miniblog.

I think you could build an interesting IDE with a tiny embedded LLM in addition to the usual tooling. Features like 'extract method' would be much nicer if an LLM could provide a name. Choosing a good name is virtually impossible from just a typed AST.

I really like the MELPA model of packaging directly from git. It solves the problem of forgetting to release something -- just merge a PR and you're done. It also makes version number bumps much less important. You could go even further in a statically typed language and also figure out when breaking changes occur.

Noodling with an interpreter for a statically typed language with reified types (e.g. a list knows what type it contains). Currently I have a single representation of types in both the runtime and the type checker. I think that's a good thing?