Your Keyboard Is Slowing You Down

I wrote this entire newsletter without touching my keyboard.

The first draft came from a walk around my neighborhood, talking to my phone. The edits came from talking to my computer. No typing. Just talking.

That's the point.

Last week, I made the case that AI meeting transcription is the easiest, highest-ROI way to start using AI at work. The core insight was simple: when the conversation is being captured, you stop splitting your brain between listening and note-taking. You can actually be present.

This week, I want to extend that idea. If you shouldn't be taking notes in meetings anymore, you probably shouldn't be typing as much as you are either.

This is true for basically everyone. The average person types somewhere between 40 and 70 words per minute. The average person speaks at 120 to 150 words per minute. That's roughly 2 to 3x faster. And yet most knowledge workers are still face-down in their phones and laptops, pecking away at keyboards, spending hours each week on something that could take a fraction of the time.

Here's the thing: your keyboard was never optimized for speed in the first place.

The QWERTY layout was designed in the 1870s to prevent typewriter keys from jamming. It deliberately slowed typists down by spreading commonly used letters across the keyboard, forcing your fingers to travel farther than they need to. The Dvorak layout, developed in the 1930s, put the most frequently used keys in the home row where your fingers naturally rest. Studies have shown it's more efficient. But switching costs kept everyone locked in to QWERTY, and here we are, 150 years later, still using a layout designed to solve a mechanical problem that no longer exists.

There are even stranger solutions out there. Stenotype machines, used by court reporters, allow trained operators to hit 200+ words per minute through chorded keystrokes. The CharaChorder takes this further with a device that lets you type entire words with single chord combinations. But almost nobody uses them because the learning curve is brutal and the muscle memory from QWERTY is too deeply ingrained.

None of this matters anymore.

AI transcription has gotten good enough, and cheap enough, that speaking is now the fastest way to get words into a computer. The keyboard bottleneck is dissolving.

Voice-to-text has been around for years. Siri, Alexa, Google Assistant. We've had these tools, and mostly they've been disappointing. Clunky, inaccurate, useful for setting timers and not much else.

What's different now is the underlying transcription models. The AI doesn't just hear your words. It understands context, adds punctuation, corrects small mistakes, and produces clean, readable text. You can ramble, pause, restart a sentence, and the output still makes sense. It's genuinely usable for real work.

And here's the unlock that most people don't realize: you can whisper and it still works. You don't have to be the obnoxious person on the train dictating emails at full volume. Modern transcription tools can pick up speech at remarkably low decibels. The social awkwardness barrier, the main reason people resist this, is smaller than you think.

There's another angle here that I don't think gets enough attention.

We're all aware of the screen time epidemic. People feeling self-conscious about how many hours they spend staring at their phones. Feeling tethered to their laptops. Trying elaborate workarounds like locking their phone in another room, bricking it with app blockers, or going full digital detox on weekends.

Voice dictation is an elegant solution. You don't have to white-knuckle your way through phone addiction. You can just...talk. Respond to emails while walking. Draft a document while doing dishes. Get your work done without your face buried in a screen.

I drafted part of this newsletter on a walk around my neighborhood. I wasn't staring at my phone. I was just talking, getting my steps in, and letting the AI capture everything. That's the future of a lot of knowledge work, and it's available right now.

My preferred tool is Wispr Flow. It runs in the background on my laptop as a small widget above the dock. When I want to dictate, I hold down the Fn key, speak, and release. The transcription prints directly into whatever window I have open, whether that's Claude, email, a Google Doc, or Slack. It also saves a backup copy in Wispr Flow's archive.

A few things that make it work:

The Fn key shortcut. It's unobtrusive. No clicking, no app-switching. Just hold, speak, release. It becomes muscle memory fast. And if you're an investment banker or consultant living in Excel, you can change the shortcut to whatever key you want. It doesn't have to be Fn.

Words-per-minute feedback. Wispr Flow tells you how fast you're speaking after each dictation. You'll see very quickly that you're operating at 2 to 3x your typing speed. For managers and executives thinking about team productivity, the math adds up: if your team spends even a few hours per week typing, that's days per year in aggregate. Dictation compresses that.

The archive. Every dictation is timestamped and saved. It's both a backup and a log of everything you've captured.

If you try Wispr Flow with your AirPods connected, it will produce gibberish. I learned this the hard way.

For whatever technical reason, Bluetooth headphone microphones don't play well with AI transcription. You need to use either your device's built-in microphone, a wired headset, or an external mic. A podcasting mic or a portable wireless mic like the DJI Mic Mini work great. This is the number one reason people bounce off voice dictation: they try it once with the wrong microphone, get nonsense output, and assume the technology doesn't work.

It works. You just need the right mic. Check your settings before you start.

Download Wispr Flow or an equivalent. Use it for one week. Dictate at least one email, one note, and one message per day. Notice what changes. Not just in speed, but in how you think. When you're speaking instead of typing, you're not editing yourself mid-sentence. You're just talking. And sometimes that's exactly what you need.

"Something Big Is Happening" took over the internet this week. Matt Shumer's viral essay comparing this moment in AI to February 2020 hit 80+ million views. Will Manidis quickly responded with "Tool Shaped Objects," a sharp critique of AI hype that produces nothing but the feeling of productivity. The Information's Abram Brown summarized the whole thing in "AI Industry's Big Flaw: No One Speaks Normie." His point: everyone is talking past each other and freaking out. 80 million views, and the tech industry still can't explain AI to normal people. Which is, incidentally, what this newsletter is for.

Referral program is live. You should now see a referral tracker at the bottom of this email. Here's how it works:

And if you have way too much time on your hands: refer 100 people to the newsletter and I'll send you a Mac Mini. I'm serious. If you know, you know. If you don't, ask me, and I’ll explain why AI nerds are ordering Mac Minis.