What is Copilot Fusion?
GitHub announced Copilot Fusion at GitHub Universe. The system determines which AI model, or combination of models, should handle each step of a coding task. It weighs what the task needs across reasoning, code generation, debugging and tool use.
Agents in the loop
Jay Parikh, executive vice president of CoreAI at Microsoft, said agents start 40% of pull requests. That shows AI agents now hold a meaningful share of code review and merge workflows.
Why it matters
Choosing models per task instead of relying on a single one could let developers benefit from the strengths of different models. GitHub's move signals that multi-model approaches in coding tools are going mainstream.
