Different AI tools require different prompting strategies. Here are best practices for the three major domains you use every day.

9A. Chat AI — Claude, ChatGPT, Gemini

These are your general-purpose thinking partners. The techniques in Sections 1–8 apply here most directly. Here are the domain-specific additions:

Best Practices

• Always set the context first. Even one line: "I’m a 22-year-old CS student preparing for a campus placement interview." This alone improves output quality significantly.
• Use conversation continuity. Do not start a new chat for every follow-up. Continue the thread — the model remembers earlier context within the same conversation.
• Ask for formats explicitly. If you want a table, say table. If you want spoken language, say spoken. If you want bullet points under 10 words each, say exactly that.
• Use the model as a sparring partner. Share your draft argument or plan and say: "Poke holes in this. Be critical, not encouraging."
• For factual or current topics, verify independently. Chat models can hallucinate confidently. Use web search tools when they are available, or verify claims that matter.

Common Mistakes

• Asking one huge compound question — break it into a chain instead.
• Accepting the first output on anything important without a single iteration.
• Using vague quality words like "better" or "deeper" without defining what that means.
• Not specifying the audience — "explain this" is always worse than "explain this to someone who knows X but not Y."

Power Move for Chat AI

At the start of a complex session, say:

<aside> ⚡

"Before we begin: ask me any clarifying questions you need to give me the most useful response. Don’t assume. Ask first."

</aside>

This inverts the dynamic — instead of you guessing what to include, the model tells you what it needs.

9B. Code AI — Cursor, Windsurf, GitHub Copilot