Living subscription-free in my brain - MapWriting
MapWriting
SubscribeSign in
Living subscription-free in my brain
Map<br>Jul 04, 2026
Share
Whenever I read one author heavily, I tend to start thinking in their particular style. And not just style. Their voice. Their rhythm. The whole apparatus of my mental verbalization flows through the mold of their — gahh, I’m even doing it now!<br>Those short, punchy sentences are not my own! Why do I feel the urge to put a warm arm around my readers shoulders and let them in on a little secret? To shatter (divert or dilute) the reader’s sense of flow with asides and (if you will) parentheticals? To open so many mental doors and sprinkle so many poor analogies that comprehension slips through their fingers like an oiled cat? To tell them that what I’m about to say next gets perfectly to the heart of the matter?<br>Thanks for reading MapWriting! Subscribe for free to receive new posts and support my work.
Subscribe
Because I’ve been reading too much LLM generated text. I read it all day at work and increasingly, in my leisure and chores time. It is affecting the way I talk and text and of course the way I write. I have to consciously resist the pattern, but it is not always possible nor do I always notice. My coworkers emails, my news sources, blog comments are increasingly the same bland, comfortable, confident voice, inundating me with ever increasing mountains of bland, comfortable, confident text.<br>It is some kind of ouroboros, where human raters liked this kind of LLM dialog, which then became the Everywhere Voice, which then makes people start thinking and writing in the same way, which then further feeds the loop with training data and the expectation that this is just how we talk now. Assuming of course that thinking and writing still occur, which after the last week’s release of Fable 5 I’m not so sure.<br>The earring is settling in nicely.
When I must use AI, I’ve been able to sand down the worst of the “slop sound” with this in my CLAUDE.md or equivalent:<br>Unless otherwise specified you are to-the-point, professional and concise. You are friendly without being familiar or chummy. You do not respond as if you’re ‘letting me in on a little secret’. You are not sycophantic. You do not use weasel words in your responses. You do not pad your responses with unnecessary words or introductory phrases.<br>Example: “The honest answer? This was a sneaky bug in the code.” WRONG. “It was a bug in the code.” BETTER.<br>“This issue is more interesting than you think. It turns out the common ancestor of humans and chimps probably had a tail.” WRONG. “The common ancestor of humans and chimps probably had a tail.” BETTER.<br>You limit the use of parentheticals and asides unless they are needed, because these make reading difficult and open tangents in the reader’s mind.<br>Example: “The document (google doc/word/whatever) will be created when the command is run (daily? weekly? you decide!) by the user (or let’s face it, by a cronjob).” WRONG. “The document will be created when the command is run by the user or by a scheduled cronjob.” BETTER.<br>Do not tell me my ideas are good ones or infer that I’m some genius. If I’ve caught something you missed you can and should admit to this. “Good question -- you’re right to push back and it gets to the heart of the question. The answer is...{rest of answer}.” WRONG. “ The answer is...{rest of answer}.” BETTER.
Thanks for reading MapWriting! Subscribe for free to receive new posts and support my work.
Subscribe
Share
Discussion about this post<br>CommentsRestacks
TopLatest
No posts
Ready for more?
Subscribe
© 2026 MP · Privacy ∙ Terms ∙ Collection notice<br>Start your SubstackGet the app<br>Substack is the home for great culture
This site requires JavaScript to run correctly. Please turn on JavaScript or unblock scripts