e.g. meta-decisions about what to watch (it does give a slight spoiler tho)

Unironically pretty good use of LLMs is to get spoiler-free info about stuff you're watching / reading

“The luddites were right”

Replicating a fairly similar response to the Jon Bois tweet using the "empathy" counterspeech prompt from that paper and the same LLM they use

E.g. compare these two:

Have you seen this paper: arxiv.org/pdf/2411.14986
To me some of these replies look strikingly similar to the "counterspeech" LLM interventions studied in this paper, which finds a backlash effect. So I almost wonder if these could be misguided attempts to do something constructive

Local comparative examples

Nature as standing reserve

perpetual-mend levis