AI Coding Tip: Dictate Your Prompts Instead of Typing Them

AI Coding Tip 018: Dictate Your Prompts Instead of Typing Them

TLDR: Maxi Contieri's latest AI coding tip is refreshingly simple: stop typing your prompts character by character and start speaking them. You talk roughly twice as fast as you type, which means you can give AI assistants far more context in the same amount of time, or the same context in half the time.

Summary: There's something almost obvious about this tip once you hear it, and yet I suspect most of us are still hunched over our keyboards laboriously typing out multi-sentence prompts to Copilot, Cursor, or Claude. The math is pretty straightforward. The average developer types somewhere around 60 to 80 words per minute on a good day. Most people can speak comfortably at 120 to 150 words per minute. That gap matters when you're trying to construct a detailed prompt that gives an AI enough context to actually do useful work.

What often happens when we type prompts is that we abbreviate. We cut corners. We leave out the nuance because writing it out feels like too much effort. "Fix this bug" instead of "This function is supposed to return an array of user IDs from the cache, but it's returning undefined when the cache is empty, and I need it to return an empty array instead." The second prompt is dramatically better. Speaking it takes maybe ten seconds.

Contieri frames this as a semi-automatic, beginner-level tip, which is the right framing. You don't need any special tooling beyond a decent dictation setup. macOS has built-in dictation. Windows has voice access. Most operating systems do. You press a hotkey, you talk, and the words appear in whatever text field you're focused on. It's not magic, but the compounding effect over a full workday is real.

The deeper insight here is about friction and the prompting habits we build around it. When typing is the bottleneck, we unconsciously learn to write terse prompts and then complain when the AI gives us a generic response. Remove the bottleneck and you naturally start giving more context, describing edge cases, explaining what you've already tried. Those are the prompts that produce genuinely useful output.

Key takeaways:

• Speaking is roughly twice as fast as typing for most people, which directly translates to richer, more detailed prompts with less perceived effort
• Terse prompts are often a friction artifact rather than a deliberate choice, and dictation removes that friction naturally
• Built-in OS dictation tools on macOS and Windows work for this without any additional setup or cost

Why do I care: From an architecture standpoint, the quality of what you feed into AI tools directly affects the quality of what comes back. I've watched teams adopt AI coding assistants and then conclude they aren't useful, when really the problem is that the prompts are too vague to produce good results. Anything that lowers the barrier to writing richer, more specific prompts is worth paying attention to. This tip won't change the world, but it's the kind of small habit shift that adds up over months of daily use.

AI Coding Tip 018 - Dictate Your Prompts Instead of Typing Them