I'm a developer. I type fast. I've never seriously considered voice dictation because — honestly — it always felt like a gimmick. Something for people who can't type, not for people who live in a terminal.
Then I built Air Wisper. And I realized I should probably eat my own dog food.
So I set myself a challenge: use voice for all non-code writing for 7 days straight. Emails, Slack messages, documentation, notes, commit messages — everything that isn't literal code syntax. Here's what happened.
The rules
- All prose writing must be dictated (emails, Slack, docs, notes, commit messages)
- Code itself is still typed (obviously)
- I can edit the output after dictation — but not rewrite it from scratch
- I track how it feels each day: friction, speed, quality
Awkward but fast
The first Slack message felt ridiculous. I'm sitting at my desk, headphones on, talking to my computer. My instinct was to whisper, which made the transcription worse.
Once I committed to speaking at a normal volume, things clicked. My first email — a reply to a client about a deadline — took about 15 seconds to dictate. Normally I would have spent 3 minutes typing, rewording, and second-guessing my tone. The AI cleanup handled the tone just fine.
Verdict: Faster than expected. Socially awkward.
Commit messages become useful
This was the first real surprise. I've always written lazy commit messages because the friction of switching from code-brain to prose-brain is real. With voice, I just say what I did after finishing a feature.
"Refactored the webhook handler to retry failed deliveries up to three times with exponential backoff, also added logging for each retry attempt."
The AI turned that into a clean, concise commit message. I started actually looking forward to committing.
Verdict: This might be the killer use case for developers.
Long-form writing feels like cheating
I had to write a technical design document for a new feature. Normally this would take me an hour of staring at a blank page, typing a paragraph, deleting it, trying again.
Instead, I opened the doc, held my shortcut key, and just... explained the design. Like I was talking to a colleague. Five minutes of talking produced about 800 words of clean documentation. I spent another 10 minutes editing and restructuring. Total time: 15 minutes.
A task that usually takes an hour took 15 minutes. That's not a marginal improvement — that's a category change.
Verdict: Voice removes the blank page problem entirely.
The social barrier disappears
I work from home, so talking out loud isn't an issue. But today I had a meeting at a co-working space. I couldn't dictate with people around.
This is the real limitation: voice needs a private-ish environment. Not silent — background noise is fine — but you need to be comfortable speaking at normal volume. Open offices are tricky. Coffee shops are fine if you don't care what people think.
At home in the evening, I caught up on all my messages in about 5 minutes. Would have taken 20+ typing.
Verdict: Environment matters. This is a home office / private space tool.
I stopped thinking about it
By day 5, dictation stopped being "a thing I'm trying" and just became how I write. The shortcut key (⌥D) was muscle memory. I'd be reading a Slack thread, hold the key, reply, release. No conscious effort.
The weirdest part: my messages got more thoughtful. When you type, you often just react — short, blunt replies. When you speak, you naturally add context and nuance. My teammates actually commented that my Slack messages were "more detailed than usual."
Verdict: Voice doesn't just make writing faster — it makes it better.
It leaks into personal life
Saturday morning, I caught myself dictating a grocery list into Apple Notes. Then a message to a friend. Then a journal entry.
The journal entry was the most interesting. I've tried journaling many times and always quit because typing my thoughts feels forced. Speaking them felt natural. I talked for about 2 minutes and ended up with a page of clean text that actually captured what I was thinking.
Verdict: Once the habit forms, you use it everywhere.
The numbers
I averaged about 21 dictations per day. Most were short (Slack messages, commit messages), but a few were long (documentation, emails). Rough estimate: I saved 30-40 minutes per day compared to typing. Over the week, that's about 4 hours.
What I'll keep doing
After the week, I dropped back to a hybrid approach. Here's what stuck:
- Always voice: emails, Slack replies, commit messages, code comments, documentation, notes
- Always type: code, terminal commands, search queries, short edits
- Depends: very short messages (under 10 words) — sometimes typing is just faster for "sounds good, ship it"
What surprised me most
It wasn't the speed. I expected that. What surprised me was how much writing I was avoiding because of typing friction. Code comments I should have written but didn't. Documentation I kept postponing. Thoughtful replies I replaced with "lgtm."
Voice didn't just make me faster — it made me write things I otherwise wouldn't have written at all. That's the real unlock.
The takeaway: Voice-to-text isn't about replacing your keyboard. It's about removing the friction between thinking and writing. Once that friction is gone, you write more, write better, and write faster.
Try it yourself
Air Wisper is free to start. 200 requests/week, on-device transcription, AI cleanup. See if voice changes how you write.
Get Started FreeAir Wisper is a native macOS app. Requires macOS 14 or later.