Полная расшифровка выпуска Code is cheap now; trust is the new bottleneck

on that. Number one is you have Copilot app projects. So right now a lot of us are creating sessions and interacting with AI. And what you end up in your sidebar is like a thousand things that are very hard to control. And it doesn't accumulate all of the knowledge and automations that you want, etc. So what we've done now is you could create a project for something that is long-lived or maybe short-lived, but it's more than one session overall. And then that just allows you to kind of take your creation to a next level without all the cognitive load. You have an orchestrator that is able to say, okay, I'm going to check on this work, check on this work. I'm going to go in and get whatever needs your attention and then present it to you. So it's a calmer creation, which is what we want. We always talk about keeping developers in the flow and projects and the orchestrator of projects allows us to do that. So that one is huge in our opinion. Then the next one on that is automations. So not only can you have a project as well, but you could have then automations that I think about these automations as things that you could set up a squad on. So you could have a set of security automations that are in a group and that will become your security squad. And then we're going to provide templates and one click for you to get these as well. But think about you're going to triage the security vulnerability, then you're going to try to reproduce and see if you really can reach it. Then you're going to go and fix it. Then you have to verify it. And then we could learn from all of that. And there's a policy coach that tells you, oh, this is the way to do it next time better and next time better and next time better. Jason Hiner: Nice. Mario Rodriguez: So this automation with this intelligence and [unclear] memory of it create an amazing experience. Even if you're a solo developer, I have a bug squasher squad. I have a security team expert squad. I have a squad to make sure my CI builds are continuously getting faster, if I need to go and divide. A lot of us are programming in Rust right now. What happens when one of those crates starts getting bigger and bigger and bigger? We have a way to break up that crate as well on a loop. So I'm really excited about that because it's clearly the future when we talk about this interconnected community. The way that you should think about it is you and I will become those nodes. But these agents and these squads will do too. And then we just through the GitHub app and canvases constantly interact with them. And the last thing that I think is very important is continuity. So you could be working on your mobile device, then pass that into Visual Studio Code, then go into the GitHub app, then go into .com, go into Slack or Teams. And at any point in there you could kind of share the session and then continue the session. So that's really important in the sense of what I am seeing the most is when you are in the flow of creation, you always have an idea and you're like, oh, let me not lose that. So you go into the project and then capture it really quick. And then when you're back at your desk, then you could just continue on it. And because these things can run in the cloud now and it's not only on my device, that just allows us to explore more and more and more, which just really increases the possibilities of what we could achieve. Jason Hiner: Now it's time for a word from this week's sponsor, Nooks. Most sales teams run outbound, inbound and deal management in three separate tools. A lead comes in Friday night, nobody calls till Monday. A rep hangs up from a great call, the recap goes out on Thursday. Deals slip and nobody can say where. Nooks puts every motion on one execution layer. AI agents work alongside your reps. They get calls answered, get emails delivered, send the follow up the same hour and prep to secure the next step. While the reps do the selling. Teams on Nooks book three times more meetings, follow up 20% faster and sales reps get back 10 hours a week. Over 1,800 revenue teams already run their go-to-market on Nooks. See what revenue agents can do at nooks.ai/Deep View. That's nooks.ai/Deep View. And we thank Nooks for their support of The Deep View. And now back to the show. That also lets AI builders have a little bit more of a life too. They're not chained to their desk with the code all the time. Mario Rodriguez: 100% 100%. Jason Hiner: You can get out, touch grass once in a while and be working from mobile on whatever their thing is when those ideas come. Mario Rodriguez: Yes, exactly. I mean, I don't know about you, but a lot of my deep thinking is walking, actually. Jason Hiner: Yes. Absolutely. Mario Rodriguez: And sometimes I have an idea and I might not have put it in my notes app for sure. But at that moment, I may want to tell the agent something, because I don't want to lose it. And by the time I get back from my walk, like my walk, say, usually between 35-40 minutes, then I already have something in there that I could react to. And that, you know what I'm saying? Like that feeling is very special. Jason Hiner: Yeah, very cool. Thanks for summing all of that up. I want to ask some of the bigger picture question, but first I want to double click on the Azure piece and the sort of the scalability and the speed, you know, as well. So, you know, the amount of customers and activity that you are facing has increased enormously right over the past year. So I assume that's part of why, you know, you feel like the need to do this, re-architect the platform to be able to handle that level of activity. You know, you've also had, like, during this year, Anthropic felt like it just had too much volume to still be on GitHub, so it made its own version. Similar to, like, Meta has done that, Google has done that. Is that part of the thinking of, like, making this on Azure, the new version on Azure, so that as you have customers that

grow really large and are doing more, they don't have to feel like they need to go off and build their own version of Git, you know, to handle their volume? Mario Rodriguez: Yeah, 100%. And so, look, we knew that AI, and the adoption was going to accelerate. Unfortunately, what happened to us is also that we ran out of power to be able to add more and more capacity into that data center that we were in. And it was just not one, it's three of them, and it resulted in all those things. And yes, that meant that to be able to deal with, like, a scale like this, and it's almost, you know, vertical. Jason Hiner: Yeah. Mario Rodriguez: Then you're going to have to go into a system that has the ability to add significantly more compute and has the power to serve that compute. We added four million CPU cores. There's no one in the planet that could give you that in less than a quarter. We added 160 petabytes of network storage. Like there's no one other than a hyper-scaler that is going to be able to do that. So we accelerated then, you know, the move into Azure. We already had it in the plan. We were executing on it. Jason Hiner: Okay. Mario Rodriguez: All that happened was an acceleration of that plan, because if not, then GitHub was going to be constantly down, right? So that is what happened. We're already executing towards it. We just accelerated. And then we had a great partnership with Azure on that. So that's kind of number one. Number two is, Anthropic is a great partner. And, you know, we actually collaborated with them in that blog post that they did on Scion. And it was great. You know, clearly they don't want to actually fragment the Git ecosystem. That is not what they're talking about. Jason Hiner: Scion is their version of Git, you know? Mario Rodriguez: Well, yeah, but it's not Git, right? So, and I don't think like the whole point of Scion was not to fragment. So it's not like something that they're going to launch or anything like that. But the point was, hey, they needed, you know, specific writes per second. And yes, right now, that is not available in the way that the infrastructure is set up. That's what we're going towards, like you said. But the reality is, there's only, I don't know, five to ten Anthropics in the world. Just like you were saying, Meta has their own, Google has their own and things like that's perfectly fine. Jason Hiner: Yeah. Mario Rodriguez: Really what we need to do is make sure that the, I don't know, almost a billion repos now can work at machine speed. And those are not like the Anthropic repos for sure. So OpenClaw and what Peter did there, that's what we're trying to mainly address. And yes, these big mono repos, that would necessitate not only what we're doing right now but even more investments as well that are more vertical in nature. So we have to do both. It's the way that I want you to think about it. And from our side, Git is still the right bet. And I just want to make sure that everyone

understands that. From a GitHub side, Git is still the right bet, and we believe that we're going to be able to scale out and then scale up that. So it's not only about the writes but also about the reads. It's not only about that even. It's about your PRs and how you do diffs and change sets. And then it's about CI and how you actually do continuous integration. So it's the entirety of the SDLC. It's not just like one specific item. We're all going to need version control, and we [unclear] Git is the right solution for that. Jason Hiner: OK, that's important. All right. How about over this past year, the journey of the AI coding agents has really changed a lot of things. And I'm sure for you all, I have to imagine that you're thinking about some of these things differently. I have people that I know that are not technical at all that now are coming up to me. And they're telling me that they have a GitHub repo. Mario Rodriguez: Yeah. Jason Hiner: And it sort of blows my mind sometimes. I never would have imagined that like a year ago. With so many more people having access to this. So there's that on the one end. People that are not technical at all now sort of getting into vibe coding and so having a presence in your community. And then you have all of the other folks like project managers, product managers that now like they're building a lot of their own prototypes. Right. I don't need to just go into a meeting with a PowerPoint. I can build a prototype and show them what I'm thinking. And then we can build off of that. And so the scope of your potential community has expanded so dramatically. So what does that mean for you as Chief Product Officer? What does that mean for how you're thinking about the way you build the product? Mario Rodriguez: I think it was in 2023 we started talking about a billion developers. And if you think about that, it was in 23, it was 24 for sure. Jason Hiner: Okay. Mario Rodriguez: And that is 10% of the world population. There's no way you're going to scale 10% of the world population in a CS degree to understand. Jason Hiner: Right. Mario Rodriguez: Like you're not going to be like do that. Jason Hiner: Yeah. Mario Rodriguez: So for us, even that early in 2024, we saw the vision of like we are going to scale a significant amount. This is why I was telling you, we're already doing the move to Azure. As an example, it's just the acceleration of that. Jason Hiner: Okay. Mario Rodriguez: At one moment we thought 2030. That was kind of what? Jason Hiner: For a billion developers. Mario Rodriguez: Exactly. That's kind of what we thought was going to happen. And it's just accelerated at a rate that maybe that happens within a year and a half or two. Jason Hiner: Because you're like at a quarter of that now, right? Mario Rodriguez: You're like at almost 20, at least publicly what we share is 225 million plus. Jason Hiner: Okay. Mario Rodriguez: But in the Octoverse report, if you see it, it says over the next two to three years, we could be six X that and that would be over a billion. Jason Hiner: That would be like one and a half billion. Mario Rodriguez: Yeah. And then so the curve goes six, [unclear], three, and then one point something. I think that's kind of the parameters. But even at two will be half a billion overall. We're still getting today for the second year in a row, one developer, one builder, every second signs up. And we made it incredibly hard to sign up now, by the way. Jason Hiner: Really? Mario Rodriguez: Yeah, because there's so many bots and accounts. Jason Hiner: Right. Mario Rodriguez: That I mean, that 225, I could have it to be 10 billion. Like if I'm going to include every single agent trying to do things. So we make it like really hard. And I get DMs constantly. It's like, "I can't get through the CAPTCHA." I can't do this. And like, we don't, you're not accepting this email address. I'm like, yeah, like we got to make it this hard nowadays. So these are people. Like, people people. Jason Hiner: Actual humans. Mario Rodriguez: Yes. So for us, is that like we were already thinking about it. Like, in fact, I've been talking more about it. John Maeda, who is the head of design, and I talked probably at the beginning of the year. He was telling me something that MIT used to say was just lower the floor, raise the ceiling, and then widen the walls. And I think that's very powerful. And started kind of talking about this in May. I went to a couple of podcasts where I talked about lowering the floor. So that means you get more and more people coming in. Which is exactly what is happening in creating, because now they could create through natural language, right? And now the tools and the models are so good on that, that everyone can build. And that's beautiful to us because what ends up happening, and I don't think we talk... I mean, I definitely do, but I don't think as an industry... We're not talking enough about this, which is you want the floor to be lower because more and more people that come in, you increase the likelihood then of a set of those people increasing their craft and becoming them professionals. And when you do that, then you raise the ceiling, essentially. So you end up then having these people having an outsized leverage effect on what creation is. So I guarantee you that when we solve cancer, there's going to be a series of developers that created tools that were given to the scientists for them to solve, right? Like it is going to be powered by software. And those people creating those tools are going to be professional developers. Maybe some of them just got in right now into this and then increased their craft very rapidly to become them professionals, right? Jason Hiner: Yeah. Mario Rodriguez: And that's the funnel that we want, because that's really how humanity goes and advances at a quick pace. Jason Hiner: OK. Mario Rodriguez: And as you get then people in, like Albert Einstein

in physics as an example, and then eventually they become great at it. And then you can actually [unclear]. So like that's what we're trying to do. This interconnected community is so we all learn from each other. So you lower the floor of what you could pick up and remix. Then you become really interested in that problem. Then you become really good at it. When you started doing this, you know, you were probably not as good as right now. Jason Hiner: For sure. Mario Rodriguez: So every single day you increase your craft and the technology then helps you as you increase in your craft to create better and better and better. And then you could reach more and more people. And then that creates then the effect. And that's one. That's what we call raising the ceiling. Jason Hiner: Right. OK. Mario Rodriguez: So someone lowered the floor for you to do all of this. So you eventually can raise that ceiling on it. And not everyone makes it. Not everyone does this, right? And like they might try a couple of things and then kind of give up and go to something else, but a few of them do. And that's the effect that you want to have. So like that is what I want people to think of. You have this. Yes, we're going to lower the floor. Yes, we're going to provide tools. We're going to connect the people. We're also going to spend a fair amount raising the ceiling as well. This is why I said in my talk, like developers are not going away, and whoever says that, F that. Like, you know, open source is not going away Jason Hiner: Yeah. Mario Rodriguez: Either. In fact, it's going to be even stronger. Jason Hiner: Yeah. OK. Last question. Mario Rodriguez: Yes. Jason Hiner: I have to ask. There's so many tools to let you build. Mario Rodriguez: Yeah. Jason Hiner: Right? Claude Code, Codex, Lovable, Replit, you know, Cursor, all of them. Why would you encourage people to consider starting with Copilot if they really want to learn this and they really want to eventually sort of raise the ceiling? Mario Rodriguez: Yeah. You know, the way that I think about it, and this might not be the answer you were expecting or looking for. I think more about it as the GitHub platform rather than a point product. Jason Hiner: OK. Mario Rodriguez: Copilot. Jason Hiner: OK. Mario Rodriguez: And the way that I think about it, what do I want in that GitHub platform? Well, I want developer choice and I want model choice. Right. So if you want model choice and Fusion is part of that model choice, that's a really big draw for people. So you don't have to then only utilize one of the families with Fusion. You could even get the same intelligence at 70% of the price as well. So that's kind of one. The second thing is we also talk about multi-surface as well. And that multi-surface really is important. So you could go from mobile all the way to GitHub.com, all the way to VS Code, all the way to Slack and Teams. And that's also really important too. And as more and more of what we're doing with the squads and automations and that intelligence layer happens. Then those surfaces and then the capabilities of it become better and better and better. So that's kind of number two. And then number three is, hey, you're kind of building with others. And there's a set of capabilities from our actions ecosystem, from our, you know, the plugins ecosystem, all the way to things like our sandboxes and cloud providers, like that provides a lot more choice. And it allows us to vertically integrate a set of experiences that end up, you know, we end up being chosen, you know, depending on the individual or the enterprise overall. Jason Hiner: Very good. Mario, thanks for the time. Can't wait to hear what you're up to next year at GitHub Universe with as fast as everything is moving. So appreciate you being on the show. Mario Rodriguez: Thank you.