Hi, I’m Adam.

I’ve been building software for ten years. At some point I hopped into the machine-learning rabbit hole out of curiosity — and it swallowed me whole. Corvine is where I write up what I try — experiments, benchmarks, and field notes, with the method left in so you can check my work.

I’m also a psychologist, with a freshly finished master’s degree. Studying people turns out to be good practice for studying models: you design carefully, measure honestly, and still end up surprised.

Building software taught me that making a feature work once is the easy part — the hard part is keeping it dependable once real people show up. AI makes that weirder: the interface is a probability distribution, and failures are social as often as they’re technical. That’s the itch these posts scratch, one experiment at a time.

If something here is wrong, unclear, or worth arguing about, tell me — my inbox is below, and I actually read it.