I’m using Cursor 47.8 on a Windows 11 PC for the first time to leverage my GPU for ML training. I usually use Cursor on a Mac with no issues, but on PC, the AI agents—Claude (3.5/3.7) in this case—seem unable to do anything beyond providing “guidance and explanations” and having conversations.
For reference, on Mac, I can instruct the agent to create a virtual environment and install all required dependencies, and in YOLO mode, it will do exactly that without any manual intervention. However, on Windows, using the same command, Claude only provides text-based instructions on how to complete these tasks myself.
When I asked why it couldn’t execute the commands like it does on Mac, it explicitly stated that it was an AI assistant and could not run commands on any operating system, including Mac. Furthermore, it clarified that while Cursor, as an IDE, might be capable of executing commands, Claude itself is just a language model and cannot.
To push further, I asked, “Aren’t you running inside the Cursor IDE right now?” The response was: “No, I am having this conversation through a chat interface.”
I’m not sure what’s going on. Is the Cursor experience really this much worse on Windows compared to Mac, or am I missing something?
I’ve attached a screenshot of my conversation with Claude for reference.
