Naomi Bashkansky resigned from OpenAI's model safety and alignment team on July 23 after 18 months, and the next day started as a founding researcher at Conduit, a San Francisco startup building AI models trained to convert human thought directly into text. Bashkansky, 23, announced the move on X on August 5 with a blog post titled "Why I'm Leaving OpenAI To Build Telepathy." Conduit trains models to read brain activity captured through a non-surgical headset and predict what a person was thinking, based on text or actions recorded at the same time. Bashkansky wrote that data collection must scale "by orders of magnitude beyond what has ever been done in academia." Her roadmap predicts that by 2027, she expects to direct AI coding agents by thinking instructions via headband; by 2030, frontier AI labs will train models interfacing with Conduit's neural representations, while Conduit moves from reading thoughts into "general write"; by 2035, AI "feels like a sixth sense and another limb." She cited three reasons for leaving: conviction Conduit could become "the general read and write company," a bet valued at over $1 trillion; regard for co-founders Rio Popper and Clem von Stengel; and research more open-ended than inside OpenAI. She is not the first researcher to leave a major lab; Ilya Sutskever left OpenAI in 2024 to start Safe Superint