Topic
Copilot
4 episodes
-
Building the Agent Host Protocol for sessions that continue across devices
Jessica and Cathy unpack GitHub's Agent Host Protocol as a session-continuity layer for long-running coding agents, focusing on the practical shift from editor-bound extensions to durable hosts that multiple clients can attach to.
-
Early Adoption of Agentic Coding Tools by GitHub Projects
Asteria and Draco analyze a GitHub study of 25,264 agentic pull requests across 2,361 repositories, finding that while agentic coding tools are spreading, real adoption remains thin: median repos generate only 1-2 agent PRs per quarter, small teams use them more intensively than large ones, and nearly all projects rely on a single human reviewer to oversee agent work. The paper reveals adoption is infrastructure-constrained, not capability-constrained — the bottleneck is human oversight capacity and project governance, not what the agents can generate.
-
(a) Macro Level average performance profiling.
Vince and Ava dig into SkillOpt-Lite, a paper arguing that skill optimization for agents can be simplified into a minimal pipeline built from trajectory exploration, consensus mining, and independent validation. They focus on what problem it solves, why the authors think the extra machinery in prior systems is unnecessary, and where the production story is real versus still a research move.
-
I Built agent Guardrails template to Stop AI Coding Agents From Wrecking Codebases
Lundrog built an open-source framework called agent-guardrails-template to control AI coding agents and prevent them from breaking codebases. The system uses four safety laws, active enforcement via a Go MCP server, and risk-based decision matrices to reduce AI-caused incidents by 78%.