Real tactics that experienced AI users figure out through trial and error
Started as a LinkedIn post about the invisible techniques that make AI actually useful.
Most people know how to use AI tools but struggle to get consistent results. It's not about the tools - it's about knowing the techniques that experienced users figure out over time.
- 🔄 When to reset a conversation vs. just nudging the context
- 🗂️ Run the same prompt in multiple tabs to compare responses
- ✍️ Edit and re-run prompts instead of long back-and-forth corrections
- 📋 Use code blocks to keep answers organized (even for non-code)
- ⏩ Type "continue" when AI cuts off mid-response
- 🔢 Break instructions into numbered steps to guide AI logic
- ❗ Add "don't guess" or "cite sources" to reduce hallucinations
- 🤔 Ask AI to critique and improve its own answers
- 💡 Get AI to help you craft better prompts
Got a technique that works consistently? Open an issue or pull request.
See CONTRIBUTING.md for details.
CC BY 4.0 - Use it however you want, just give credit and link back. 🙏
⭐ Star this repo if you find it useful!