GitHub Unveils Copilot Coding Agent at Build 2025

GitHub has introduced a cloud-based coding agent that can draft and iterate on pull requests directly within GitHub, leveraging an infrastructure similar to GitHub Actions. The agent can handle tasks autonomously, creating branches, iterating on PRs based on code review comments, and updating commits until the work is accepted. Developers can assign tasks directly via GitHub or VSCode, and the agent can even grant access to external tools and data by configuring servers directly in the repository settings. The community has reacted with a mix of excitement, curiosity, and healthy skepticism, praising its time-saving capabilities and expressing concerns about the choice of when/if to merge. Some users fear that offloading coding to AI might devolve into janitorial oversight, while others believe it will create new and exciting opportunities for junior devs. Security and policy questions have also been raised, such as the potential exposure of sensitive information. GitHub is evolving into an AI-enhanced development platform, abstracting away grunt work and enabling developers to focus on higher-level logic. The Copilot agent can lead to new roles like "AI orchestration engineer" and new skills like prompting and supervising AI in coding. To prepare for this AI-centric future, developers should focus on writing clear, detailed task descriptions and automated tests to build trust with AI agents. The end-game is to become comfortable letting AI draft code and guiding/refining it, and validating it with tests.