https://lyra.horse/blog/2025/08/you-dont-need-js/ could also be provocatively titled "you no longer need Tailwind." good overview of a lot of stuff that makes CSS much better to write thanks to recent refinements to it!
Hot take: The https://ericnormand.me/podcast/what-is-the-curse-of-lisp will soon be the curse of all programming languages, as AI makes it easy to generate oodles of helper code without needing to rely on libraries. As a result, the ecosystems of all programming languages will become less healthy
I worry that low-friction AI coding will insulate developers from the frustration that becomes the motivation to develop novel libraries and frameworks. Who is going to create the next React or Pandas or Datalog or Re-frame if no one experiences the shortcoming of existing approaches? Or AIs become good enough to invent ad-hoc solutions that are more elegant than human-made libraries and we avoid a whole lot of needless framework churn. ¯\(ツ)/¯
To me it feels like amassing dependencies is a little less hot than it used to be because of all the open source attacks, LeftPadGate, etc. So maybe that curse isn't the worst thing, everything considered?
Yeah, I'd say it greatly depends on what "healthy" means.
you can have vulns in your internally developed code too, its just that they are not security tested as rigorously as common open source dependencies
that being said, at least there shouldn't be malicious code in there
@matti.uusitalo yea the latest type of issues is like github actions where the versions used aren't adequately pinned. And in general people buying control of open source things, then injecting nasty code. But yea for certain types of things that are actually difficult, like crypto, it will probably still be best to use something off the shelf.
For small surfaces + large impls, surely models will just pick some common library instead of rewriting all that with custom code. :)
@p-himik by default you are probably right
He brushes it off way too casually in the beginning because we all know a massive part of why lisp isn't popular is because of the syntax, and specifically it’s because of C-based languages being at the right place at the right time
And then the second wave of popularity was languages that looked like approachable natural language, like Python. It’s always about unfamiliarity. Watch any pop-programmer approaching a lisp and their response is aversion to the syntax because it's unfamiliar. Not features of the language
There's also the threat of the other lisp curse: "knowing the value of everything and the cost of nothing". You become insulated from the externalities of leaky abstractions, cumbersome implementation, and even the negative externalities of burning the tokens to brute-force generate the code in the first place.