I did the following:
- Open an obsidian vault (= just markdown files) in cursor, get it indexed.
- Ask questions on my notes.
- Refactor my notes.
I find issues in part 2 & 3. The prompting of cursor AI agents is heavily tuned for
suggesting code edits
vs answering a question. See issue for markdown: Cursor AI gets confused for markdown edits
Secondly, I want to point out that there is an opportunity here. The rest of the world is building RAG-like-agents for answering questions on dedicated knowledge-bases / documents / memories
etc. If a set of markdown files (=obsidian vault) are treated as first class citizen; cursor AI can very easily solve this use case of becoming Super Obsidian
and can also nudge people to migrate!!