ai security researcher

kirill novikov.

i work on cybersecurity and ai safety, mostly ai control and the security of ai agents.

tokens through claude code, per day. the orange one is the busiest. by model

writing

coming soon. write-ups of fixed bugs and research notes will appear here.