An OpenAI Alignment Researcher Left After 18 Months to Build Non-Invasive Thought-to-Text Models at Conduit
naomibashkansky.com / Hacker News·low signal
Naomi Bashkansky wrote on 2026-08-04 that she spent a year and a half on OpenAI's alignment team, including the AGI onboarding presentation and the alignment blog, and is leaving to be a founding researcher at Conduit, which trains thought-to-text models on non-invasive neural data. The stated ambition is scaling non-invasive neural data collection far past prior academic efforts so people can interface with AI systems directly by thought. 129 points and 206 comments on HN, where the reaction focused on the alignment-to-BCI career path.