← All posts



2 posts

Anthropic is subsidizing our AI coding at 21x. How long will it last?
We measured what our team's Claude Code usage would cost at API prices. It runs about 21x our seat price on average, and 113x for our heaviest engineer. Here are the numbers, how we measure them, and the script to measure your own.

Introducing Modelplane: the control plane for AI inference
Today we're open sourcing Modelplane, a control plane that operates AI inference across a fleet of GPU clusters, on cloud, neocloud, and on-premise, as one inference platform.