About two months ago I started an experiment. Every task I normally typed, I tried doing by voice instead.
Not voice-to-text dictation. I've tried that before - it produces one long wall of words that still needs an hour of editing. I mean actual voice-first thinking, where I talk through a problem and the AI handles the rest.
The results surprised me. Some things got dramatically faster. Some things stayed the same. A few got worse.
What got faster:
First drafts of anything. When I'm writing an email, a Slack message, or a post like this one, talking is about 3x faster than typing. I don't have to think about spelling or sentence structure - I just say what I mean and fix it after.
Capturing ideas. This was the unexpected one. When a thought surfaces in the middle of something else, I can voice-capture it in under 10 seconds without stopping what I'm doing. I've now saved probably 50 ideas that would have evaporated by the time I found a keyboard.
Client update emails. I used to dread writing these. Now I talk through what happened this week, what's blocked, what's next - takes about 2 minutes. The emails are actually better because they sound like me.
What stayed the same:
Code review comments. You need precision that voice is slow at.
Anything with technical terms or product names. Getting Supabase or Vercel spelled correctly by voice is more work than typing.
What got worse:
Deep writing that requires real thinking. Voice is fast but it's also surface-level fast. When I'm working through a genuinely hard problem, typing still works better because the friction slows me down enough to actually think.
The tool I use for this is Genie 007, which handles the voice-to-action part so I'm not manually cleaning up dictation afterward.
https://genie007.com/?utm_source=indiehackers&utm_medium=post
Two months in, voice-first saves me around 45 minutes a day. Most of that comes from first drafts and idea capture.
What's surprised you most about where AI tools actually fit into your workflow - and where they don't?
Voice-first thinking vs voice-to-text is an important distinction. Dictation still leaves you editing a wall of words; talking through a problem externalizes the messy middle that typing usually compresses too early.
I have been trying a lighter version for product decisions on a Discord discovery directory — narrating "why this server fits" out loud before writing ranking rules. It catches assumptions I would otherwise polish into the UI.
After sixty days, what stuck as permanent vs what quietly slid back to keyboard when stakes felt high?