miniblog.

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.

Today I learnt that Racket *intentionally* doesn't have a traditional REPL workflow. The authors were concerned about students not understanding the state between the current session and the code on disk. (Arguably Jupyter has some of these features now.) https://blog.racket-lang.org/2009/03/the-drscheme-repl-isnt-the-one-in-emacs.html

The word "agent" is so overloaded in the AI space. Sometimes it means a sophisticated interaction system, but other times it just means API. I think it's partly a sign of how new the space is. We don't have consensus on the best way to use these systems yet.