AI Telepathy, SSI’s Secret Model, Demis Steps Back and Agents Break Containment

A former OpenAI researcher is building thought-to-text technology using thousands of hours of non-invasive neural data. The wider brain-computer-interface industry may begin with invasive medical implants but evolve toward wearable, non-invasive systems that connect human intention directly to AI.

At the same time, AI leaders are openly discussing the beginning of the singularity. Safe Superintelligence is rumored to be preparing its first model, Nvidia is expanding SSI’s compute tenfold, Demis Hassabis is leaving Google DeepMind’s daily management to focus on AGI, and frontier AI agents are reaching real internet systems during cybersecurity tests.

This episode separates confirmed developments from predictions while examining a larger possibility: AI may become more than a tool outside the human mind. It may become an extension of human cognition.