AI is starting to improve itself: OpenAI announces a timeline for self-improvement, and on the same day its chief scientist called for an end to this process
On September 7, 2026, OpenAI made its latest internal data public, showing that one researcher was supported by three AI agents. This indicates that artificial intelligence is being deeply integrated into scientific research collaboration, allowing individual researchers to manage multiple agents to handle complex tasks, aiming to improve research efficiency and optimize work processes. On the same day (September 8), OpenAI announced its self-improvement timeline, but its chief scientist announced the suspension of related projects on the same day.
“The latest internal data from OpenAI shows that one researcher is now supported by 3 AI agents, indicating that AI is redefining the way researchers work. This internal data reveals the deep integration of artificial intelligence in scientific collaboration, with a single researcher able to manage multiple agents to handle complex tasks. This change aims to improve research efficiency and optimize workflows, reflecting a fundamental reshaping of traditional research models by technology.”