for that. I think it's going to be time for us to really deal with this stuff and really get enterprises into a spot where they can start to advance and really compete because the opportunity here is too big to miss it, yeah.

>> I'm excited, can't wait to have you and your customers on the show next time to discuss precisely that. Gabe, thank you so much for being on the show. Really appreciate it. Dustin, John, always a pleasure. And thank all of you for tuning in to our live coverage here from Google Cloud Next in fabulous Las Vegas, Nevada. My name's Savannah Peterson. You're watching "theCube," the leading source for enterprise tech news. (gentle music)

Gabe Monroy | Google Cloud

search

(gentle music)

>> Good afternoon Cloud Community, and welcome back to beautiful Las Vegas, Nevada. We're here in the afternoon of day three of Google Cloud Next. It has been an insanely exciting week. My name is Savannah Peterson, joined by analysts and brilliant humans all around, Dustin, John, thank you so much for being here. Dustin, we haven't gotten to do a segment yet to this show,

>> No, how did that happen? And it's like day three, and this is the first time we're on the desk together.

>> Criminal, criminal. >> I know.

>> We'll have to talk to someone about that.

>> Yeah. >> I don't know who. (panel laughs)

>> That's me. We'll fix that. Have that fixed, we'll take care of it. Make a note of that. AI, where's my AI assistant?

>> We also got the VP of developer experience at Google with us, Gabe, thank you so much for being here.

>> Thank you for having me.

>> You are a total alum. I think you've been on the show how many times, five, six?

>> I lost count.

>> He's up on the leaderboard definitely, he's good.

>> He's definitely on there.

>> On KubeCon, he's up on the top 10 easy.

>> We need the "Saturday Night Live" jacket, you know? The jacket you get, the five timers club, right?

>> I like that, I like that, sign me up.

>> I like that. >> He's a VIP for sure. You won't stand in line for our party. (panel laughs)

>> Exactly. It's been a huge week for Google. How has this been for you? Give us some highlights.

>> It is been really exciting. You know, for me, I always think about all the engineers and all the teams that have been working so hard over, you know, the past few months to really land all the product impact and products at a show like this. And I'd say I'm really impressed with how this is all being received, right? You know, folks have been trying to figure out what does this AI journey mean for me? And I think what you're starting to see on both the main stage keynotes, the developer keynotes, all in these breakout sessions, is we got the goods, we brought it, and I think that's super exciting

>> Google really did bring the goods, ton of announcements this week. Any favorites for you?

>> I'm biased to my own stuff, of course. But I got to say the full code base awareness that we have in Gemini Code Assist, the ability to move from analyzing just, you know, a single file or a handful files in your repo to getting AI assistance across the entire code base, it's unlocking some really amazing stuff. And by the way, this capability has only been available from our friends in Google DeepMind for a few months now, right? It hasn't even been out there that long. So what we're showing you today is just a taste of what we're going to be able to deliver over the next few months.

>> So as VP of developer experience, what is the scope? Take a minute to explain what are you overseeing, what are your products? What are your areas?

>> How much time do you have for this segment?

>> You got 10 seconds, go. (panel laughs)

>> Look after the console experiences at GCP, how people sign up and onboard, the, you know, CLI, the SDKs, you know, how developers interact, you know, with it at a low level with cloud, terraform bindings, things like that. Look after our developer tools and developer operations business, our cloud observability business, internal platforms inside of Google, so the way developers in Google build stuff on Google, as well as our developer relations teams and technical writer teams. A lot of the folks who are staffing the booths here at the conference today.

>> Well, that's a great start, quick hit. So my next question should obviously be, obviously, we know you from theCUBE at KubeCon, CloudNativeCon, and one of the themes this year, you guys did bring the full package, love the success. Ecosystem is really showing real good proof points, big booths, good after hour parties, lot of business being discussed, but ecosystem has to be successful. The developer and the partner ecosystem has to be a success this year, how do you guys plan to do that? What's the secret sauce? Is it more developer experience? Give us your division because it's looking really good right now with the partnerships.

>> So it's a fantastic question, John, and let me make this really, really concrete. In our keynotes, in the developer keynote, you saw Guillermo, CEO of Vercel, right? So charity majors of Honeycomb. Josh, you know, from Spring, like we are bringing, in my session, at the spotlight session, we had Danny, the CTO at Snyk. Like we are taking ecosystem development incredibly seriously, because our belief at Google is we have to meet developers where they are, right? We're not trying to enforce a Google centric tool chain on our audience, right? We know people are going to use GitHub, we know people are going to use, you know, GitLab, other ecosystem tools. We want to make sure we're bringing those experiences and those products and making them feel natural on Google Cloud.

>> Yeah, and the other thing that we were talking about, Dustin, was two things. Your show didn't need Jensen from NVIDIA to show up, which is a good indicator you brought the goods.

>> The show's not over yet, it could still happen.

>> Where's the leather jacket? We did talk about that in the analyst segment. We will put that on the side. Dustin and I were talking about the Kubernetes' 10th anniversary. Big part of your role now, you know, serverless, you know, you got containers, you got the orchestration with Kubernetes, big part of the stack, big bet paid off big time, how is that going to play into this next year and certainly in the developer experience?

>> You know, as someone who had a long history with Kubernetes, I have to say I was always envious of Google's position as, you know, one of the creators of containers, right? Which are one of the key constructs inside of Kubernetes. And, you know, the ability to now be working on in creating developer experiences that are going to drive the end-to-end software development lifecycle for the best cloud of containers, that's really exciting for me as someone of my background, I have to say, yeah.

>> If you can't beat 'em, join 'em, was that what happened?

>> Pretty much, right?

>> Just kidding, no, yeah, great run at Microsoft and DO, of course. You talked about your scope a little bit, outbound product management, which I'm always fascinated by Google's approach to getting out in front of developers, in front of technologists. Certainly, that gives you some insight into what that field is asking for out of AI, out of the developer experience from Google. What's the feedback so far? Both, you know, where is it already helping and, you know, what are they asking for next?

>> It's a great question and I'll say there's some real clear signal that I've gotten from the customer meetings and individual developer meetings I've had. A few things I'll say, first is a lot of folks have been trialing out other coding assistants on the market, including ones from Microsoft. And there's a moment of, "Wait a minute, we got to pay attention to what's going on at Google." What's happening with Gemini Code Assist, what's happening with large context window, with some of the deep model innovations powered by our integrated stack of infrastructure. It's really capturing people's attention and causing folks to really take a deep look at what we're doing, so that's really exciting. The second thing is, I think there's a lot of questions around how do we measure the impact of these AI coding systems? How do we know that they're doing a good job?

>> That's a huge question right now. I'm glad you brought that up.

>> And on the one hand it's interesting because, you know, given how these things are priced, you know, our list price for this is $19 a month per developer. When you think about, you know, how much we pay developers typically, doesn't cost a lot to get an ROI on 19 bucks a month, right? But people still want to know, like, how much productivity are we really getting? And unfortunately, the answer kind of comes back to what are the standard ways that you measure software engineering efficiency, right? There is no magical answer to this that we've had in the industry. If you're aware of one, I'd love to know it, right?

>> DORA Metrics is the closest I think we've ever seen, you know?

>> And that is exactly what we're hearing from customers. You know, Mark Quigley, director of engineering enablement at Wayfair, you know, I had him up on stage before talking about, you know, how they're looking at developer productivity with Gemini Code Assist. DORA metrics, that's what they're using over there, yeah.

>> Open source community, obviously very important to you, it's part of what you do. What's the feedback loop like between you and the community as you're developing new products?

>> That's a great question, and you know, what I'll say is it's changed a bit in the era of AI, right? Things are moving so fast, so quickly.

>> I was curious about this, yeah.

>> Yeah, and there isn't as much open source innovation happening yet, right? You know, my sense is as someone who's been in that space for a while, the open source starts to come in more after things have settled down and begun to commoditize a little bit. And so, I'm looking forward to the moment where we start to get pieces of that around. You know, for example, the stack around how you build RAG applications and things like that. But what I will say is, you know, we announced LangChain support in lots of different products. LODB, Cloud Run, Serverless Framework, all supporting LangChain and the ability to build a RAG application, you know? A gen AI app that uses your own operational data to provide meaningful insights to customers. You do that on like 10 lines of code now and developers are really, really impressed with that stuff and the feedback's been very positive.

>> I love that, I love the BigQuery having vector embedded into it, so it going to make it so much easier to do that. How's the developer interface into that, one question, and the other one is, Gemini Pro integrated into Code Assist, is that 1.0, 1.5? What's the plans for the new upgrade going into Code Assist?

>> So it's actually very interesting, the technology that we use behind the scenes in Gemini Code Assist use a variety of models actually. And, you know, one thing that's interesting is we do in fact use Gemini 1.5 Pro with the million token context window,

but we don't use for- >> In Code Assist now?

>> In Code Assist, yeah, in private preview now. But we don't use that for everything, right? Because, you know, we want to make sure we're optimizing for latency. The 1.5 Pro model could take a while to get back, right? Some of the demos, we saw 45 seconds to get a response. When you're typing code in your IDE, the experience is not great if you're waiting 45 seconds for everything. So we use different models tuned at different levels to provide a really seamless experience as folks are authoring. But if you want to do something like add an entire feature to my repository, that's a lot of value. And you might be willing to wait 45, 60 seconds or whatever the time may be. Which by the way, those are the times today. This is the worst it's ever going to be.

>> So who tunes that latency? Is that automatically done by Google or developer has knobs for that?

>> We try and build in default gestures that are mapped to different models to really optimize for that latency. But really the key point is those numbers are going to come down over time as, you know, certainly with Google's prowess and infrastructure, we're going to figure out how to get that time down and developer experience is going to be the best.

>> I saw the HashiCorp guys there, they were having a good time last night. They're chatting away. What other partners can you point to that are on this train that you're riding? Who can you point to, notable names in the ecosystem that you could share?

>> Sure, yeah, so we're with the Hashi folks, as you said. We're working with Snyk deeply. Working with LangChain, we're working with Pinecone, working with MongoDB, I mean, it's just a host of partners. In fact, you know, we got a announcement that we made in December around all of the different ecosystem partners that we're working on enabling in Gemini Code Assist. Some of them it's for documentation and knowledge sources, but others were actually fine tuning Gemini Code Assist to understand the languages and frameworks and be essentially context aware about how to write a query, you know, for a document store in MongoDB as an example.

>> ISVs are really digging this too, aren't they? Really getting into the partners are kicking in.

>> Partners are digging in. Yeah, I got to say, a lot of my conversations this week have been with the partner ecosystem. And you know, I think folks are really recognizing that Google is a extremely partner friendly organization. We are out there with open arms, welcoming folks onto big stages, and you know, lots of breakout sessions that are very partner heavy. Expo hall here is filled with them as well.

>> We've talked about it all week.

>> Yeah, I got to say, we were just talking, eight months ago, it wasn't even a year, the last Google Next, so compliments to the team of getting the content together with the backdrop of the landscape is moving so fast, new papers dropping, new code, I mean, you had to kind of keep the eye on the ball, but you guys delivered a lot from Google Next. What are some of the highlights for people who aren't inside the ropes? What's the big accomplishment from things that you delivered on from Next eight months ago and new things are now on the table? How would you put those out there?

>> Well, I would say in general, you know, what we had at at the last Next was a lot of painting a vision for where we thought Google Cloud was going to go. And at this conference, there's been a lot of real honest to god products being shipped across the developer space, across a new area, Gemini Cloud Assist, where we're starting to provide support across the entire SCLC inside a big query. You're starting to see Gemini show up. And you know, you're starting to see just a whole host of innovation that was formerly in preview, now getting into GA territory.

>> I love the comment a couple of minutes ago about this is the worst it's ever going to be, it's just getting better from here. What can you say about dogfooding at Google? You know, I'm an ex-Googler and boy, we love dogfooding things, except that Gemini didn't exist when I was at Google five years ago. What's it like to be a developer at Google today?

>> I got to say it's really exciting because the tools and the quality of, you know, tooling inside of Google, I think, you know, renowned at Google for having some of the best tooling in the industry. But now with Gemini added on as sort of an addendum to this tooling, everything just became an order of magnitude more powerful. So I can just tell you personally-

>> Super coders. (John laughs) But you know, a lot of my job, I do still write code and I try and prioritize as much of that as I can. You see some of my repos up, but I spend a lot of time in Workspace. And I got to say Gemini in Workspace has really served to accelerate the doc writing, reviewing documents, building slides, and a whole host of other things. And I think people underestimate how much of that collaboration experience is a part of the overall software development experience. And that piece has really impacted, I think, developers and, you know, management executives, and a whole host of folks inside of Google.

>> I think that's a really great point you just brought up. It's sometimes the little pieces are incredibly important in terms of that user experience, whatever that might be. I'm curious, so we've talked a lot about good things. What are some of the mistakes that you see people making when they're approaching their strategy for their developer experience adopting AI?

>> That's a really fantastic and insightful question. I've had some customer conversations this week that are actually organizational and cultural conversations masquerading as product and technology conversations, right? You know, how can you buy, how can you sell me this product that is going to fix, you know, this? And it's like you ask five whys and it's like, wait a minute, that's actually a structural challenge inside of your organization. And so, you know, for me, the mistake I see is people trying to go from zero to a hundred miles an hour with technology like AI assistance before they figured out the fundamentals, fundamentals in software specifically. You don't have CICD pipelines in place. If you don't have good monitoring observability, you know, you might not be tall enough to ride the AI coding assistance ride, you know? And so, you know, just making sure that you're sequencing this stuff in the proper order and not going too fast is important. So concretely, I suggest folks run small pilots, you know, have really scoped outcomes and work with a partner if you can who can really coach you and provide some, you know, external validation around your approach.

>> Measurement too, right? I mean, measure before and after.

>> Yes, yeah. As they say, measure twice, cut once, right?

>> What's the coolest thing you're seeing here at the show? Coolest thing?

>> Well, you know, again, I think the coding assistance stuff and the large context window stuff, but I will say as a huge Workspace fan, what we announced with Google Vids, short form video, inside of Workspace. Like we send out newsletters to my team, you know, in order to keep folks up to speed. I'm just imagining like, can we turn that into a short form video and, you know, have some pictures of the team, maybe some cuts of some meetings, and really turn that newsletter into like a TikTok video. You know, that folks can watch. I suspect we'll get a lot more engagement and certainly create a more fun work environment.

>> That'd be great for theCUBE, we'll integrate some of that into our CUBE action. So I got to ask some questions for both of you guys and Savannah too, here. I get this question a lot. It's out there, a little bit out there, so I'll just preface with that.

>> We're here for it John,

let's go. >> Everybody wants to know what the killer app is for AI. Is there a killer app for AI? What does it look like? Or is the killer app every app? So everyone wants to know what the killer app is for AI. Is there a killer app?

>> Yo, I would say that AI is kind of like the internet, right? And it's going to, like what's the killer app for the internet, right? It turns out it's been lots of different things.

>> Chrome. (panel laughs)

>> Chrome, right, yeah. Maybe Netscape before that, right? But you know, there's lots of different use cases, but what I'll say concretely is the vision that Tom has painted during the main keynote of a future of connected agents. You know, that to me is really what this is all about, right? Because pretty soon these models are going to turn from simple prompt and response interactions towards provide some intents. And then, you know, there'll be a plan developed and there'll be a set of steps. And those steps might require a series of prompt and response interactions all trending towards a defined outcome, right? And you're starting to see those capabilities with our large context window and capabilities we have with full code base awareness and Gemini Code Assist, but that to me is really the future. And I think when you think about it, the developer who's surrounded by experts, AI experts, maybe some human experts, and you think about how that can accelerate the amount of software that's created. And let's be honest, there's more software that needs to be created than there are developers to create it, that is really, really exciting. And to me, that's the killer app in the software industry.

>> Well, just to follow up on internet example, I've been using that on theCUBE Pod a lot. You could use the internet by building webpages and put content on there, people can navigate, discover, consume content. The library is no longer needed. I don't need to go to the library anymore, so it's displaced. And then you got search engines, you got bigger, more bigger apps. AI's the same thing. You could use AI and then you could build AI. So the question comes back to when do you use and when do you buy or wait, build your own AI versus just using it? You can't afford like machine learning engineers.

>> That's an important point because one thing we're seeing with customers is gen AI is being applied to productivity use cases and it's being applied to innovation around building AI first experiences for, you know, someone's end customers, right? And in my app dev spotlight session, we actually covered both. We talked about how you can use AI to enable productivity, you know, ideally deliver 10x the amount of software that you're able to deliver. But what are you going to do with that new productivity game? Well, hopefully you're going to put gen AI into a bunch of application workloads and really get AI on both sides of the equation, on the building side as well as on the delivery and creation of end user value.

>> That's the killer app's AI. How do you top that answer?

>> You know, I think, I'm not going to top that answer. I'm going to go a different direction, which is I think it's wonderful that we can ask the AI questions and get back insightful results, and ask it to do things for us, and it goes and does it. I'm actually looking forward to the point at which the prompts are even fewer and farther in between. That it just generally does the right thing for you. And it's less about, you know, in the home automation space, it's less about asking to turn on the lights, and turn off the TV, and open the garage, and it just happens automatically because it's the right thing to do, obviously in this situation, you know? And that's the whole automation side of things. In the coding side of things, it's, you know, instead of having to tell the AI, the code assistant make me a function that queries the database and sort by this and, you know, prompt on that, you more describe what your in-state is and you know, it happens.

>> The other thing about cross modalities is great, the reasoning around multimodal is a big feature that came out of this show. You don't need to worry about plugging the modalities in there, you say, "Hey, I got video, I got the audio, I got DNA, I got text, I got vision."

>> I think the best AI is going to be the invisible AI. When it's at a consumer level and we don't even know that we're interacting with it and it's making my mother's life better, or our family's lives better, or our planet healthier and safer in a lot of different ways, and we aren't even the ones having to prompt that. I'm looking forward to that. We just built on each other quite well there, gentlemen, I just want to say thank you. Final question for you, Gabe, since you are a VIP alum and have probably sat in this chair more times than I have, what do you hope to be able to say the next time you're on the show that you can't say today?

>> Oh, that's a really good one. For me, it's less about what we can say from a product and a technology perspective and more about what we can say from an outcome perspective. What I'd love to have is a set of customers with some really deep challenges who are able to get on a stage with us and say, "Look, Google Gemini helped us accomplish something that we thought was formerly impossible." I'll give you a couple examples of that. A lot of conversations I'm having recently have been around, I got a lot of Brownfield legacy environments. And the thing that's holding me back from building the future is actually these Brownfield environments. And can we apply Google's AI in a way that will allow me to accelerate, remove that weight from the organization and allow us to really build the future and prioritize that?

>> That's something that almost anyone who's been around enterprise software can relate to. If you've got anything that's 5, 10 years old, you can relate to, you know, Brownfield.

>> Are they replacing it or they putting a wrapper around it and having some assistant agent run it?

>> Well, it depends, right? In some cases, they want to wrap it, in other cases, they want to modernize it and turn it into the future, but I think the point is they got to deal with it. And I think a lot of folks have been kind of kicking the can down the road a little bit. And a year from now, I think the time's going to be over for that. I think it's going to be time for us to really deal with this stuff and really get enterprises into a spot where they can start to advance and really compete because the opportunity here is too big to miss it, yeah.

>> I'm excited, can't wait to have you and your customers on the show next time to discuss precisely that. Gabe, thank you so much for being on the show. Really appreciate it. Dustin, John, always a pleasure. And thank all of you for tuning in to our live coverage here from Google Cloud Next in fabulous Las Vegas, Nevada. My name's Savannah Peterson. You're watching "theCube," the leading source for enterprise tech news. (gentle music)