The habit started on my work computer
I first used Wispr Flow on my Windows work computer. That is where this whole experiment started.
The reason was simple. It understood me.
I regularly move between English, Czech, and Slovak. Flow worked well enough across all three that speaking stopped feeling like a demo and became part of my actual workflow.
That changed more than my typing speed. When I have a half-formed thought, speaking lets me get it out before I polish it to death. Sometimes the pause between an idea and the keyboard is exactly where a useful detail disappears.
With dictation, I can catch the whole messy thought first and edit it later.
A quick correction: I used to call these text-to-speech agents, but that is backwards. They turn speech into text. The best ones also clean and adapt the words instead of merely transcribing them, which is why they can feel like agents.
Voice made me better at giving AI context
The biggest benefit showed up in my work with large language models.
If you work with AI, your output depends heavily on the context you provide. The model cannot use the assumptions that stayed in your head. Yet when I type a prompt, I tend to compress it. I skip background and constraints because they feel obvious, then wonder why the answer missed the point.
Speaking changes that behavior. I explain the situation more like I would explain it to a colleague. I mention what I already tried, what failed, what the output is for, and what "good" should look like. I also catch contradictions while I am saying them out loud.
That does not make the model smarter. It gives the model a better brief.
