OpenAI hits automated research intern milestone and targets 2028 AI researcher

OpenAI has now publicly confirmed that it hit its long signposted goal of an automated AI research “intern,” and it is openly planning for a more autonomous AI researcher by March 2028. The accompanying data shows agents are no longer a side tool but the main engine of day to day research work inside the lab, measured in multiple agent workdays per human workday. This is one of the clearest quantitative signals so far that AI is starting to accelerate its own development loop rather than just speeding up end user tasks.

Jakub Pachocki’s “An Alien Mind” essay makes this even more consequential. He frames automated research not just as an efficiency play, but as the pathway to recursive self improvement, and simultaneously questions whether any lab is ready to keep such systems aligned. The post explicitly ties further scaling to the ability to monitor agent behavior in realistic environments and to slow or stop when safety bars are not met. For competitors, this raises the bar on both capability and transparency: frontier status is no longer just about model benchmarks but about how far you have pushed automated R&D. For the broader race to AGI, it marks a shift from speculation about self improving systems to a concrete internal milestone that one of the leading labs says it has already crossed.