miniblog.

I'm having fun writing a simple type checker, but I'm learning firsthand why syntax-directed checking doesn't work. It prevents inference. My checker catches real bugs, but it can't handle cases like this: [1, 2].map(fun(x) { x + 1; }) I think I need bidirectional checking.

I've dabbled with 'conventional commits' for a personal project but I found they slowed me down. It's not always easy to categorise a commit as a fix, a chore etc. Sometimes refactorings also fix bugs. Do you use them? I can imagine a large, mature project benefitting more.

Whilst LLMs don't always give an accurate answer, the UI is really compelling. I keep finding users whose favourite way of doing research is an LLM.