Open Source Summit NA 2025 | David Nalley & Antje Barth, AWS

Clips
News
More from Open Source Summit NA 2025

David Nalley

Governing Board Chair

AAIF

Antje Barth

Principal Developer Advocate

AWS

Code, community and the future: 13 takeaways from Open Source Summit NA

Beyond uniting developers under shared ideologies, goals and pain points, Open Source Summit North America ignited conversations at the intersection of code, community and change.TheCUBE goes live with insights at Open Source Summit NA

play_circle_outline Introduction of speakers Paul, Antje, and David at Open Source Summit North America 2025.

play_circle_outline Advancing Communication Standards: The Linux Foundation's A to A Agent Project and Its Impact on Industry Collaboration

play_circle_outline Empowering Organizations: Embrace Open-Source Tools and Experiment with Agentic AI for Future-Ready Technology Careers

Info
Transcript

David Nalley & Antje Barth, AWS

David Nalley

Governing Board Chair AAIF

Antje Barth

Principal Developer Advocate AWS

In this Open Source Summit North America 2025 interview, David Nalley, head of developer experience at AWS, and Antje Barth, principal developer advocate at AWS, join theCUBE Research’s Paul Nashawaty to explore how open source and agentic AI are converging to shape the future of application development. The conversation dives deep into recent ecosystem shifts, including the Linux Foundation’s A to A Agent Project and the rise of multi-agent systems that demand new communication standards and protocols. Nalley and Barth break down the challenges and opportuni... Read more

explore Keep Exploring

What are the names and roles of the individuals participating in the Open Source Summit North America 2025? add

What was announced by the Linux Foundation and what implications does it have for technology communication standards? add

What steps should individuals take to get started in open source related to agent development? add

bolt Powered by CUBE AI

David Nalley & Antje Barth, AWS

search

Paul Nashawaty

>> Hello and welcome to Open Source Summit North America 2025. My name is Paul Nashawaty and I'm covering the AppDev space and all things open source. I'm joined by Antje and David. How you doing?

Antje Barth

>> Excellent.

David Nalley

>> Doing great.

Antje Barth

>> Thanks for having us.

Paul Nashawaty

>> Thanks for joining me today on the desk here. It's really, really great to have you here. Why don't you introduce yourself to the audience and tell everybody what you do?

Antje Barth

>> Yeah, Antje Barth. I'm a principal developer advocate at AWS, and working with developers mostly around agentic AI topics.

Paul Nashawaty

>> Awesome. Great. And there's not a lack of work there, I'm sure.

Antje Barth

>> Absolutely.

Paul Nashawaty

>> David?

David Nalley

>> Hi, I'm David Nalley. I lead developer experience at AWS.

Paul Nashawaty

>> Thank you for being on. Again, thanks. This is awesome. So there's a lot of announcements that are coming out today. When we look at the growth of the ecosystem, especially when we look at agentic AI and we look at the applications and how applications are growing, there's a lot of collection of information that occurs. Whether it's using heritage data or using new applications in the environment, there's different ways for technologies in order to connect together. One of the announcements that came out was the A to A agent project that the Linux Foundation was announcing today. That was pretty exciting because it talks about the expansion and growth that you see with regards to connecting different technology stacks together. I guess let's start here. David, maybe I'll start with you. What are your thoughts on just this announcement and what it means to the environment?

David Nalley

>> Well, I think it's interesting that we're seeing this announcement come out because I think it does signal that we're going to have to push closer to a de facto standard. That doesn't mean that it's the only way that agents are going to communicate, but I think that there's a recognition in the industry of the fact that we're going to need some protocol, some single way of communicating. And again, we'll probably end up with multiple of those in the long term, but I think this is a strong signal that we're starting to see some coalescing from the industry around ways for agents to communicate to each other.

Paul Nashawaty

>> Absolutely, absolutely. I think, Antje, when I talk to organizations and I talk about this growth of bringing applications together, in my practice I talk about past, present, and future. And as we know, AI is happening fast. Things are changing real fast. So I always get conversations around, do we use MCP? Do we use A to A? Do we use just APIs in order to connect things together? What are your thoughts around this, and just the growth of how this is happening so fast?

Antje Barth

>> Right, and absolutely. The growth and the speed of innovation in AI is just mind-blowing, right?

Paul Nashawaty

>> Right.

Antje Barth

>> Just a year ago or two years ago, you had to write so much code around to even create your agents and give it a structure. But with the reasoning models being so capable, so powerful these days, there is much more capability to have them go through the user requests and then plan, reason, and everything. And the next step is then really kind of give those agents access, as you said, to tools. And the tools might be, depending on the company you're working in, access to specific APIs. You want to give it maybe the ability to call to get more to perform an action, but also maybe connecting to your company data sources to get more context-aware information about specific things to answer user requests. So you have this tooling side, which there are protocols evolving in the net space. And then obviously as you're building out your system of agents, you're likely going to have that case where you have not just one agent, but then you're suddenly in this multi-agent world where you have different specialized agents handling specific tasks and there's the need to connect them. And then if you look into the future part of this, where this ends up is really kind of this world where you have agents performing a lot of specialized tasks and you need standards and protocols to really build those systems to make sure they communicate in a efficient way.

Paul Nashawaty

>> Yeah, that absolutely makes a lot of sense. I mean, when we think about, David, you were talking about the adoption of protocols. And when we look at that agent to tool approach that you were just referring to, right now I know that I talk to CIOs, and CIOs tell me, look, they have a budget. Historically they just had an IT budget. This is what they're going to spend their money on. Right now there's a carved out piece of the budget for AI. And what I'm seeing a lot is organizations going, "All right, well we just need to have some type of AI action. Let's just spend money on something that's AI." But the reality of it is they're trying to bridge the gap between old and new. Because what I hear the number one challenge that organizations run into is to modernize their applications, and they need to overcome things like complexity and skill gap issues. And those are big challenges, especially with the explosion of growth in the AI ecosystem. There's so much happening so fast. So what are your thoughts around this explosion of AI and the ecosystem?

David Nalley

>> I think that's a natural progression where people see a lot of opportunity. They get excited about a technology. They want to build atop that. I think people are having a lot of fun building things, both with these tools and building the tools themselves. And I think that excitement is driving a lot of interest and it's driving a lot of noise and-

Paul Nashawaty

>> Oh, and innovation....

David Nalley

>> well, I mean, we haven't completely maxed out on innovation yet. Still a lot of room for that. But right now people are excited because they're starting to see the potential. And so I think when I look at what people are doing, I think that they're focused right now on what can the technology do. I'm not sure that we've gotten to the point where we've identified the whole list of problems that are going to be solved.

Paul Nashawaty

>> Sure.

David Nalley

>> But I see people poking around looking at how do we... Agents weren't an idea materially a couple of years ago. Then we had agents and we built agents to tools and some connectivity for that. And then the idea of agents having to talk to each other, it's kind of that next generation. The space is moving really fast though.

Paul Nashawaty

>> It sure is. It sure is. And when we look at this, I like what he was talking about with the experimental, the code. And from a developer perspective, I'm thinking vibe coding, right? You're going to try some new things. You're going to try to make things work and see what sticks. And if it works, great. But the harmonization of connectivity across the ecosystem, the thing that a lot of organizations are honestly scared to death about is the governance, compliance, and regulation. You got to make sure that once you have these... One thing was like APIs talking to data sets, and that's controllable to some extent. Then you have humans connecting in, and that's somewhat controllable, you know, identity management and machine management. But then you have agents going and talking to other agents. What does that mean from a developer's perspective?

Antje Barth

>> Right. It's definitely a big topic, and obviously we need to build agents in a secure and responsible way. There's a couple of things I think that can help on this journey. First of all, we want to make it really easy for developers to start building agents. A few weeks ago we released Strands Agents SDK and open source SDK that takes a model driven approach. Just connecting to what I said earlier about the models are being so capable of figuring out this agentic loop and which tools to call. But then the other part to that is validating they're doing the right thing, to your point. And so Strand's agents also includes capabilities to really implement your specific validation path so you have that control and can ensure the agent is acting as you want it to act. And then as we go along, this will obviously be a big topic that we are all as a community contribute towards.

Paul Nashawaty

>> Sure, sure. Where does somebody get started on that in this new announcement, Strand Agent?

Antje Barth

>> Yeah, absolutely. Join us on GitHub. If you look up Strands Agents SDK, it's an open source project. We do appreciate the community coming together and developing this with us as well.

Paul Nashawaty

>> That's really cool.

David Nalley

>> I've actually been really impressed with the Strands community in particular because every week new people that are not employed by AWS are showing up and contributing to that. That's really exciting. It does show me that there is some excitement and people are willing to make that kind of investment, which is not a trivial investment today when you've got limited places where you can spend your time and invest your contributions in.

Paul Nashawaty

>> Time is incredibly valuable, especially when you have everything. Especially from a developer's perspective, everything's shifting left, and the developers in there with the catcher's mitt catching everything that comes their way. So it's kind of crazy. But let's talk about use cases, because I think use cases is really whether rubber hits the road. Let's talk a little bit about that. We were talking before, before we got online here. So tell me a little bit about the use cases.

Antje Barth

>> Absolutely. We see a lot of use cases starting really kind of automating business processes, workflows internally, kind of where you see efficiencies really playing out quickly, like processing large amounts of data, optimizing specific internal tasks. And then the next step is then also once you have those multiple agents connected, I think this is kind of the next level where you have the agents reasoning through more complex things pulling from different areas of the business. For example, you could use that also for driving research. I think one important part which is exciting right now is can we employ agents to do research for us? Like looking into different aspects of the business, find maybe synergies in the business. We see this in the academic research space, but I think there's a huge potential also in a company to kind of see where agents can help identify those.

Paul Nashawaty

>> Yeah, I think that's really, really exciting. Right now I'm a huge fan of human in the loop at this point. I don't think we're ready to just kind of let agents just go out in the wild and do what they want to do. I think we will get there. I think we'll get there, especially when it comes to the tedious tasks and things that are kind of draining from developers and such. It's like, "Let the agents take over that." I'll throw out to both of you on this one. We talk about agentic AI and the expansion into the ecosystem. What does this mean for where you're coming from? Let's start, David, with you.

David Nalley

>> Here's what I think. I think we're far from the end of the innovation that's going to happen. But when I look at how agents are changing, what I really see is really tight specific focus on specific problems and being able to really narrow down and focus on those problems specifically and then hand it off to something else. We have this idea of humans being really good generalists. I think that models have done an okay job with general knowledge, but this idea of being able to narrow the focus down to one problem set and to do that really well is really exciting. What I'm hoping for is that we see lots of these agents working in parallel on disparate problems and then working collectively to drive outputs and outcomes that are much faster than what we could do even if we were assembling a team.

Paul Nashawaty

>> Yeah. Well, I'm seeing in theCUBE research that I'm doing that I'm running is 67% of respondents indicated that they're looking to hire generalists over specialists. Not necessarily because they want to hire generalists, but they can't find specialists, so they end up hiring generalists. So I think that the agentic, the AI approach is really going to allow for the generalists to do more of what a specialist has traditionally done. Antje, what are your thoughts on agentic AI and the ecosystem?

Antje Barth

>> I think it will really kind of augment, and that's what I'm hoping for, augment our abilities, our skills. So I think it's really critical for everyone to kind of start exploring the area if you haven't done so already. Because in the future, we will see agentic AI and agents helping us in the jobs, helping us augment, being more efficient in our jobs. So it's an interesting topic, and it's not just right now. We started speaking about generative AI, helping maybe writing things, generating images and things. But in this new agentic AI world, it's doing much more for us. It's kind of maybe even becoming a personal assistant to us where we can say, "Hey, help me do this task." And then agents, as David pointed out, we see a lot of interest in these asynchronous agents that maybe they get a task and they go off and research something, and then help us ideate on projects. Maybe that's a thing that I can see coming together where I say, "Okay, I have this task and now let me help with a couple of agents." Like, "Let's do some research here. Let's explore there." And they go off, run, and then at one point come back. And then I have this information that I need to complete my tasks. So I think this is asynchronous model we'll see a lot of coming.

Paul Nashawaty

>> Yeah, that makes a lot of sense. I really like the approach that you put there on how to get that in motion. It's really meeting customers, clients, whatever organizations wherever they are in their journey. Because if they're early in their journey and they don't know how to get started, they can start without dipping their toe in the water. But if they're really mature, they can take on more mature projects to get going, which I think is incredibly important. David, position on open source? Organizations starting out and getting going, what's your thoughts?

David Nalley

>> Getting started on open source, or?

Paul Nashawaty

>> Yeah, getting started in open source in relation to the conversation around the agents?

David Nalley

>> First of all, Antje just said get started now. I think one of the powerful things that people can do is go build an agent. Figure out how to connect it to your data sources. Be experimenting with things like MCP, A to A. Figure out how to connect these agents to start doing work. I think that's vitally important. I think you need to be experimenting with these tools, especially if you plan to continue to having a career in technology. I think you ignore what's happening in AI and specifically agentic AI at your peril. One of the reasons we released Strands that Antje was talking about earlier was we needed to speed up the ability for people inside Amazon to be able to create agents. So I would personally, and I know I'm biased, I would personally encourage people to start with Strands. Figure out how to use Strands to use MCP to connect to different tools. Maybe you have A2A to use those agents talking to each other. But these are tools that whether it's Anthropic or Google or AWS have released to make agentic AI a lot more approachable and a lot more simple to get started.

Paul Nashawaty

>> Absolutely. Well, I really appreciate both of you joining me today. The insights were very valuable. I think the audience here that's going to watch this and understand this, they're really going to have to get their head around what approaches they're going to take. I think you've clarified a number of ways, maybe also opened up a number of questions for people to think about, which is good. But I think it's really great. I really appreciate you joining me today on today's session. I want to thank you for watching our session today. My name is Paul Nashawaty. I'm coming to you live from the show floor at Open Source Summit North America 2025. Thank you for watching theCUBE, the leading source for tech news.