Comment by koiueo

4 months ago

These days people rather switch to a browser window, open an LLM of their choice in a new tab and in verbose English ask "how do I do X in this popular program Y?".

Then get a hallucinated answer and come to you to complain about a missing cli option, while it's literally there, in their terminal, just one -h away. True story (had to vent out, thanks for listening).

Hey, now.

When I want to ask an LLM how to do something in emacs I `SPC $ g g` and ask it in a gptel buffer.

  • The only 2¢ I can add here is that LLMs are surprisingly good for solving tasks that involve Elisp. There's large corpus of Emacs Lisp in the wild - the amount of it on GitHub alone is shocking.

    For comparison - whenever I try using a model to write some Neovim config stuff, LLMs hallucinate badly.

    Using Emacs these days is so much fun - you just ask a model and you can immediately try things - not only without restarting - you don't even have to save anything, anywhere.

    You can even make Emacs do things that involve tools and languages that have nothing to do with elisp, e.g., "write elisp that would open all nested dirs in a given folder, and then run magit-log for each project, searching for specific pattern... and if found, issue npm or uv pip install with arguments...", etc.

    • This matches my experience as well.

      I'm a Spacemacs user, and Claude knows the difference between Spacemacs and stock emacs. It basically wrote tools for itself to hook into the emacs LSP package and look up C++ code (in our codebase, which is, uh, challenging for FOSS language servers) from gptel. I was completely fucking unable to do this via publicly-available MCP servers in the VSCode agent mode. In general I have been able to resolve a bunch of niggling config issues that I had just been ignoring in favor of doing Actual Work.

      As you said, I can evaluate the elisp Claude wrote right there in the gptel buffer, try it out, and iterate before actually pasting it into .spacemacs. It it doesn't work, Claude knows a ton of debugging tricks I can try. trace-function, advice to dump a backtrace, etc. It knows how to do everything in org-mode. Super helpful. Way better than the AI Assistant that my employer bundles with our product (which is outperformed by the Google AI Overview).

      2 replies →