A framework for structuring AI-assisted development into three feedback loops: the inner loop (fast execution with focused AI prompts and multi-agent setups), the middle loop (managing project state, task handoffs, memory files like Agents.md, and work sessions), and the outer loop (long-term architectural planning and API design). The system draws from Steve Yegge and Gene Kim's 'Vibe Coding' book and includes practical tips like using CLI task managers, leaving background work for AI overnight, and maintaining modular architecture so agents can work in parallel without conflicts.

Sort: