Verbosity Is Lazy Now

It didn't use to be this way. A longer document used to mean more labor for the writer and more insight for the reader. But now a large body of eloquent speech looks like slop, while a few punchy sentences feel more genuine. That's because it's easier to get an LLM to generate content that looks like a plausible piece of work than to put the actual work into figuring out what you want to say and how to say it best.

Frankly, I will think twice before reading a wall of text unless I know for sure that a real person has put some meaning into it. I don't need the product of someone else's prompt — I can prompt a model on the same sources myself, and that product will suit my taste and interests better (for starters, it will be shorter).

Of course, the verbosity of LLM output is a fixable problem. You can easily prompt a model to be more concise. But the fact of the matter is that most LLM output posted right now is verbose, and that feels like a relic of the olden times when verbosity was valued. Maybe we, as consumers of AI, will eventually learn to value brevity. If so, we will at least waste less time chewing through a text to extract whatever valuable information is encoded there — or to discover that there is none, and that the text was generated only to take up space and signal expertise.

If you have nothing to say, maybe it's better to say nothing at all than to copy an LLM and add to the noise, of which there is already plenty in our lives.

If you actually want to say something, say that something and nothing else. An LLM can help format raw thoughts into content that is easier to read, but make sure it's your thoughts in that content. Cut the paragraphs and sentences that stand in the way and don't deliver anything new.

And so, the paradox: a shorter piece of text signals more effort and more genuine human involvement than a longer one. At least in 2026.