Codex usage reportedly returns to normal after configuration changes
ThursdAI says a Codex user's usage returned to normal after removing a 1M context-window setting and moving sub-agents to a cheaper model.
TLDR
ThursdAI describes a Codex user whose configuration had a 1M context window, which it says “kills OpenAI's cache.” According to its account, usage returned to normal after the user removed that setting and moved sub-agents to a cheaper model. It recommends checking codex.toml for anyone running through their limits.
Combined views
—
2 Sources, first seen ago
