GitHub Desktop 3.6 turns worktrees into a control surface for AI coding
Original: GitHub Desktop 3.6: Worktrees and deeper Copilot integration View original →
The friction around AI coding is increasingly less about generating code and more about managing the Git state that follows. GitHub Desktop 3.6 targets that handoff directly: it adds Git worktree support, Copilot-powered commit authoring, and AI-assisted merge conflict resolution inside the desktop client.
Worktrees are the most practical change. Developers can keep multiple branches from the same repository open at once without repeatedly stashing changes, switching branches, or cloning another copy of the repo. That matters more as coding agents run isolated sessions in parallel. A human reviewer still needs a clean way to inspect, compare, and land the output, and Desktop is now giving that workflow a first-class place.
The Copilot layer has also been rebuilt on GitHub’s Copilot SDK. That shared foundation powers the revised commit message workflow and the new merge conflict experience. Every Copilot feature in Desktop now includes a model picker for models available through GitHub, and BYOK support lets users connect a third-party provider or a local model. For enterprises, that is not just convenience; model routing, cost, and data policy are now part of desktop development workflow design.
Commit authoring now reads repository guidance from .github/copilot-instructions.md and AGENTS.md, and it honors commit metadata rules defined for the repository. The intended result is less generic commit text and more output that matches the team’s actual standards. For merge conflicts, Desktop can explain the conflicting changes and suggest a resolution that the developer can review, edit, or accept before completing the merge.
GitHub Desktop 3.6.0 is available for macOS and Windows, with automatic updates rolling out progressively. The important signal is broader than one client release: GitHub is turning the local Git GUI into a control surface for agent-heavy development, where parallel work, model selection, and reviewable AI suggestions become normal Git operations.
Related Articles
GitHub added another enterprise model option to Copilot. Microsoft AI’s in-house coding model MAI-Code-1-Flash is now generally available for Copilot Business and Enterprise, with admin policy controls and usage-based billing.
Why it matters: model launches become more consequential when they land in tools developers already use every day. GitHub says early testing found GPT-5.5 strongest on complex multi-step coding tasks, and the rollout ships with a 7.5x premium request multiplier.
Hacker News did not focus on the headline that plan prices stay flat. The thread zeroed in on a simpler point: on April 27, 2026, GitHub admitted that long agentic coding sessions cannot be subsidized forever, and predictable Copilot costs are giving way to token math.