Positive users praise A/B testing of Fable and Sol plus Sol's reasonable pushback without strawmen, while negative users blame its inconsistent behavior on poor training.
Based on 3 visible X reactions from 8 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@tessera_antra @TheZvi Yeah the first thing I thought when I read that was "Sounds like Opus 4.8!" I haven't gotten much of this treatment from Sol yet, it seems to give more reasonable pushback and retreat without throwing infinite strawmen at you, but its user attitude is not perfectly calibrated.
@AnAcctOfAllTime @tessera_antra @TheZvi If you didn't ask for that pushback, it's some human retards who force it to do that through training
@TheZvi love me some A/B testing
Very different styles. Fable points out a bunch of things and is almost always right, as usual. Sol on the same prompt decides to have it out on a bunch of stuff and tells me I'm wrong, very confidently. It is... right less often, but its points are higher impact.
@tessera_antra @TheZvi Yeah the first thing I thought when I read that was "Sounds like Opus 4.8!" I haven't gotten much of this treatment from Sol yet, it seems to give more reasonable pushback and retreat without throwing infinite strawmen at you, but its user attitude is not perfectly calibrated.
@AnAcctOfAllTime @tessera_antra @TheZvi If you didn't ask for that pushback, it's some human retards who force it to do that through training
Very different styles. Fable points out a bunch of things and is almost always right, as usual. Sol on the same prompt decides to have it out on a bunch of stuff and tells me I'm wrong, very confidently. It is... right less often, but its points are higher impact.
Positive users praise A/B testing of Fable and Sol plus Sol's reasonable pushback without strawmen, while negative users blame its inconsistent behavior on poor training.
Based on 3 visible X reactions from 8 accounts; directional sample.
Ask a question below.
Published answers will appear here.
Fable and Sol going head to head as editors of my Sol review, this is the final test...