Hacker News Reader

Awesome ai generated comment. Seriously when will @dang integrate pangram into comments?

Does pangram work?

Alas, tl;dr: no. (it caught the one above, easily, but didn't catch ones from Opus 4.6, it's very confident they're 100% human. as I would be :/ )

Why is that a tldr no if it did indeed catch the blatant ai comment? Are you funded by the competitor? Lol

Because it caught one of four :) because it's unclear what "works" means, I spelled it all out. :)

I guess if it catches some but not all, that's better than catching zero, but then what is the false positive rate?

Randomness would also catch 'some but not all', after all.

I would think that you would just send posts with a prediction rate > X to the mod queue as auto-flagged

Assuming that the false positive rate is low enough. That's the question, no?