đ Typing Is the New Phone Number
Iâve been using voice to input text for almost everything now. DJI mic clipped to my shirt, VoiceInk running in the background, talking through my code, my emails, my messages. I wrote about the workflow in Misheard Lyrics for Robotsâhow voice transcription errors donât matter when your listener is an LLM.
But something else is happening that I didnât expect: my typing is getting worse.
Use It or Lose It
After only a few months of voice-first input, Iâve noticed my typing accuracy dropping. When I do sit down to type something manuallyâa short command, a quick messageâmy fingers stumble more than they used to. Keys hit in the wrong order, words misspelled, sentences that need backspacing.
But it mostly doesnât matter. When Iâm typing into an LLM conversation, the model just works around the errors. It recognizes my intent through the noise the same way it handles garbled voice transcription. A typo-riddled sentence carries the same meaning as a clean one, and the AI doesnât judge your spelling.
MapleStory Typing
The last time I remember typing this sloppily was in MapleStory.
If you played MMOs in the 2000s, you know the rhythm. Youâre in the middle of fighting something, your hands on the movement and attack keys, and someone messages you. You need to respond. So you quickly switch to the chat bar, mash out a reply as fast as possible, hit enter, and get your hands back on the game controls before your character dies.
Nobody typed carefully in MapleStory. You typed âcomignâ and âwaht lvl r uâ and âmeet me at hensys.â And it was fine. The other players understood you perfectly because the context was so strong. Youâre both in the same game, doing the same thing, and the meaning bleeds through the typos.
Thatâs exactly what talking to an LLM feels like now. The context is so richâthe conversation history, the codebase, the task at handâthat my mangled keyboard input still lands.
The Precision Gap
Between MapleStory and now, there was a long stretch where typing quality mattered. School essays, professional emails, code reviews, Slack messages to coworkers. Every context had an audience that would notice sloppy text, or worse, a compiler that would reject it entirely.
I learned to type carefully. Everyone did. The tools demanded precision, and so you gave them precision.
Now the tools donât demand it anymore. When my primary audience is an AI that can reconstruct âteh functon shoudl retrun a list of usresâ into a working database query, the incentive to type cleanly evaporates. And without the incentive, the skill quietly starts to fade.
Phone Numbers
This reminds me of phone numbers.
People used to memorize dozens of them. Not because memorization was fun, but because you needed them. Your address book was at home. Youâre at a pay phone, or at a friendâs house, or at work, and you need to call someone. So you kept the important numbers in your head.
Then phones got contact lists. Then smartphones made the contact list permanent and portable. The skill of memorizing phone numbers didnât just become unnecessaryâit actively atrophied. People who used to know fifty numbers by heart now struggle to remember more than two or three. Their parents, their partner, maybe their own.
The numbers didnât get harder to memorize. People just stopped practicing, because the environment no longer required it.
Shedding Skills
Typing accuracy feels like itâs on the same trajectory. Not disappearing overnight, but gradually fading as the environments that demanded it are replaced by environments that donât. When every text input field has an AI on the other end that can parse your intent regardless of your typos, clean typing becomes a nice-to-have rather than a requirement.
I donât think this is necessarily bad. Nobody mourns the loss of their phone number memorization system. The mental capacity freed up just gets used for other things. Or it doesnât, and thatâs fine too.
But it is strange to feel a skill degrading in real time. To sit down at a keyboard and notice that your fingers donât land where they used to, and to realize that you donât really care.
I've been using voice to input text for almost everything now. [DJI mic clipped to my shirt](/10-bits-per-second.md), [VoiceInk](https://tryvoiceink.com/) running in the background, talking through my code, my emails, my messages. I wrote about the workflow in [Misheard Lyrics for Robots](/misheard-lyrics-for-robots.md)âhow voice transcription errors don't matter when your listener is an LLM.
But something else is happening that I didn't expect: my typing is getting worse.
## Use It or Lose It
After only a few months of voice-first input, I've noticed my typing accuracy dropping. When I do sit down to type something manuallyâa short command, a quick messageâmy fingers stumble more than they used to. Keys hit in the wrong order, words misspelled, sentences that need backspacing.
But it mostly doesn't matter. When I'm typing into an LLM conversation, the model just works around the errors. It recognizes my intent through the noise the same way it handles garbled voice transcription. A typo-riddled sentence carries the same meaning as a clean one, and the AI doesn't judge your spelling.
## MapleStory Typing
The last time I remember typing this sloppily was in MapleStory.
If you played MMOs in the 2000s, you know the rhythm. You're in the middle of fighting something, your hands on the movement and attack keys, and someone messages you. You need to respond. So you quickly switch to the chat bar, mash out a reply as fast as possible, hit enter, and get your hands back on the game controls before your character dies.
Nobody typed carefully in MapleStory. You typed "comign" and "waht lvl r u" and "meet me at hensys." And it was fine. The other players understood you perfectly because the context was so strong. You're both in the same game, doing the same thing, and the meaning bleeds through the typos.
That's exactly what talking to an LLM feels like now. The context is so richâthe conversation history, the codebase, the task at handâthat my mangled keyboard input still lands.
## The Precision Gap
Between MapleStory and now, there was a long stretch where typing quality mattered. School essays, professional emails, code reviews, Slack messages to coworkers. Every context had an audience that would notice sloppy text, or worse, a compiler that would reject it entirely.
I learned to type carefully. Everyone did. The tools demanded precision, and so you gave them precision.
Now the tools don't demand it anymore. When my primary audience is an AI that can reconstruct "teh functon shoudl retrun a list of usres" into a working database query, the incentive to type cleanly evaporates. And without the incentive, the skill quietly starts to fade.
## Phone Numbers
This reminds me of phone numbers.
People used to memorize dozens of them. Not because memorization was fun, but because you needed them. Your address book was at home. You're at a pay phone, or at a friend's house, or at work, and you need to call someone. So you kept the important numbers in your head.
Then phones got contact lists. Then smartphones made the contact list permanent and portable. The skill of memorizing phone numbers didn't just become unnecessaryâit actively atrophied. People who used to know fifty numbers by heart now struggle to remember more than two or three. Their parents, their partner, maybe their own.
The numbers didn't get harder to memorize. People just stopped practicing, because the environment no longer required it.
## Shedding Skills
Typing accuracy feels like it's on the same trajectory. Not disappearing overnight, but gradually fading as the environments that demanded it are replaced by environments that don't. When every text input field has an AI on the other end that can parse your intent regardless of your typos, clean typing becomes a nice-to-have rather than a requirement.
I don't think this is necessarily bad. Nobody mourns the loss of their phone number memorization system. The mental capacity freed up just gets used for other things. Or it doesn't, and that's fine too.
But it is strange to feel a skill degrading in real time. To sit down at a keyboard and notice that your fingers don't land where they used to, and to realize that you don't really care.