/architect: Reduce Fable tokens by 80%, Fable orchestrates/reviews, Codex builds

Article excerpt

A developer has released an open-source tool called /architect that claims to reduce token usage by 80% when working with large language models. The system uses a three-stage orchestration pattern: Fable handles planning and reviews, Codex handles building, with the goal of optimizing API costs and efficiency when using AI coding assistants. The project gained modest traction on Hacker News with 47 points and 25 comments, suggesting interest from the developer community in token-efficient AI workflows.