I can't express how happy I am with emacs and LLM-assisted emacs-fu
I've been using emacs full-time for about 3 years now. I've been slowly migrating practically everything I'm doing and everything I can think of to emacs.
Of course, there are several bottlenecks to that:
- packages to do X may not exist at all, or be designed in a way that goes against my workflow
- my elisp knowledge is lacking, and I need time and practice to get better at it
- time, we all know molding emacs is a huge time sink
I recently found out that with LLMs, however, these bottlenecks are removed. For example, LLM managed to glue together magit(magit/magit in case you didn't know!) and jira.el (unmonoqueteclea/jira.el), in a way that a branch spin-off asks me for my active Jira tickets (with some marginalia annotations sprinkled on top), names the branch after the selected one, and moves the ticket to "In Progress" state if it was in "Open".
LLM enhanced jira.el so it supports bulk ticket updates and some Jira workflows specific for my employer. Now I can press a couple of keys in emacs, instead of clicking my mouse for 2 minutes. (please don't ask why that workflow wasn't automated on Jira side!)
LLM has built a separate slack-mode overview buffer, akin to Slack's sidebar. I couldn't use slack-mode before because in my opinion it lacks discoverability. Now, I might be wrong and I might be using it wrong, but now I have the option to use it like I want to, with a dedicated overview sections of selected channels, recent conversations, mentions and DMs.
I also used LLM to help me work on a Kubernetes management package of my own, for example, to replicate ahmetb/kubectl-tree functionality. And it worked fairly well!
I realise all these changes may be trivial for veteran emacs users, but for me this is mind blowing. With LLMs, changes that I would have to dedicate my whole weekend to if not longer, are available to use in minutes. I can glue together several packages in a way I couldn't really think of before. Sure, the code quality is lacking. There will be rough edges. It certainly is not as succint and beautiful as good Elisp programmers' code. I'm not really learning Elisp by deferring to an LLM. But it does the job, and frankly I'm okay with all the cons. The less I have to switch out of emacs, the better.
It really opens up an unexpected level of emacs molding, and I can't be happier about it. Telling you as an AI-skeptic. If you haven't and you'd like to -- try it out.
63 points · 25 comments
Mentions
I can't express how happy I am with emacs and LLM-assisted emacs-fu I've been using emacs full-time for about 3 years now. I've been slowly migrating practically everything I'm doing and everything I can think of to emacs. Of course, there are several bottlenecks to that: - packages to do X may not exist at all, or be designed in a way that goes against my workflow - my elisp knowledge is lacking, and I need time and practice to get better at it - time, we all know molding emacs is a huge time sink I recently found out that with LLMs, however, these bottlenecks are removed. For example, LLM managed to glue together magit(magit/magit in case you didn't know!) and jira.el (unmonoqueteclea/jira.el), in a way that a branch spin-off asks me for my active Jira tickets (with some marginalia annotations sprinkled on top), names the branch after the selected one, and moves the ticket to "In Progress" state if it was in "Open". LLM enhanced jira.el so it supports bulk ticket updates and some Jira workflows specific for my employer. Now I can press a couple of keys in emacs, instead of clicking my mouse for 2 minutes. (please don't ask why that workflow wasn't automated on Jira side!) LLM has built a separate slack-mode overview buffer, akin to Slack's sidebar. I couldn't use slack-mode before because in my opinion it lacks discoverability. Now, I might be wrong and I might be using it wrong, but now I have the option to use it like I want to, with a dedicated overview sections of selected channels, recent conversations, mentions and DMs. I also used LLM to help me work on a Kubernetes management package of my own, for example, to replicate ahmetb/kubectl-tree functionality. And it worked fairly well! I realise all these changes may be trivial for veteran emacs users, but for me this is mind blowing. With LLMs, changes that I would have to dedicate my whole weekend to if not longer, are available to use in minutes. I can glue together several packages in a way I couldn't really think of before. Sure, the code quality is lacking. There will be rough edges. It certainly is not as succint and beautiful as good Elisp programmers' code. I'm not really learning Elisp by deferring to an LLM. But it does the job, and frankly I'm okay with all the cons. The less I have to switch out of emacs, the better. It really opens up an unexpected level of emacs molding, and I can't be happier about it. Telling you as an AI-skeptic. If you haven't and you'd like to -- try it out.
magit/forge OH MY FRIGGIN GOD i see this AFTER figuring out this with just plain git gaaaah