DHH's trick: make Claude and Codex review each other via CLI—no fresh context needed
zeeg · x · 2026-10-08
DHH shares his adversarial code review technique: while driving in Codex, he prompts "Review this with claude"; in Claude, "Review this with codex". The models know how to kick off reviews using each other's CLI, take turns, and settle arguments—no magic involved. zeeg adds that you don't even need a separate model or context window: asking in your existing session yields similar results. Compaction paranoia is unnecessary, and "fresh context" often doesn't help.
More from coding & agent
- Agent harness study: prompts and tool schemas drive 3x cost gap on identical models — dair_ai · 2026-10-08
- Anthropic, Google, OpenAI and 7 more offer free official AI courses, incl. full MCP track — anthara_ai · 2026-10-08
- Haiku 5.5 costs 20x less than Sonnet 5.5, dev uses it for parallel sub-agents — RLanceMartin · 2026-10-08
- MCP server drives a real vehicle CAN bus — agents can read and transmit live frames — sa1vatorre · 2026-10-08
- ServeLearnBench: agents still show large learning gaps in continual self-improvement — BeidiChen · 2026-10-08
- Cloud agent runtime plus Tailscale: the "local hands" setup devs prefer — blelbach · 2026-10-08