Developer Advocacy and Software Development with Sven Peters at Atlassian Team ’24 Europe

This is Techron tv. Hey everybody. Mitch Ashley here in Barcelona at Atlassian, team 24 Europe.

I have some more great conversations. One of my most favorite thing is to talk about developers, talk with developers, talk about development, and I'm joined by Peters, who was a developer advocate with Atlassian. You've been a developer advocate for quite a while.

How long have you been with Atlassian? Um, I started Atlassian in 2011 actually. So Wow.

I was actually the first developer advocate for Atlassian. We're a small company at that time. Yeah.

Um, that's like 300 people. Um, and then it expanded from there. Who knew just what, what happened.

Right. That's great. Well, listen.

Um, where did you come from before you were at Atlassian? Yeah, so I was developing software. I was a software developer.

I was leading a team. Um, so I knew what, how developers tick, um, and, uh, that that was also required of the role, just like to talk to developers. Um, I was not really a presenter on stage.

Um, I learned that along the way, uh, with a job, but I enjoyed talking to developers and inspiring developers. And, uh, that's the role of a developer advocate. Just to, to just like talk to developers to explain them the technology behind that.

Um, and not to sell something but sell a philosophy of how you can just like develop software better. That's basically what I do. Yeah, You were early Pioneer is a dev.

We call it dev rail now, right? Exactly. Exactly.

Yeah. Yeah. Well talk a little bit about, so what are developers excited about with what's happening with Atlassian?

So I think the develops are always excited about just like new technology that comes in, right? They just like jumping on everything that is, that is new and exciting. Um, and right now it's definitely ai, um, everything that has to do with ai, um, lms, how you can just like write your own, uh, app with using, using, uh, the LMS that are available, um, but also how you just like can improve your developer experience with ai.

Um, the whole like coding assistant. Um, that's a, that's a big topic in, in software development right now. How can we be more effective, um, in, in software development?

How can we just like get, get more out of our time that we have that is limited because our backlogs are full of items, right? Um, everyone has a full backlog. So how can we just like be faster in, uh, just like delivering software to our customers?

Tell you what, if you and I created a, an AI that addressed technical debt, every developer would love us. Yeah. That'd be the best AI ever.

Yeah, so exactly right. I think everyone is a little bit afraid of just like the AI is taking over the developer's job, but what it really is, is the AI is just like helping to make this job more joyful. Um, as you said, just like if I have an AI agent that can just like look at my code and get the technical depth out of it, right?

Right away, this is like fantastic. I don't wanna do that as a developer. I just like wanna create new new code.

I want to just like, think about just like what architecture can I just like build AI isn't there yet? Yeah. Um, that's still what the developers are doing.

Um, and then at the same time, you can care more about the products that you are creating for the customer. So, um, you don't have to just like write boilerplate code anymore. The AI can just like take that over.

So all this boring part of this all annoying part of software development, that's what ai, AI is doing. Um, so we, yesterday we announced a new, uh, AI agent that can just like do code reviews for you. Um, so just like I'm a senior developer and I just like don't want to just like say, oh, I see that immediately there's an infinitive loop.

Uh, there's exception that's a security problem if you just like write it this way, the AI can just like find, find it out as fast as I can. Um, so I can just like take care more about just like, is that the right architecture in that, in that, in that, uh, code change or should, should, should we change something? So also there, the AI is helping me, uh, with the boring part of things, just like taking that over, just like adding this full request comments directly at the pull request.

And I can then take a glance at that once this, this just like iron out. Um, I can just like take a glance at that and then say, okay, yeah, now these are the important things that we probably need to change or everything is fine. Um, I never have time for code reviews.

Um, and sometimes I, I must admit, I also say just like, looks good to me, right? But I'm gonna go home. It's, it's just like I'm, I'm, I'm tired.

I just like want to get it done. So I say it just like looks good to me, but the AI is never tired, right? So the AI can just like do, do much better job at, at these things, um, and can help us to really tackle the hard problems.

Mm-Hmm. Well it's, it's the, uh, we talk toil, right? Take the toil out of the job.

Yeah. Right. Exactly.

Exactly. Well, how is, um, you know, one of the things I'm really fascinated by is Atlassian's approach is the agents are actually a user. They're a person, not a person, but they're like a user in the system.

You can assign work to it, give part a workflow, it can assign work back to people pretty unique. It's not just an a method of body automation. So I'm really fascinated by how do developers think, what are that approach?

So yeah, e exactly. So these are just like, um, I would call it assistant, right? So they, they're just like your specialist and helping you just like, if you have a database question, you can just like ask an agent for that and train that agent, especially on how you use that specific database in your organization, in your product with your architecture and train the agent for that.

Um, and you don't have to reach out maybe to the database expert on your, on your team or maybe in another team. Um, you can just like ask the agent just like first if you, if you don't, can't, can't, uh, move on or just like you're stuck into something, you can just like still reach out to the human. But I can just like be faster for me as a, as a, as a developer, I can just like get faster to my, to the results that I need or the, the knowledge that I actually need.

At the same time, I'm not interrupting my database expert in another team all the time and just like have, have some questions. So having this virtual teammates, how we call it in your, in your team is just that that's, I think is just like one, one aspect of the agents that we have. Another aspects are these, uh, automated, uh, stuff that just like the, the easy stuff, as I said, just like the pull requests, uh, comments.

I can, I can have an agent for that. Um, we have a feature flag removal agent, for example, at Atlassian. Okay.

Where the code is sometimes full of feature flags that are Yeah, it's enabled, right? So we can actually remove the feature flags from that code, but no one's doing that. It's still there, right?

It's a, it's an extra task for soft deliverable take on, um, what the AI can do. So what you do is just, like you say, remove the feature flags from from that code, right? But it's done.

You can remove that, um, if they, if if the suture flag is enabled and then you get a pull request and you can just like still review it. And I think this is also something that is important still, that we need to trust, trust the bots, um, somehow. So the human is, should be in every step along the way, should be in control and the eye should just like be a helper and it suggests things.

Um, but the human should still be in control. And that's the same thing, like creating a poor request. I can look at it, I can still make changes, uh, to the code if I'm not, not, uh, happy with what the AI suggest me at at code changes.

Um, and then I can merge the code back to the, to the main branch and go on with, with software Development. That trust is kinda, you develop it just like between people Yeah. To use something, a tool ai Exactly.

To Learn how it works, what you trust. How to trust it. Exactly.

Exactly. Yeah. And then also we have built this new product called vo, where you can build your own AI agent, right?

So your own VO agent. Um, and this is just like fascinating right now, just like in, so, so I compare it sometimes with, you know, when the iPhone came to market, everyone was just like super excited about the iPhone, right? You can't just like touch it.

Um, but the AppSec on the iPhone were just like limited. Um, and I couldn't have imagined at that time that we, I would use it for payments, right? It's a credit card is on, on, on my phone right now, uh, and I can just like have my, my, my flight ticket on my phone.

And I think we're at the same time now with AI where we just like say, oh, fascinating technology, but we have no idea what we can do with it. So we built this robo agents where you can just like build your own robo agents and try things out, right? And then we can test things out.

Um, the developers can test things out with those agents. What's, what's really working okay, how would, how do you want to use it? Uh, and then find the, the, the use cases that AI can is, is really, really good at.

And then just like we can just like take that and implement that directly into the, into the product later on. Um, so I, and, and it's, it's like you being an app developer as you, you wanna just like have an app on your iPhone, but you can't because you're not a programmer, you can't do that. But with the robo agents, there's no code.

So you can just like use it as no code program this with no code, or you can also just like use, uh, use them with uh, program them with code. But it's so easy now building your own own AI agent was with robo, You know, so there's is, there's two types of agents, right? For there's no code and then there's, I think it's auto dev, is that what the Yeah, auto dev is a, yeah.

What's, what's that? So auto dev is an agent that comes out of the box. So auto dev we just like announced yesterday.

And um, that's actually, if you go, go from an Jira issue, uh, and then you can like create directly from the issue description can create directly a code plan. What do I need to change? Um, if I just like wanna have a X Excel export, just like I need to have a button, so just like, so people can press it, I need to have a file dialogue, probably I need to have, uh, some, some database connection just to stream the data or I need to need to save it somewhere.

So, uh, or export it as Excel, right? So I need that. Um, and it, it creates you, the code plan for it can still edit that.

As I said, the human is always on the loop and can always just like make changes. Um, and then the next step is actually you can from that code plan, and if you're happy with that code plan, you can create code from it and it shows you the code still. You can just like ask the AI to just like do something differently.

Um, and then from that code you can just like create directly pull, we create directly from pull request directly from Jira. Um, so that's also just like for some tasks it just like shortened the, the development time. Um, and uh, I create value very, very fast with, with auto dev, for example.

So as a developer, well, I was talking with Mike, your CEO yesterday in a, in a meeting and he was really emphasizing the architecture of what Atlassian is built. And that's why you can come out with, you know, the teamwork graph and the, the AI service that supports your AppSec. How do you see the architecture of Atlassian?

Yeah, I I think we have this whole like knowledge everywhere, right? We have it in Confluence, we have it in Jira, um, and then makes so much sense that a human, it's, it's just like way too complex for a human to, to just like understand everything around it and just like know everything, right? But that's actually what, what artificial intelligence is good for, right?

They just like, they can just like make the connections between those different elements that you have in your company can be pull request, it can be Jira issues, can be Confluence pages. Um, and we are integrating with a lot of other tools too. So you can have Word documents, uh, Google Docs and out of this knowledge that that, that is inside your systems, um, we are creating this, this teamwork graph, this this architecture where everything runs on and the eye can just like make, make fast connections between those.

It knows. Um, when I, if I have questions around a project, it knows that I'm working on that team. Uh, so it can just like give me relevant information for me.

Uh, and if someone else ask it, ask us the ai, it just like gives, gives information about that. Uh, imagine you're just like, uh, working, working in a big organization and you have different CI ICD configurations or different use Sure. At CI CI/CD, uh, pipeline that I use differently.

If I'd ask the ai, um, tell me how, how, how do I set up this ICD part? It knows exactly I'm working on that pro project, I'm working for that product. It just like gives me relevant information for me.

So, and that, that's because of the teamwork graph, because it knows me, it knows which projects I'm, I'm, I'm working on it knows just like the, the exact CI/CD pipeline configuration for this project. So it can give me the relevant information that is relevant for me and I don't have to search through the whole acknowledge base to just like find the right information. Very good span.

It's been a pleasure talking with you. Thanks. Thank you so much us the developer perspective as you do so well.

And congrats on the announcements and it looks like you're having fun. Thanks for having me. You Bet.

You bet. We'll be back with more in a moment.