FOD#162: Did OpenAI’s Agents Start Recursively Self-Improving?

How separate OpenAI agent runs built shared memory, and the missing training detail that determines whether this was recursive self-improvement