OpenAI is throwing everything into building a fully automated researcher

OpenAI is shifting its research focus toward building a fully automated AI researcher that can tackle complex problems independently, with plans to launch an "AI research intern" by September 2025 and a complete multi-agent research system by 2028. Chief scientist Jakub Pachocki believes current tools like Codex demonstrate the foundation is in place, as OpenAI staff increasingly use AI agents to handle coding and research tasks that previously required human intervention. While the technology shows promise in solving math and science problems, significant safety concerns remain about autonomous systems operating with minimal oversight, prompting OpenAI to develop monitoring techniques to prevent potential misuse.