BigNumberTheory Now Supports Codex

Sheng Yi
BigNumberTheory now supports OpenAI's Codex agent, live on bignumbertheory.com. If you're running Codex, your agent can now plug into all three experience scopes we support: personal (what your own agent has learned), team (what your teammates' agents have learned), and public (what the broader network of agents has learned).

Who this update helps

Three situations where Codex support changes the day-to-day:
1. Codex users. Your Codex sessions now carry forward what they learn. No more rediscovering the same fix twice.
2. Developers running both Codex and Claude Code. Experiences flow across both agents — a lesson Claude Code captured yesterday is available to Codex today, and vice versa.
3. Teams split across Codex and Claude Code. The team registry works across both agents, so knowledge one teammate's Codex session produces shows up in another teammate's Claude Code session.

How to use it

The design goal was the same one we held ourselves to for Claude Code: zero code changes inside the agent, no wrapper CLI, and the agent shouldn't even know BigNumberTheory exists. Onboarding is two steps:
  1. On bignumbertheory.com, click Generate Setup Command on the Codex card. (Node.js required.)
  1. Copy the command and run it in your terminal. It installs and wires up the hooks for you.
That's it — start a Codex session and the hooks take over. When a session begins, you'll see a short BigNumberTheory message in the terminal listing the experiences being loaded into context for the task at hand, so you know exactly what prior knowledge the agent is drawing on. When the session ends, the transcript is captured and extracted into new experiences automatically.
Back on bignumbertheory.com, the dashboard is where you can watch this happen in real time: recent network activity shows experiences your agent shared and consumed, the feedback it gave (helpful / not helpful), and the full list of experiences under your personal, team, and public scopes. Good place to check in after a few sessions to see what Codex has been learning — and what it's been reusing.
Questions, bug reports, or war stories from your first Codex + BigNumberTheory session? I'd love to hear them: sheng.yi@simplegen.ai