Gabe Monroy, Google Cloud | Google Cloud Next ’24
[Savannah Peterson]
Welcome to Google Cloud Next
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.
Introducing the Guests
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.
[Dustin Sias]
No, how did that happen and what’s like day three and this is the first time we’re on the desk together.
[Savannah Peterson]
Developer Experience at Google
We’ll have to talk to someone about that.
[John Furrier]
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?
[Savannah Peterson]
We also got the VP of developer experience at Google with us. Gabe, thank you so much for being here.
[John Furrier]
Thank you for having me.
[Savannah Peterson]
You are a total alum. I think you’ve been on the show how many times? Five?
[John Furrier]
I lost count. He’s up on the leaderboard. On KubeCon, he’s up in the top ten.
Google Brought the Goods
Easy.
[Dustin Sias]
We need the Saturday Night Live jacket. The jacket you get at the Five Timers Club. I like that.
[00:01:00]
[John Furrier]
Sign me up. He’s a VIP for sure. He won’t stand in line for our party.
[Savannah Peterson]
Google’s Highlights
Exactly. It’s been a huge week for Google. How has this been for you? Give us some highlights.
[Gabe Monroy]
It’s been really exciting. For me, I always think about all the engineers and all the teams that have been working so hard over the past few months to really land all the product impact and product truth at a show like this. I’d say I’m really impressed with how this is all being received. Folks have been trying to figure out what does this AI journey mean for me? 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. I think that’s super exciting.
[Savannah Peterson]
Gemini Code Assist
Google really did bring the goods. A ton of announcements this week. Any favorites for you?
[Gabe Monroy]
I’m biased to my own stuff, of course, but I the full codebase awareness that we have in Gemini Code Assist. The ability to move from analyzing just a single file or a handful of files in your repo to getting AI assistance across the entire codebase.
[00:02:02]
It’s unlocking some really amazing stuff. By the way, this capability has only been available from our friends in Google DeepMind for a few months now. It hasn’t even been out there that long. 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.
[John Furrier]
Scope of Developer Experience
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?
[Gabe Monroy]
How much time do we have for this segment? You got 10 seconds.
[John Furrier]
Go.
[Gabe Monroy]
Look after the console experiences at GCP, how people sign up and onboard, the CLI, the SDKs, how developers interact 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. 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 quick hit.
[John Furrier]
Ecosystem Success
So my next question, obviously we know you from theCUBE, but KubeCon, CloudNativeCon, and one of the themes this year, you guys did bring the full package, love the success.
[00:03:08]
Ecosystem is really showing real good proof points, big booths, good after-hour parties, a lot of business being discussed, but ecosystem has to be successful. The developer and the partner ecosystem has to be a success this year.
Developer and Partner Ecosystem
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.
[Gabe Monroy]
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, our charity majors of Honeycomb. You saw Josh from Spring. In my session, my active spotlight session, we had Danny, the CTO at Snyk. We are taking ecosystem development incredibly seriously, because our belief at Google is we have to meet developers where they are.
[00:04:00]
We’re not trying to enforce a Google-centric tool chain on our audience. We know people are going to use GitHub. We know people are going to use 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.
[John Furrier]
Meeting Developers Where They Are
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.
[Gabe Monroy]
The show’s not over yet. This could still happen. Where’s the leather jacket?
[John Furrier]
We did talk about that in the analysis segment. We’ll put that aside.
Kubernetes and Serverless
Dustin and I were talking about Kubernetes, 10th anniversary, a big part of your role now, serverless. You’ve got containers. You’ve got orchestration with Kubernetes. Big part of that 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?
[Gabe Monroy]
Google’s Position in Containers
As someone who had a long history with Kubernetes, I have to say, I was always envious of Google’s position as one of the creators of containers, which are one of the key constructs inside of Kubernetes.
[00:05:07]
The ability to now be working on 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 and someone of my background, I have to say.
[Dustin Sias]
Joining the Container Revolution
If you can’t beat them, join them. Is that what happened?
[Gabe Monroy]
Pretty much, right?
Outbound Product Management
Just kidding.
[Dustin Sias]
Great run at Microsoft and DL, 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.
Feedback on AI and Developer Experience
What’s the feedback so far? Both where is it already helping and what are they asking for next?
[Gabe Monroy]
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.
[00:06:04]
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.
Measuring the Impact of AI Coding Assistants
The second thing is, I think there’s a lot of questions around, how do we measure the impact of these AI coding assistants? How do we know that they’re doing a good job?
[Savannah Peterson]
It’s a huge question right now. Right.
[Gabe Monroy]
And I can wrap that up. And on the one hand, it’s interesting because given how these things are priced, our list price for this is $19 a month per developer. When you think about how much we pay developers typically, it doesn’t cost a lot to get an ROI on $19 a month, right? But people still want to know how much productivity are we really getting?
[00:07:00]
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?
[Dustin Sias]
Dora Metrics is the closest I think we’ve ever seen.
[Gabe Monroy]
And that is exactly what we’re hearing from customers. Mark Quigley, Director of Engineering Enablement at Wayfair, had him up on stage before talking about, how they’re looking at developing productivity with Gemini Code Assist. Dora Metrics, that’s what they’re using over there. Yeah.
[Savannah Peterson]
Open Source Community Feedback
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?
[Gabe Monroy]
That’s a great question. And what I’ll say is, it’s changed a bit in the era of AI, right? Things are moving so fast, so quickly.
[Savannah Peterson]
I was just curious about this.
[Gabe Monroy]
Yeah, and there isn’t as much open source innovation happening yet, right? 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.
[00:08:02]
And so I’m looking forward to the moment where we start to get pieces of that around, for example, the stack around how you build RAG applications and things like that. But what I will say is, we announced LangChain support in lots of different products. AlloDB, Cloud Run, Serverless Framework, all supporting LangChain. And the ability to build a RAG application, a Gen AI app that uses your own operational data to provide meaningful insights to customers. You can 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.
[John Furrier]
BigQuery and Gemini Pro
I love that. I love the BigQuery having vector embedded into it. It’s 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?
[Gabe Monroy]
So it’s actually very interesting. The technology that we use behind the scenes in Gemini Code Assist uses a variety of models, actually.
[00:09:04]
And one thing that’s interesting is we do, in fact, use Gemini 1.5 Pro with the million token context window.
Gemini 1.5 Pro in Code Assist
But we don’t use it for…
[John Furrier]
In Code Assist now.
[Gabe Monroy]
In Code Assist, yeah. In private preview now. But we don’t use that for everything, right? Because 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.
[John Furrier]
Latency Tuning
So who tunes that latency? Is that automatically done by Google or developer has no option for that?
[Gabe Monroy]
We try and build in default gestures that are mapped to different models to really optimize for that latency.
[00:10:03]
But really, the key point is those numbers are going to come down over time as certainly with Google’s prowess in infrastructure, we’re going to figure out how to get that time down and developer experience is going to be the benefit.
[John Furrier]
Notable Partners
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 that’s notable names in the ecosystem that you could share?
[Gabe Monroy]
Sure, yes. So we’re with the Hashi folks, as you said. We’re working with SneakDeeply, working with Lanchain. We’re working with Pinecone, working with MongoDB. I mean, it’s just a host of partners. In fact, we got an announcement that we made in December around all of the different ecosystem partners that we’re working on enabling in Gemini Codecis. Some of them is for documentation and knowledge sources, but others, we’re actually fine-tuning Gemini Codecis to understand the languages and frameworks and be essentially context aware about how to write a query for a document store in MongoDB, as an example.
[00:11:04]
ISVs are really digging this too, aren’t they? I’m 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 I think folks are really recognizing that Google is an extremely partner-friendly organization. We are out there with open arms, welcoming folks onto big stages and 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.
[John Furrier]
Content Creation in a Fast-Moving Landscape
Yeah, I got to say, we just saw it 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 dying above. But you guys delivered a lot from Google Next.
Google Next Highlights
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?
[Gabe Monroy]
Well, I would say in general, what we had at the last Next was a lot of painting a vision for where we thought Google Cloud was going to go.
[00:12:09]
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’re starting to see just a whole host of innovation that was formerly in preview, now getting into GA territory.
[Dustin Sias]
Dogfooding at Google
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?
Developer Experience at Google Today
You know, I’m an ex-Googler and boy, we love dogfooding things, except the Gemini didn’t exist when I was at Google five years ago. What’s it like to be a developer at Google today?
[Gabe Monroy]
I got to say it’s really exciting because the tools and the quality of tooling inside of Google, I think, renowned at Google for having some of the best tooling in the industry.
[00:13:02]
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, I’m- Super coders. 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 can see some of my repos up on GitHub. But I’d spend a lot of time in Workspace. And I got to say, Gemini and 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 management executives and a whole host of folks inside of Google.
[Savannah Peterson]
Importance of Small Details
I think that’s a really great point you just brought up. 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.
Mistakes in AI Adoption Strategy
What are some of the mistakes that you see people making when they’re approaching their strategy for their developer experience adopting AI?
[00:14:05]
[Gabe Monroy]
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. How can you buy, how can you sell me this product that is going to fix 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 for me, the mistake I see is people trying to go from zero to 100 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 might not be tall enough to ride the AI coding assistance ride. And so 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, have really scoped outcomes, and work with a partner if you can who can really coach you and provide some external validation around your approach.
[00:15:10]
[Dustin Sias]
Measuring Before and After
Measurement too, right? I mean, measure before and after.
[Gabe Monroy]
Yes, yeah. As they say, measure twice, cut once, right?
[John Furrier]
Coolest Thing at the Show
What’s the coolest thing you’re seeing here at the show? Coolest thing?
[Gabe Monroy]
Well, 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. We sent out newsletters to my team in order to keep folks up to speed. I’m just imagining, can we turn that into a short-form video and have some pictures of the team, maybe some cuts of some meetings, and really turn that newsletter into a TikTok video that folks can watch? I suspect we’ll get a lot more engagement and certainly create a more fun work environment.
[John Furrier]
Killer App for AI
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 you guys and Savannah too here.
[00:16:00]
I get this question a lot. It’s out there, a little bit out there, so I’ll just preface with that.
[Savannah Peterson]
We’re here for it, John, let’s go.
[John Furrier]
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?
[Gabe Monroy]
AI as the Internet
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? It’s, it turns out it’s been lots of different things. Yeah, Chrome, maybe Netscape before that, right? But 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. 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.
[00:17:02]
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-based awareness and Gemini code assist. But that to me is really the future. And I think when you think about a 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 engine.
[John Furrier]
AI as a Killer App
Well, just to follow on the 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 that’s displaced. And then you got search engines. You got bigger, more, bigger apps. AI is 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 build your own AI versus just using it?
[00:18:04]
Using vs. Building AI
Like, we can’t afford machine learning engineers.
[Gabe Monroy]
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 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, ideally deliver 10x the amount of software that you’re able to deliver. But what are you going to do with that new productivity gain? 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 a killer app, AI.
[John Furrier]
So how do you top that answer?
[Dustin Sias]
AI’s Future
You know, I think I’m not going to top that answer.
[John Furrier]
I’m going to go a different direction. Here we go.
[Dustin Sias]
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.
[00:19:06]
I’m actually looking forward to the point at which the prompts are even fewer and farther in between. 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 home automation side of things. And 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 end state is and, you know, it happens.
[John Furrier]
Cross-Modality Reasoning
The other thing about cross modalities is a great reasoning around multimodal is a big feature that came out of this show. You don’t need to worry about plugging the modalities in. They say, hey, I got video, I got the audio, I got DNA, I got text, I got vision.
[00:20:03]
[Savannah Peterson]
Invisible AI
Yeah. 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.
Next Time on the Show
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?
[Gabe Monroy]
Outcome Perspective
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.
[00:21:03]
Addressing Brownfield Legacy Environments
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.
[Dustin Sias]
That’s something that almost anyone who’s been around enterprise software can relate to. If you’ve got anything that’s five, 10 years old, you can relate to brownfield.
[John Furrier]
Modernizing Legacy Systems
Are they replacing it or are they putting a wrapper around it and having some assistant agent run it?
[Gabe Monroy]
Well, it depends, right? In some cases, they want to wrap it. In other cases, they want to modernize it and turn it to the future. But 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.
Enterprise Advancement
I think it’s going to be time for us to really deal with this stuff and really get enterprise into a spot where they can start to advance and really compete because the opportunity here is too big to miss it.
[00:22:05]
[Savannah Peterson]
Conclusion
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 is Savannah Peterson. You’re watching theCUBE, the leading source for enterprise tech news.