On day 12 of OpenAI's 28-day consecutive release plan, OpenAI made context and request handling optimizations to Codex and ChatGPT Work, including reusing context across sub-agents, extending cache duration, and improving request routing to increase cache hit rates.