It's Dosu Drop Day! 🗄️ Claude Code, Codex, and Cursor logs already hold the answers your team has already. Dosu reads and writes what it finds to your Library. 🧵 go.dosu.dev/drop-aug26-x
We wanted to understand how coding-agent spend breaks down, so we replayed 60 public pull requests using Claude Code and Codex and analyzed 112 session logs. Planning and context-gathering accounted for 67% of the modeled API-rate spend in those sessions.
We launched a new OSS project today called Decant!
What did your coding agent do yesterday? Decant reads the logs Claude Code and Codex write and splits each run into categories of Context, Planning, Code, and Communicating. We'd love to hear what you think and what you learn
It's Dosu Drop Day ☀️ This month is addition by subtraction. We removed the waitlist, pared down our MCP to essentials, and helped your agents use fewer tokens. 🧵 go.dosu.dev/drop-jul26-x