2025-08-26 –, Potter Auditorium - Kenneth C Rowe Management Building
Developers dread debugging failures in CI—but what if they didn’t have to? In this session, we’ll dive into strategies for building self-healing CI that’s trustworthy, reliable, and cost-effective.
Have you ever dreamt of a CI that's always green? With DevOps principles, we've built the tools and processes to enable developers to run as much validation as possible before pushing to CI, and yet the errors still slip through.
In this session, we'll explore a completely self-healing CI system that is trustworthy, reliable, and cost-effective.
A trustworthy system is one that we believe will produce correct output. For an AI Agent, it means it cannot ever give us what we've all come to know as "AI slop".
A reliable system is one that we know will consistently fix the problems in CI. The success rate needs to be so high that we're not still spending almost as much time debugging CI failures as we did without the agent.
A cost-effective system means we're not sending our entire budget to one of the big AI companies. We must be able to self-host a small and specialized LLM capable of being trustworthy and reliable.
With these three tenets achieved, we can eliminate wasted developer time and live the dream of an always green CI!
Kyle is a member of the ecosystem team at Dagger, helping shape the future of CI/CD. With a background in DevOps across companies of all sizes, he enjoys sharing insights from both his successes and failures.