Independent AI safety researcher, UK. Linguistics background (BA Hons). Seven years self-employed building technical systems. Published the first empirical evidence of vocabulary-activation correspondence in LLMs. I find gaps others miss and I ship results without institutional support.