As a developer, I rely on AI-powered tools like Cursor AI to speed up my workflow. While it offers great coding assistance, one major issue disrupts the development process—lack of persistent project memory.
Every time the model changes, it forgets the development goals, project structure, and past interactions. This means:
Constantly re-explaining context and project requirements.
Losing track of previously implemented solutions.
Re-training new models from scratch, adding unnecessary friction.
For Cursor AI to be truly developer-friendly, it must: Maintain long-term project memory across sessions. Seamlessly adapt to model updates without retraining. Provide better recall of past decisions to ensure continuity.
These improvements would make AI-powered development more efficient and less frustrating. Who else is facing similar issues? Let’s discuss!
I think this an bug because without this feature cursor ai is not complete. just assume you explain all thing to one employe in few months and then he leave you and another emplye come then again need to train! lol!!
I have hundreds of such files but it depends on the project. Some are simpler, then few files are enough. Some are complex and big, then many files are necessary.
I put core requirements in rule files and features in MD files.
Im sure that currently so many companies are training their staff how to use cursor.
As with any software the evolves and changes they will need to update the staff on changes unless the staff check by themselves.
But also AI models change so much over a few months. So prompts that worked on old models dont work so well now, where new prompts work better than ever.
Maintain long-term project memory across sessions.