“OpenAI has a publicly stated goal to develop a working prototype of an "AI researcher" capable of recursive self-improvement in 2026.”