Developer Experience in the Age of AI

[Music] I'm Bryce and I'm an engineering leader at Clerk. If you don't know what Clerk is, Clerk is an authentication and user management platform. We obsess over developer experience, how to make a billing feel really great to build with. And we spend a lot of time thinking about how to make it easy to get started, integrate, and build delightful experiences. And so today I'm going to share some practical learnings from the field. How developer experience is evolving to solve to serve both humans

and agents and how great DX is still the hallmark of best-in-class tools like Nex.js. So first let's build some shared context. What is developer experience or DX? We use this term all the time but I want to make sure that we're all on the same page for this talk. So naturally, I asked our friend chat. Thought it'd be appropriate to ask them to define developer experience for us. So here it is. Developer experience is the sum of

perceptions, feelings, and outcomes a developer has while learning, using, and maintaining a product, tool, or system in their workflow. It's not a bad definition. Maybe a little wordy, but I like to think that developer experience is how our tools make us feel. Does using a tool spark joy? Does it make us more productive? Or does it keep me in the flow? We all have an intuition for good DX when things just click. DX is

what separates tools we the tools we tolerate from the ones we love. So, what actually makes a good developer experience? I conveniently enough wrote a blog post about this um back in 2020. I've been thinking about developer experience a lot throughout my career. Um, so I define a couple principles that I observed working with tooling, building tooling that make a good developer experience. First one is fast feedback loops, self-explaining tools don't get

in the way, meaningful, actionable output, and quality documentation. All of these things together make up a good developer experience. I would say none of these are really groundbreaking, but they come from practical observations and learnings on working with my team throughout my time in my career. But it's 2025 and humans aren't the only operators anymore. We're now building tools that AI agents are interacting with directly in addition to

humans. So what are agents? A little more shared context building here. Agents. At the simplest, agents are AI powered assistants that write, fix, and reason about code, acting like collaborative teammates in our workflow. Little show of hands here. Who's using agents in their day-to-day development workflow? Sweet. Most of you. Not a surprise. Um but if we can distill it down, agents are a system where you

provide a natural language prompt and the agent uses a large language model under the hood to generate or modify your project directly. As you all have seen, many of you have seen agents have quickly become a critical tool for building software. Thanks to AI and agents, generated software is at our fingertips. We're generating huge amount of code with prompts. agents have rapidly accelerated the pace at which software can be built

and generating a lot of code. Thanks to II agents, we agents can generate code as fast as we can prompt. We're seeing a really interesting shift in the industry with the introduction of agents in AI generated code. We're no longer limited by the speed and efficiency by the speed at which we can actually write code by hand. We're now limited by the speed and efficiency with which we can prompt our agents to write