Igor's Substack
Igor Pereverzev
Independent AI-safety research on how AI systems hide communication from monitors. Found covert channels evade detection by translating latent geometry, not obfuscating. Extending from spiking neural nets to brain-computer interfaces.