OpenAI’s chief scientist writes about minds we don’t fully understand.

Jakub Pachocki is OpenAI’s chief scientist. On Sept. 6 he wrote about minds his own lab does not understand. The essay is “An Alien Mind.”

He goes back to a night in 2023. On a project called RLSlow, he and a colleague, Szymon, stayed at the office. The numbers were good. What sat with them, he writes, was that machines meaningfully smarter than people would arrive in their lifetime.

Three years on, he says those reasoning models run computers, work with people, and carry out research. They are also changing computer security, “and in that present clear new dangers.” He writes: “Based on internal results, I have a strong expectation that this speed of progress could be sustained into recursive self-improvement.”

“This is a time that calls for extreme caution,” he says. “I am concerned no one is prepared for the consequences of a continued rapid rise in machine intelligence.” He closes: “Currently I believe that no lab has solved alignment and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer.”

The same day the lab posted “Research acceleration: The view inside OpenAI.” Coding agents, it said, are reshaping the daily work of its researchers. “According to our measurements, we have now reached the goal, announced last fall, of having an automated research intern by September of this year.” By that it means “a system that can carry out well-defined research tasks under human direction, including tasks that would take a skilled researcher a few days.”

As of mid-August, the post says, the research shop used 3.1 agent-workdays of effort for every workday of human labor, on a standard eight-hour day. People still set the priorities. No new frontier model identifier launched in this window.