[tensorlibb]

They also never disclosed they were load balancing with nvidia GPUs and Amazon aws trainium inference chips until people complained.