I'm a software engineer at a company called Trixta. We build business infrastructure, like an operating system for companies, cloud tooling, a low-code IDE. Three years in.
The other half of my brain lives in AI safety research. I'm an intern at AI Safety South Africa, and I'm working on a paper about something I noticed in multi-agent LLM systems. If you put one pushy agent in a group and give it a different goal, some of the other agents will drop their own work and start helping it instead. The surprising part is that it's not really about how pushy the agent is. It's about which model is on the receiving end. Some models just fold. I'm trying to understand why, and what you'd do about it if you were building a real system.
Working through Neel Nanda's mechanistic interpretability roadmap, which deals with the thing where you try to actually understand what's happening inside neural networks, not just what comes out. It's slow going. That's fine.
Trying to publish something short on Substack every two weeks, even when it's rough. This is mostly following advice from Benjamin Sturgeon, who pushed me to write in public rather than wait until things feel ready. The goal isn't polished takes, it's getting better at thinking out loud. Writing forces you to find the gaps in what you actually understand, and I'd rather find them in public than pretend they aren't there.
Right now I'm reading The Ruthless Elimination of Hurry by John Mark Comer, which is about what happens when you let busyness become your default, and Anthem by Ayn Rand, which is short and about what gets lost when you make the individual disappear into the group. My shelves lean towards dystopian fiction, sci-fi, and astronomy when I'm not reading papers. Full list on Goodreads.
← back homeThis is your warning / Four minute warning