"Somebody is always paying the token bill": AI and Open Source with Brian Proffitt & Jason Brooks

<p>Generated code is arriving faster than maintainers can read it. RedMonk's Steve O'Grady discusses this new reality with Jason Brooks and Brian Proffitt, both managers in Red Hat's Open Source and AI Program Office. Proffitt's teams are adding what he calls human speed bumps upstream: PR templates that ask whether AI wrote the patch, and if so, what the prompt was. The point is transparency, not blocking the flow. Brooks argues for the importance of human decision makers because writing code may cost close to nothing now, but getting another person to accept it still costs plenty, and someone always pays the token bill. The conversation also touches on maintainer burnout, why saying "no" is still a legitimate answer to a pull request, Fedora Hummingbird and its agentically assembled packages, and the thing Proffitt worries about most: trust.</p><p></p><p>This RedMonk video is sponsored by Red Hat.</p><p></p><p>Show notes: <a rel="noopener noreferrer nofollow" href="https://redmonk.com/videos/brian-proffitt-jason-brooks/" target="_blank">https://redmonk.com/videos/<b>brian-proffitt-jason-brooks</b>/</a></p><p></p><p>Chapters</p><p>00:04 Introduction to AI and Open Source Governance</p><p>00:32 Guest Introductions: Jason Brooks and Brian Proffitt</p><p>02:10 AI's Ubiquity and Industry Impact</p><p>03:06 Guardrails for AI Code Quality in Projects</p><p>04:22 Implementing Human Speed Bumps for AI Contributions</p><p>05:17 Encouraging Transparency in AI Contributions</p><p>06:46 Automated Review and Guardrails in Open Source</p><p>07:50 Verifying AI-Generated Code in Fedora</p><p>09:10 Human Oversight in AI Processes</p><p>13:39 Developer Burnout and Managing Contribution Floods</p><p>16:09 Trust and Transparency in AI-Driven Ecosystems</p><p>18:30 Evolving Policies for AI in Fedora</p><p>20:16 Future Challenges and Community Trust in AI</p><p>21:11 Closing Remarks and Future Outlook</p>