The Art of Talking to AI: Prompts That Actually Work for Coding

The article argues that AI tools like ChatGPT, GitHub Copilot, and Cursor are useful for coding, but they need clear instructions. It says beginners often get better results when they describe what they are building, their skill level, the languages or tools they want, and the output format they expect.

Reading notes

#

• AI works better when the prompt is specific instead of vague.
• A good prompt says what you are building, such as a to-do list app or a contact form.
• Mentioning that you are a beginner helps AI give simpler explanations.
• The prompt should name the languages or tools you want, such as HTML, CSS, JavaScript, or Python.
• Asking for explanations turns the interaction into learning, not just copying code.
• Clear output instructions can ask for separate code blocks, comments, or testing steps.
• The article gives before-and-after examples for a webpage, a JavaScript function, and fixing broken code.
• It offers templates for learning concepts, building projects, debugging, and code review.
• It warns against assuming AI knows your setup, asking for everything at once, or skipping follow-up questions.
• It suggests a back-and-forth conversation where the user refines the request and asks for changes.