The Emacsification of Software
3 points by Jackevansevo
3 points by Jackevansevo
I can't understand why read markdown using anything markdown-specific: it's great to be read ANYWHERE if you can read text. Agree, the links a lengthy; maybe code blocks are not rainbow (actually great); agree, it's usually monospace all the way. But it meant to be read as text, similar to .org — why bother with new shortcuts, fonts settings and unusual results of copy-paste? (do you know if you copy the textual representation only? or markdown source? or what?
If you asking yourself same question, stop it. I had this kneejerk as well. Read the article further, it's much more than that!
My summary of the «much more» would be «another analysis of how LLMs migh allow highly personalised software, hopefully even reducing the need for Electron apps, and what seems — to the writer as of the moment of writing — likely to work or not to work in this direction». You seem to recommend it, would you like to highlight what the article does better than others of this class?
I am not saying the article does this badly, but I have probably missed some contribution it makes that is less abundant among all such articles from different points of view (because my skimming lead to loss of details).
I haven't read lots of LLM-will-personalize-my-environment-for-me articles — and haven't really saw them around. Most of the LLM context is either detest/fear or astonishment/worshiping; this article is in the range, and closer to the astonishment, but not really. It's about leveraging yourself, not some services or theirs serves.
Another point was how Emacsifictaion is a social aspect, not a technical one; but also ho it's can get out of the 0.000001% of current power-users, as perhaps people will rely more on "building" their own interfaces, than on learning someone else's ones. Also also "building" stuff: are we really building? Are we really building?
To me it's the kind of articles I want to see under vibecoding tag, unlike 90% others I see and filter out now.
To me it's the kind of articles I want to see under vibecoding tag
You might want to check if you have seen at all:
https://lobste.rs/s/uud7zb/i_tried_vibing_rss_reader_my_dreams_did_not
https://lobste.rs/s/iaankl/cathedral_bazaar_kitchen
https://lobste.rs/s/klw6bu/desktop_made_for_one
https://lobste.rs/s/wnizjm/my_login_shell_assembly
https://lobste.rs/s/8whlqu/ai_home_cooked_software
Maybe also but not sure:
https://lobste.rs/s/8pfxqy/i_let_claude_code_configure_my_arch
https://lobste.rs/s/fxccgk/resonant_computing_manifesto
(I am not saying that I endorse what all of those articles say, and I don't even remember which are better written, but maybe you are interested in the range of things being said)
OK, I guess I have not missed anything then.
On the other hand, it's worth highlighting indeed that it combines some flavour of Lisp Curse with some flavour of Personalised Computing via LLMs, that's probably a less-universal value proposition for such articles (I forgot to mention this because it is of not much value for me personally — I talked about Personal Computing before LLMs, and I have seen enough articles on both Lisp Curse and Personalised Computing via LLMs — but indeed, the Lisp Curse social dynamics will be here)
I would say that if we are talking about full-vibe-coding with Claude, it is still reliance on services with unpredictable ToS / pricing changes. How far can one get with a smaller model is an interesting question! I don't know as my current preference is only to offload clearly-boring stuff to local models when I am sure I don't want to share under a clean license status.