hi
currently an independent researcher working on ai safety via interpretability1. i’m interested in approaches that search for the true features, and this might look different from more standard approaches. some examples to illustrate: this, this, and this. feel free to add me on discord @ ajskateboarder if you’re interested in similar topics (or any normal topics idk), or email me at: me [at] [my name] [dot] com
i post on lesswrong, less thought-out stuff in notes
aging github (i miss programming…)
i don’t skateboard but i think the rest of my username is pretty accurate
(updated september 2026)
since ~october 2025. i’ve been aware of the basic case for agi since 2022, and started taking it more seriously in 2025 ↩