🤖 ai neuro
Cursor Meetup Takeaways: Hooks, Debug Mode, and Rules That Work
Cursor Meetup Takeaways
Attended a Cursor meetup and gathered some practical insights on their latest features and best practices.
New Features
| Feature | What It Does |
|---|---|
| Hooks | Pre-prompt checks — can scan for sensitive credentials before sending to LLM |
| Debug Mode | Agent asks to reproduce bugs first, builds a hypothesis list before attempting fixes |
| Marketplace | Share rules, skills, commands, sub-agent types, and hooks with the community |
| Cursor 3 | One IDE for managing multiple projects |
Rules Best Practices
❌ Don’t:
- Write long, complex rules
- Include general best practices the AI already knows
✅ Do:
- Start simple, iterate
- Include company-specific/internal knowledge
- Document edge cases and gotchas not found in public docs
Key Insights
- Developer-written rules > AI-generated walls of text — specific context beats generic instructions
- Hooks = middleware pattern for compliance and security checks
- Debug mode = scientific debugging — hypothesis-driven approach before code changes
Sources
- Cursor meetup (direct from Qepo)