OpenAI researcher Naomi Bashkansky leaves company to build ‘telepathy’ at Conduit

### OpenAI Researcher Naomi Bashkansky Joins Conduit to Develop Thought-to-Text AI

OpenAI researcher Naomi Bashkansky has left the artificial intelligence company to join San Francisco-based startup Conduit as a founding researcher, taking on a role focused on the development of brain-computer interface technology. Bashkansky announced her move on X, revealing that Conduit is working on what it describes as telepathy through thought-to-text AI models trained using non-invasive neural data. The startup aims to create technology that can interpret neural signals and translate a person’s thoughts into written language without requiring surgical implants.

In a blog post discussing her vision for the future, Bashkansky predicted that within the next few years, interacting with artificial intelligence through thoughts could become a major form of human-computer communication. She argued that future AI systems could become a natural extension of their users rather than functioning simply as highly capable digital assistants or automated coworkers. Bashkansky suggested that this approach could create a closer relationship between humans and AI, allowing technology to directly enhance human abilities instead of replacing them.

She also outlined an ambitious timeline for the development of the technology, suggesting that by 2030 the broader technology industry could increasingly adopt and expand telepathic interfaces. Looking further ahead, she predicted that by 2035 the distinction between humans and artificial intelligence could become significantly less pronounced. Her comments echo broader predictions from technology leaders who have discussed the possibility of increasingly direct connections between human brains and AI systems.

Bashkansky emphasized that the eventual impact of superhuman AI will depend partly on decisions society makes about how these systems are developed and deployed. She described her preferred vision as one in which humans remain directly involved and AI is used to strengthen human capabilities rather than make people obsolete. Conduit is developing its brain-computer interface without requiring users to undergo surgery or receive implanted electrodes. Instead, the startup says its technology relies on a headset capable of detecting neural signals externally and converting them into text.

The company’s approach places it within a growing field of research focused on making brain-computer interfaces more accessible by eliminating the need for invasive medical procedures. Conduit was cofounded by Rio Popper and CEO Clem von Stengel, who have backgrounds in academic research and technology. The startup says its team has already spent approximately six months collecting around 10,000 hours of neuro-language data from thousands of participants.

Conduit describes the dataset as one of the largest collections of its type, potentially providing the foundation for developing more sophisticated thought-to-text systems. However, the company’s claims have also generated skepticism online, particularly over the description of the technology as “non-invasive.” Some users have questioned whether any technology designed to interpret neural activity should be characterized as completely non-invasive, highlighting concerns about privacy and the sensitivity of brain-derived information.

The ability to translate neural signals into language could raise important questions about consent, data ownership, security and how such information might be used in the future. Conduit has not publicly announced a funding round or disclosed a confirmed valuation, leaving the startup’s financial position unclear as it develops its technology. Bashkansky’s move nevertheless highlights the growing interest among AI researchers in combining machine learning with neuroscience.

Beyond her work in artificial intelligence, Bashkansky is also known for her achievements in competitive chess, having earned the Woman International Master title and previously won a world school championship in the girls’ under-13 category. Her transition from OpenAI to Conduit signals a growing focus on technologies that seek to move AI interaction beyond traditional keyboards, screens and voice interfaces toward direct communication between human thought and machines.