Hi! I'm Brian. Many people know me online as betaveros.
I'm currently (spring 2025) doing interpretability research at Anthropic — check out our newest papers, On the Biology of a Large Language Model and Circuit Tracing: Revealing Computational Graphs in Language Models, and other updates on Transformer Circuits Thread. Previously (2020–2022) I was building end-to-end encryption at Zoom — see our ASIACRYPT 2022 paper, Rotatable Zero Knowledge Sets; before that, I did a Masters of Engineering in computational music at MIT (2020).
I also love solving and writing puzzlehunt puzzles. I ran the 2024 Galactic Puzzle Hunt with ✈✈✈ Galactic Trendsetters ✈✈✈.
More broadly, my interests include information security, web development, UI design, and programming language theory/design/perversity; as well as singing, mystery visual novels, and dragons. 🐉
Personal Links
Projects and Creations
- noulith, a functional-ish programming language with too much syntax sugar, likely best known for the fact that I topped the Advent of Code 2022 leaderboard with it — here are my solutions
- gph-site and Puzzlord, Django apps for running and writing puzzlehunts, respectively; Puzzlord was built from scratch for the 2021 Mystery Hunt
- DP Puzzle Hunt, a puzzlehunt targeting beginners. All the puzzles are still solvable online, check them out!
- and more…
Blog highlights
Some combination of popular and personal favorites:
- Designing a Programming Language to Speedrun Advent of Code (2023)
- Introduction to Code Golf and Golflangs (2022)
- Interpreting Some Toy Neural Networks (2022)
- Just Enough Elliptic Curves to Be Dangerous (2021)
- Blogging Advice For People Exactly Like Me (2021)
- Advent of Code: How to Leaderboard (2020)
- C++ Rvalue References: The Unnecessarily Detailed Guide (2020)
- Olympiads: The Infinitely Overdue Retrospective (2019)
