swyx burned 20B+ tokens on Astra and reports about $6/hour, rising fast with parallelism.

The practical finding from the Latent Space hands-on is coherence across billions of tokens in a single agent thread, plus the ability to choose and train models, label data for active learning, instrument and read logs, and command fleets of subagents running other models. The $6/hour figure comes from 33 tokens per second at $50 per million max rate. Their caveat is the one to internalize: on Ultra with 20-50 parallel agents the hourly cost goes far past $6 precisely because the model parallelizes well. Artificial Analysis independently confirms Astra is more token-efficient than Sol and Fable. Latent Space