Open source projects rarely become under-maintained because the architecture stopped mattering. More often, they drift because maintainers run out of time, funding, or energy.
In this session, William Hill invites us to explore whether AI agents can meaningfully help sustain and evolve a lightly maintained open source system by using Conduit, a Go-based data streaming platform, as the case study. Rather than treating AI as an autonomous replacement for maintainers, he will present a human-governed maintenance workflow that combines context retrieval, task selection, change generation, automated verification, and architectural review before any change is approved.
The session will cover the design of a multi-agent workflow built around key maintenance concerns: understanding repository architecture, selecting appropriate tasks, proposing narrow changes, validating correctness, and reviewing alignment with system boundaries. We will discuss the tradeoffs involved in structuring agent roles, the kinds of guardrails that matter, and where human judgment remains essential.
William will also share a scorecard from the experiment: what kinds of tasks agents can realistically handle today, what failure modes appear in real repositories, and what characteristics make an open source codebase more or less suitable for agent-assisted maintenance.
Attendees will leave with a practical framework for evaluating their own repositories, a reference approach for human-in-the-loop AI maintenance, and a grounded perspective on where AI can help open source teams without replacing maintainership.
About the Speaker
William Hill is a Senior Software Engineer at Zocdoc. He previously worked as a Staff Software Engineer at Meroxa, where he built data streaming solutions and open source tooling in Go. His work spans software architecture, data engineering, AI agents, and developer platforms, with a focus on turning complex systems into practical workflows teams can use. William has spoken on AI agent workflows, data systems, and modern software architecture, and he is especially interested in how open source projects can remain useful and sustainable as maintainer bandwidth shifts over time. He approaches AI workflows as an engineer first, prioritizing reliability and governance over novelty, and combines hands-on experimentation with a grounded view of engineering tradeoffs.
About the Open Source Architect Community
The Open Source Architect (OSA) Community is an invitation-only group for seasoned software architects who are passionate about open source technology. Request to join the OSA Community: https://forms.gle/7efbynVzYhhH2LCQ7
We review each application carefully. If it’s a fit, you’ll get an invite to join the space where it all happens.
For the latest updates on all things open source, follow our public feed on LinkedIn.