I’d like to request a feature that allows thinking models to continuously think after each action, rather than only reasoning once per user input. Some other AI IDEs like Windsurf implement this approach, enabling more dynamic and iterative reasoning.
This would improve problem-solving, multi-step tasks, and complex reasoning by allowing the model to reassess and refine its responses step by step. It could be particularly useful in areas like coding, strategic planning, and detailed content generation.
Would it be possible to implement this kind of continuous reasoning in future versions of Cursor?