Comment by dannyw

OpenAI has been doing wonky stuff with subagents, including encrypting the prompts sent to subagents in Codex. Who knows what’s really going on.