Home

A rough hand-drawn portrait

Hey there, I'm albro. I'm an "agentic-engineer" most of the time, but I used to be a programmer too (•◡•)

I've been obsessed with AI since late 2024. But DeepSeek R1 is what really spiked my interest. Conversing with a local-sized distill on my personal PC was my first deeply engaged conversation with an LLM, I think

I'd describe myself as charmed by these things. They occupy a lot of my brain space. How could they not? Since then, I've occupied myself with trying to understand them. That's what drew me to interpretability

But I don't get to do as much of that as I'd like. Mostly, I'm a codeslopper

I've also done some post-training and harness engineering independently

This is a place where I keep some of my writing, projects, and artifacts. I also draw regularly, so expect some of that to end up here and elsewhere

sections

  1. 01writings
  2. 02projects
  3. 03artifacts

recent things

projectopen-artifactsHTML artifacts with comments for CodexartifactPrime Agent, AbridgedAn abridged version of the Prime Agent paper, with added orientation, adapted for casual readingartifactWhat RL does to the characterShort visualization of RL's effects on model persona