AI writes at maximum density

You know how we are all taught in college writing “tighten it up, cut the filler”? but because we don’t want to sound weird, we definitely don’t speak that way?

Claude writes like this: “The architect overflowed no count-based limit”

when it could have said: “the architect didn’t overflow a count-based limit”

This inversion is the language equivalent of coding one liners — save a token, lose the reader.

I think this is one symptom of why it’s so easy to notice AI writing. It’s almost always operating at maximum density

Is this because it’s trained on so many academic papers? Or is it trying to be token efficient?

In real life, the slack in how we talk gives our brains natural breaks in density of meaning, and helps us stay tuned in.

Repetition, filler words, using more words to minimize awkward inversions.

I experience conversation as a buffer, I only have so much depth to fill up before I have to say “wait… what?”

I’m all for good, punchy writing; but the densifying of information is going to make us all start to think about our literal, physical weaknesses a lot more.