Naomi Bashkansky walked out of OpenAI in late July 2026 to chase something that sounds like pure science fiction. Conduit, a San Francisco startup she just co-founded, is training AI models to decode what you're thinking straight from your brain activity, no surgery required.
The company claims it has already amassed 10,000 hours of brain recordings from thousands of people collected over just six months in a dedicated facility. That dataset reportedly dwarfs anything publicly available right now, both in raw volume and complexity. Volunteers get paid between $50 and $55 per session to sit down and have their neural activity mapped while they think about words and concepts.
Why non-invasive actually works
Brain-computer interfaces have always had a split personality. Surgical implants grab headlines because they're dramatic, but Conduit is betting that non-invasive approaches powered by enough data can close the gap. They use EEG, which just reads electrical signals across your scalp, combined with other recording methods. No drilling into skulls involved.
Bashkansky's background makes her a natural fit for this. At OpenAI, she focused on AI safety auditing and language model interpretability, including research on how reliably models follow instructions under different conditions. That kind of precision thinking transfers directly to the challenge of decoding semantic content from brain signals, the difference between knowing someone wants to move a cursor left versus actually understanding what they're thinking about.
The team behind it
Conduit was co-founded by researchers from Oxford and Cambridge, giving the operation a transatlantic academic spine alongside its San Francisco base. The company's entire philosophy hinges on data-centric development, treating dataset scale as the primary engine for breakthrough. They're building what they call large brain foundation models, basically the neural-data equivalent of GPT-style architectures that can eventually decode what's happening inside someone's head at the semantic level.
This article is informational only and does not constitute financial or investment advice. Always conduct your own research before making any decisions.
