[spats1990]

When you encounter these comments/sentiment, pretend that LLM = Loweffort Long Mumbling. In other words, poor writing.

Detection of "LLM" is a red herring. Quality is what matters. Always has been. Assess comment quality holistically, and you'll be fine.

[krapp]

If "quality" is all that matters and maximizing quality is the goal, and if LLMs can generate higher quality comments more consistently than humans, we should close all user accounts. Don't even have this be a forum anymore. Have LLMs crawl the web, post articles then generate threads discussing them from various simulated points of view. No direct human participation, no Eternal September. Then readers can have their own agents summarize the threads for them.

We can consider this the carcinization of online discourse - everything evolves towards the optimum of LLM summarization.

[spats1990]

> if LLMs can generate higher quality comments more consistently than humans

Do you believe this?

[krapp]

No, because my definition of "quality" for comments implicitly includes human intent, which LLMs lack.

But I suspect a lot of people on HN only view these threads as data and that for them "quality" only exists within the semantics and structure of the text itself, and the human element doesn't matter to them.