/blog

Field notes from a terminal-first lab.

Engineering posts, release notes, retros from production. Written by the people who built the thing \u2014 never by marketing.

5 postsRSS · /feed.xml

May 23, 2026·2 min

Learn how KV cache works in LLMs, why it matters for AI-assisted coding, and how our caching system saves 85–98% on input tokens.