@deanwball @rustbeltjacobin Nowt wrong with tokenmaxxing
@HaydnBelfield @rustbeltjacobin I want to try this with the new approach I'm using which is maybe 50x more token intensive
DeepMind's Haydn Belfield dubbed this high-compute trend tokenmaxxing.
@deanwball @rustbeltjacobin Nowt wrong with tokenmaxxing
@HaydnBelfield @rustbeltjacobin I want to try this with the new approach I'm using which is maybe 50x more token intensive
@HaydnBelfield @rustbeltjacobin yeah I just blasted 4m tokens on a book but I think there’s substantial room to run. trying a new iteration of the agent org structure that may well get me closer to 20m.
@deanwball @rustbeltjacobin Nowt wrong with tokenmaxxing

@deanwball @rustbeltjacobin Interesting to think how in/efficient it is compared to eg interview transcripts, taking notes and writing drafts for a traditional book
Eg 4m-20m tokens for 10-100k words book
DeepMind's Haydn Belfield dubbed this high-compute trend tokenmaxxing.
@deanwball @rustbeltjacobin Nowt wrong with tokenmaxxing
@HaydnBelfield @rustbeltjacobin I want to try this with the new approach I'm using which is maybe 50x more token intensive