I’ve contented that a primary, detrimental byproduct of the LLM age will be an unimaginable torrent of content whose magnitude is beyond which any human can possibly consume.
I don’t know about you, but my habits are already changing. A year ago if I’d been linked a lengthy blog post, I would’ve sat down and read it without a second thought. Today, unless it’s from an author I have a preexisting no-LLM trust relationship with, two paragraphs in my eyes glaze over and I’m skimming before I know it. Maybe it’s not LLM-generated, but it probably is.
Personally, I don’t generate writing via LLM that’s intended to be read by other humans because I don’t think it’s respectful, but I know it’s a Luddite take that’ll be hard to convince anyone else of.
Instead, I wonder if this would be a more reasonable ask: if you’re going to generate content via LLM, keep it short.
If it took 30 seconds to prompt some prose into existence that’d take me 10 minutes to read, that’s an unforgivable asymmetry. But if it takes you 15 seconds of prompting and me 15 seconds of skimming, that seems … fine? The LLM is there to clean things up and act as the world’s best editor and proofreader, but it’s not there to add a bunch of fluff to waste my time.
So how about this: tell Claude, tell Codex, 3-5 paragraphs and a code example. Possibly a couple bullet points. And maybe, just maybe, the input prompt was already enough?
Did I make a mistake? Please consider sending a pull request.