I Haven't Typed to AI in Weeks — Here's Why I'm Not Going Back

The problem most people have with AI isn't that they're bad at prompting. It's that typing is friction. And when there's friction, you type less. And when you type less, you give AI less context. And when AI has less context, its answers feel off-target. And then you get frustrated and you blame the tool.

I saw a few YouTubers talking about this a few weeks ago. My first reaction was — another productivity hack I'll try once and forget. But I was curious enough to actually try it. Not because I was unhappy with typing, honestly I was quite fast at it — around 80 words per minute consistently. But I figured, why not see what the fuss is about.

What I didn't expect was how much it changed the quality of outputs I was getting back.

I've been recommending this to everyone I know since then. Not because it's cool or trendy, but because I've genuinely seen the difference — and it's not subtle.

The real unlock isn't some advanced prompting framework. It's simpler than that: talk more.

The context is the lever

Here's what I mean. If you describe your situation to a senior colleague — someone with years of experience in your world — you don't have to spell everything out. You say three sentences and they get it. They fill in the gaps from their own knowledge. They ask a sharp clarifying question because they understood the nuance.

If you give the same three-sentence brief to an eager intern who's just started? You'll get something back that technically addresses what you said, but missed the point entirely. Because they didn't have enough context to know what mattered.

AI is more like that intern. The more context you give it, the better it fills in the gaps on its own.

Typing takes effort. And when something takes effort, you do less of it. You compress your thoughts, you cut corners, you simplify. What should have been a five-sentence dump of your actual situation becomes two sentences that barely scratch the surface.

Voice gets you context for free. Now I use TypeWhisper on my Mac — it's free and open source for personal use, and it runs entirely offline using local models. The only requirement is a computer with enough performance to run those models. The accuracy is genuinely impressive. I almost never have to go back and correct anything. Even with technical terms and abbreviations, it gets it right the first time. My Android phone gets a similar experience through FUTO Keyboard, which also runs local speech-to-text on-device. Near-identical to the Mac, just not quite as seamless.

And thinking out loud is faster than thinking in full sentences. I can articulate a messy, half-formed thought in 30 seconds that would have taken me three minutes to type out properly. It doesn't matter if your English isn't perfect — the AI is good at picking out the relevant signal from the noise. Even if your first language isn't English, as long as you can get the idea across, it still works.

I can't go back to typing. Full stop.

Try this today

The lowest-friction starting point is right in front of you. Whatever AI you use — ChatGPT, Claude, Gemini — they all support voice input. On ChatGPT's free tier, the voice button is right there. Tap it, describe your situation in your own words, and hit send. Don't clean up your sentences. Don't edit as you go. Just talk.

If you're on Mac, TypeWhisper is worth the setup. It's free, completely private since it runs offline, and the accuracy is in a different league from built-in dictation. There's a Windows beta version available too if you're on PC. On Android, FUTO Keyboard gives you a comparable experience.

The goal isn't to find the perfect tool. It's to feel the difference in output quality when you're not fighting your keyboard.

What to watch for

A few honest caveats before you go all in:

• Transcription errors happen. Don't assume the transcript is exactly what you said. Rarely — but occasionally — a word comes out slightly different from what you intended. Always skim before you hit send, especially if the output is going somewhere real.
• Both tools run entirely offline. Nothing leaves your device. If you're working with sensitive information, this is as private as it gets — no server, no cloud component.
• Speak naturally, not formally. Don't perform for the tool. The AI underneath is good at figuring out what you meant, even if your sentence wasn't grammatically clean. Let go of the instinct to phrase things correctly. Just talk.

The real win

I keep coming back to this. The win isn't that you save time. The win is that the output is better — more specific, more useful, closer to what you actually had in mind when you started.

And that changes how you feel about using AI. It's not a chore anymore. It's not "I have to go type this out." It's "I already know what I want to say, let me just say it."

That's the shift. Once you feel it, you don't go back.