Building a speech-to-Markdown app with three coding agents (opens in new tab)

I wanted to know more about current web speech recognition APIs. I tried three different coding agents to set up a proof of concept app for me.

To my surprise, Cursor's composer model beat both Claude and Gemini.