It’s simply a warning to crocodiles that there’s no lifeguard on duty.
Scaled sort usually gives good results.
The LLM is just trying to produce output text that resembles the patterns it saw in the training set. There’s no “reasoning” involved.
He definitely bought Twitter to “preserve free speech” and totally not to collect data and manipulate elections, right?
It’s another reddit alternative.