miniblog.

One subtle behaviour of Claude that wasn't obvious to me: whilst each conversation is transient, permissions persist across conversations. So if you've given permission to run e.g. 'cargo test' or even 'cargo run', you need to be sure that all future invocations are safe too. You can see the current permissions with /permissions.

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.

The lisp model of programming is generally: write a function, evaluate it, interactively call it with some arguments, iterate. Jupyter notebooks are similar. Why not automatically evaluate definitions (not expressions) whilst working? It seems like it could be a satisfying way to work with code.