Skip to main content

Prompt Engineering

Writing questions for AI models in clever ways to get better answers.

Ask better, get better. No wizardry, just clarity.

Clear beats clever

Describe the output you want. "Summarize this in three bullet points" beats "You are a concise summarizer, now summarize this." The model doesn't have an inner voice you need to talk to.

Context and examples

Tell the model your role, the stakes, and the format. "I'm a startup founder, help me pitch this to investors in 2 minutes" is better than "write a pitch."

Give examples if you can. "Here's a good one: [example]. Now do this: [task]" gets more consistent results.

Common mistakes

One mistake is being too short, like "summarize this" without telling it how. Another is being too fancy, like "engage your inner analyst." A third is changing your request mid-way, so stick to one task per prompt.

The last mistake is expecting perfection on the first try. Iterate instead, and ask follow-up questions to refine the answer.

Iterate, don't fight

If the first answer is close but not quite, ask for a revision instead of a completely new prompt. Models are good at tweaks.