The engine that powers Copilot Chat in Visual Studio just got an upgrade. Microsoft says the new experience can complete more coding tasks despite using the same AI models as before. The updated Copilot experience now uses the same technology that powers GitHub Copilot CLI, the GitHub Copilot app, and VS Code.
Microsoft tested the new and old Copilot experiences by having each perform 157 coding tasks. The new engine resolved up to 21 more tasks, completing 124 tasks compared to 103 for the older experience.
Tests were performed on C# projects across .NET Framework and .NET Core, Azure and Aspire, and WinForms and desktop apps. Microsoft also tested the new engine on C++, Python, and other languages.
Generally, developers don't judge AI by whether it can answer a question. They judge it by whether it actually saves time. If Copilot can fix a bug, review code, find an issue, or complete a task in fewer attempts, it provides real value to developers. Microsoft says that's exactly what's happening.
Developers often spend as much time verifying AI-generated code as they do writing it. If Copilot can complete more tasks correctly on the first attempt, it can save quite a bit of time.
If you've ever used ChatGPT, Copilot, Gemini, or another AI tool, you've likely run into the same problem. AI tools have a habit of sending users down the wrong path. The difference between a useful AI response and an almost useful AI response is often several minutes spent rewriting prompts, correcting mistakes, or double-checking information.
Source link







