Voice is the Missing Link
How speaking bridges the gap between your thinking, text, and AI
Welcome to Unknown Arts — I’m Patrick, your field guide to the creative frontier. Join thousands of builders around the world navigating what’s next.
Last week, I wrote about why text matters for your AI workflow: it’s the format where human thinking and AI capability naturally meet.
But then comes the obvious question: “How do I actually get all my thinking into text?”
The answer: speak more.
I grew up watching my dad dictate medical reports into a tape recorder. He’s a doctor who came of age before personal computers, when typing was a burden rather than a baseline skill. Like many physicians of his generation, he relied on dictation to move quickly through patient analysis. He’d flip through hundreds of pages of medical records, speaking his findings into little tape cassettes, then hand them off to a secretary who’d transcribe everything.
I saw dictation work and knew it was valuable. But I never did it myself because the tools weren’t accessible, and I didn’t really need them. I could type well enough to get by.
That’s changed now. Not because typing got harder, but because the barrier to dictation has disappeared entirely.
Text is still the goal, voice is just the faster path
The chain from last week hasn’t changed: human thinking becomes text becomes AI context. Text remains the format that AI models read, reason over, and transform. This week’s addition is how we get there.
Voice is the natural bridge between thinking and text. When you speak your ideas, you explain things the way you would to a friend: messy, detailed, with all the context still fresh. You tap into your innate improvisation. Then AI transcription handles the mechanical work of turning that speech into reusable text.
This solves three problems at once.
First, it removes the blank page problem. Starting to write often feels heavy, like you need the idea fully formed before you can begin. Speaking sidesteps that. You just start explaining what’s in your head.
Second, it captures richer context. When you try to type a quick note, you often write something cryptic that doesn’t make much sense later. Speech lets you capture the detail quickly without the friction.
Third, it matches how thinking actually works. Ideas come in bursts, half-formed, connected to other thoughts. Speech preserves that natural flow in a way that typing rarely does.
You’re not trying to create perfect text through speech. You’re capturing the raw material that becomes valuable because it lives in text afterward.
The old voice memo problem is solved
In the past, voice memos were useful but had a fundamental limitation: they stayed as audio. If you recorded a two-minute idea, you had to spend two minutes listening to it again later. Audio files are hard to search, hard to build on over time, hard to share with AI.
Tools like Wispr Flow, Superwhisper, and Monologue change this. You speak naturally, and the output is text, immediately available in whatever application you happen to be using.
Even ChatGPT has dictation built in. You can speak entire conversations, then ask ChatGPT to export the output as markdown. The transcription quality is good enough that your rambling thoughts become readable text with minimal cleanup.
So rather than speaking your thoughts into a hard-to-use audio file, now you can speak them directly into text with AI handling the translation.
When to speak fast, when to speak slow
Voice works across different modes of creative work. You just need to modulate how you use it.
When you’re in ideation mode, speak freely. Let the thoughts flow, interrupt yourself, circle back. The messiness is actually useful because it gives AI rich context to work with. Your job is just to get ideas out before they disappear.
When you’re documenting what you know (what your work process looks like, what you like about an example, etc.), speaking also works beautifully. Just explain it out loud like you’re teaching someone. You’ll capture detail you’d never think to write down otherwise.
When you’re moving into execution (writing final drafts, building specific features, etc.), you need to slow down a bit. Not necessarily switch to typing, but modulate your speech. Speak more deliberately. Choose words more carefully. Pause between your thoughts. Take a moment to edit the text after you capture it.
I dictate most of my direction to AI now, even when I’m doing something technical like designing software. But I’ve learned to speak slowly and precisely when giving technical instructions. The medium is still voice, but I’m using it with the intentionality I used to reserve for writing.
This is a learnable skill. The free-flowing, improvisational voice tends to come more naturally. The precise, intentional voice takes practice. But developing both means you can stay in flow across all phases of work.
Start awkward, get fluent
If you’ve never dictated your thinking before, it will feel strange at first. That’s normal.
You’re learning to externalize your thought process in real-time. To speak ideas that aren’t fully formed yet. To trust that the words will come as you talk, rather than planning them all in advance.
But here’s what I’ve learned: speaking your thoughts is already something you do constantly—it’s a foundation of daily life. You’re just redirecting that capability to take full advantage of the powerful tech now at your disposal.
The tools are finally good enough and accessible enough to make this skill worth investing in. What was once a specialized ability requiring a human transcriptionist is now open to anyone with a phone or laptop.
If AI runs on text, and voice is how you naturally create it, then learning to dictate fluently is one of the highest-leverage skills you can develop right now.
The quiet advantage goes beyond having your thinking in text. It’s being able to put it there as fast as you can think.
Until next time,
Patrick
🎯 Try this
Practice dictation for 5 minutes right now
The goal: start to feel how speaking “writes” faster than typing ever could.
What to talk about:
A quick thought currently on your mind
How you do something in your work
An example of work you admire and why
Pick one of these two approaches:
Option 1: Direct dictation (Wispr Flow)
Start a free trial of Wispr Flow
Open your notes app (Obsidian, Apple Notes, etc.)
Hit the Wispr hotkey and start speaking
Your words appear as text directly in your note
Best for: Quick capture into your existing system. What you speak is what you get.
Option 2: Conversational dictation (ChatGPT)
Open ChatGPT and click the dictation button
Speak your thoughts for 2-3 minutes on any topic
Ask: “Format these thoughts as a markdown note”
Copy the output to your notes system
Best for: Getting AI to synthesize your rambling into something more structured. You get your thoughts plus a layer of synthesis.
📚 Go deeper
“10x Your Whole Workflow With One AI Tool” — This Week in AI
A 10-minute video demo-ing WisprFlow and comparing it with Apple Dictation.
“Thinking Out Loud: How to Use Your Voice in Knowledge Work” — Anne-Laure Le Cunff / Ness Labs
Explores the cognitive science behind verbal thinking—why speaking activates working memory and the self-explanation effect differently than writing.
“Vibe Code With Your Voice in Google AI Studio” — Google for Developers
A one-minute demo of using voice dictation for vibe coding. While I don’t use Google AI Studio, this short video shows the basic workflow that applies to dictation across any AI tool
Find this useful? Share it with someone who might also get value.
Not subscribed yet? Get my writing directly — 7k+ already do.
I love this! I speak my thoughts all the time. When I write a post on Substack, I'll dicate the the whole story first and then rewrite it. I hadn't really thought about why I do that, but your post really resonates. 🙂
Excellent post, Patrick. I am speaking this comment into Voibe, a fast, private and easy AI voice-to-text tool I recently found. The upside: It costs half or less the price of the apps you mentioned The downside: It's only available for Mac. For any of your readers who use a Mac, I suggest checking out Voibe at getvoibe.com.