The Ruby AI Podcast Innovating Development: The Future of GitHub Agents and AI in Rails
11 snips
Feb 24, 2026 Kinsey Durham Grace, a member of GitHub’s coding agent core team and active Ruby speaker, explains Agent HQ and agent workflows. They discuss AI-generated artwork, using custom agents and Codespaces, making agents modular and transparent, and how observability, intent capture, and repo artifacts shape future developer workflows.
AI Snips
Chapters
Transcript
Episode notes
GitHub Agent HQ Is A Multimodel Platform
- GitHub built an Agent HQ to let users assign issues to different external models like Claude or Codex from the issues view.
- Kinsey describes it as a platform play to host multiple models so users can pick what works for them and their workflow.
Use AI As A Pair Programmer Not A Drop-in Replacement
- Treat AI as a pair programmer and steer it; expect to guide and review most outputs rather than ship them verbatim.
- Kinsey uses Copilot to write tests she dislikes and then inspects and corrects them, saving time while retaining oversight.
Monoliths Hinder Agent Productivity
- Large monolithic codebases create developer experience and deployment pains that hinder agent-driven workflows.
- Kinsey contrasts a small Go service powering the coding agent with the sprawling GitHub monolith where deploying can take a day.