ive yet to see a post like this that wasnt bullshit. share the texts bud! your claims are fully verifiable
Replying to @LincolnMargison @victormustar
bro what trend did you expect to see on that line
Replying to @cremieuxrecueil
this has been an issue in codex for like 6 months
Replying to @NickADobos @github
imagine if you didnt have thousands of followers
Replying to @larakiara
xitter.cf/gradientpull/status/20… were so back
Replying to @januarycomputer
are you sure though? its only 2x cheaper on input tokens compared to batch luna, and actually 4x more expensive on cached input.
theres no way you could possibly get near 1000x cheaper, your outputs on luna would have to be long as fuck.
i recently ran a comparison on something like this and got 20% cheaper (and worse accuracy)
Replying to @basedjensen @ctjlewis
the algorithm really does let you choose if you want to be retarded or not








