Astra writes code, Claude Code reviews it, all inside Codex
The author embedded Claude Code inside Codex so Astra writes a rate limiter and Claude flags issues; Astra scores strongly but still has blind spots.
GPT-6 Astra is here so I tried it as my primary engineer. Astra writes the code. Claude Code reviews it. Inside Codex. One prompt. Back in July OpenAI shipped a plugin that puts Codex inside Claude Code. I built the reverse: Claude Code inside Codex. Astra builds a rate limiter, hands it to claude, claude gives out findings that astra can then accept/reject, fix and ship. Honest take after a day on it: Astra is a monster. Beats Fable 5.1 on DeepSWE, Terminal-Bench, Terminal-Bench Science by 12 points, doubles every published Claude score on ARC-AGI 3, and does it at about half the cost per task. That is why it is the engineer here. But Claude still tops the independent Coding Agent Index, and in this run it caught things Astra wrote and walked right past. Different lab, different blind spots.