2025-10-31 –, Auditorium
Developer experience isn’t just for humans anymore. As AI-powered coding agents becomes a part of many developers' "team", they're running into the same friction points that frustrate us: slow iteration cycles, confusing APIs and cryptic error messages.
This talk explores how good DX helps both people and AI agents write better software faster, and some key principles that have turned out to both delight humans and keep agents on track.
So what does “agent-friendly DX” actually look like in practice?
At Modal, we set out to make cloud infrastructure feel like writing ordinary Python scripts: quick to iterate on, easy to test, and intuitive when things go wrong. A guiding principle was to delight developers who use our Python SDK. Along came AI powered coding agents, and it turns out some of our choices made these agents very efficient users of the SDK too!
We’ll walk through a few design choices that turned out to matter just as much for machines as for humans, and share what we’ve learned about building SDKs that work well for both. If you care about APIs, infrastructure, or the future of programming, this talk will give you a fresh lens on what “good DX” means when the developer on the other side isn’t always human.
Elias works as a software developer at Modal (https://modal.com/).
Always looking for ways to make software for other software developers just click.
Original co-author of the Luigi workflow orchestration engine, and many other internal tools for data teams over the years.