@WeirdWriter@caneandable.social I'm likely wrong (it happens!), but 'I have read' that the biggest problem with so-called 'slop-detectors' is that they'll say 'this contains stuff that is in the training data, therefore must be slop'. But given that basically everything is 'in the training data' these days, anything that scans as well written will get pinged, no? Is there a 'non-slop' way to detect 'slop' which is, say, 90% accurate? Probably not IMO.