Crypto
Home›Crypto›Regulation›Conduit hires Naomi Bashkansky to build AI agents from…
Conduit hires Naomi Bashkansky to build AI agents from neural signals
Conduit says it has collected about 10,000 hours of neuro-language data, but it has not published aggregate performance for decoding free-form thought into prompts.
Former OpenAI alignment researcher Naomi Bashkansky said she resigned from OpenAI on July 23 and joined Conduit the following day as a founding researcher, according to CryptoSlate.
At Conduit, Bashkansky will work on models that translate non-invasive neural recordings into text meant to direct AI agents, a concept she describes as “telepathy.” In an Aug. 4 essay, she predicted a headband could decode rough intentions into prompts for an AI coding agent by 2027, with more ambitious “read and write” neural technology scenarios outlined for later years.
CryptoSlate reports that Conduit previously said it gathered roughly 10,000 hours of neuro-language data from thousands of participants, with people wearing multimodal headsets while typing, speaking, reading, or listening alongside a language model. The company published a limited set of claimed zero-shot examples, while excluding aggregate performance metrics, its full evaluation protocol, and third-party replication.
The report says Conduit’s next evidentiary hurdle is to establish public aggregate performance by connecting its 10,000-hour dataset to a portable system that can decode free-form thought. As of now, CryptoSlate notes the company has not shown that result.