Cursor does not appear to have any awareness of when edits are being applied to Jupyter notebooks. For example, when I ask a question inline via ctrl-K, the suggested result includes all of the contemplation cruft which obviously breaks the code in the cell.
Could you share more info about this issue per Common Issues and Troubleshooting Guide (see Documentation at top right in forum). Cursor team would need more info to check why this is happening.
Only specific AI models that are not from OpenAI/Claude/Googleespond with ‘contemplator’, can it be that you use 3rd party AI API providers?