A 30-year software veteran shares a dual-agent workflow built after 2 years of AI-assisted coding, where one agent (Claude Code) writes code and another (Codex CLI) reviews it in a structured loop. Inspired by pair programming and the Ralph Loop concept, the author automated what was previously a manual cross-agent review
Table of contents
The problems I kept findingPair programming, but with AIsAutomating the loopThe real test: my first open source PRIndependent convergenceLimitationsThe deeper questionJust shipped v0.3.0Sort: