Kamimoto(かみもと)
@sep_is_heim · Sep 20
A Japanese developer let Jev pick a sparse-attention rate for each layer during video generation, cutting a run from 6m07s to 3m34s on an RTX 4070.
TypeSafe AI· Released Sep 15
Kamimoto(かみもと)
@sep_is_heim · Sep 20
A Japanese developer let Jev pick a sparse-attention rate for each layer during video generation, cutting a run from 6m07s to 3m34s on an RTX 4070.
Christopher Lee
@chris_not_busy · Sep 22
Chris built a clinical simulation where Grok proposes moves and Jev evaluates them at speed, exploring over twenty possible futures for a cancer patient in minutes.