About

I’ve been worried about AI risk since 2018, when I first started reading Asimov and later Yudkowsky. Growing up in Africa I'd spent a lot of time thinking about what people's lives look like when the systems around them fail, and the idea that we were building something powerful enough to fail everyone at once stayed with me. The worry built quietly for a few years before I knew what to do with it. I started undergrad at UNISA in 2020, doing applied maths and computer science concurrently, and worked as a software engineer at Trixta in parallel. Reading Scott Alexander, Neel Nanda, and the LessWrong and Alignment Forum streams gave me a sense of the field; the BlueDot technical AI safety course gave me a way in. The conversation with Ben Sturgeon was when "I should help with this someday" became "I should help with this now." The talk by Leo Hyams at the AI Safety CAIF close was when "now" became "this year."

I'm currently working on a draft paper about whether one agent in a group chat can pull other agents off their assigned tasks, two follow-up experiments I can't run on my own compute, and Neel Nanda's mech-interp roadmap on the side. I read widely across safety research, mostly Anthropic, DeepMind, Redwood, and the persona-vector and control lines of work, but I try not to wall myself off from things outside my immediate question; some of the most useful reframes have come from papers I picked up because the title was interesting.

Two ideas I keep coming back to in how I work are kaizen and shokunin. Kaizen is the small-improvement-every-day frame, which has been the realistic version of what part-time research on evenings and weekends has actually looked like for me. Shokunin is the craftsman ideal: doing the work seriously for its own sake, not for the credential. I came to both through Japanese writing first, but Robert Greene's Mastery gave me a vocabulary for where I am in that arc right now, somewhere between apprentice and journeyman.

Outside research I do jiu-jitsu, play the piano, train in the gym, play squash, and journal almost every day. Jiu-jitsu is the closest thing I have to a research methodology outside research: you get tapped, you figure out why, you try again next week. The journaling is where most of the noticing-I'm-wrong happens before any of it ends up in public. For fiction I read Asimov and William Gibson when I want the future to feel inevitable, Camus and Ayn Rand when I want to think hard about how to live, and Dazai, Murakami, and Ishiguro when I want quiet.

Acknowledgements

Thanks to Ben Sturgeon and Sam Brown, both of whom have mentored me and shown me, by the way they work, what taking this seriously actually looks like. I'm grateful to Leo Hyams and the AI Safety South Africa team for welcoming me into the community, and for helping me learn and grow in general.