Login
You're viewing the functional.cafe public feed.
  • Oct 27, 2026, 10:17 PM

    does anyone remember when it was generally agreed that a large amount of boilerplate in a project was a sign of bad design, and that you should seek to improve things through abstraction or refactoring rather than simply automating and codegen-ing your way through?

    💬 19🔄 105⭐ 301

Replies

  • 💬 0🔄 0⭐ 5
  • 💬 0🔄 0⭐ 6
  • 💬 0🔄 0⭐ 4
  • 💬 6🔄 0⭐ 9
  • 💬 1🔄 0⭐ 7
  • 💬 1🔄 0⭐ 1
  • 💬 0🔄 0⭐ 1
  • Oct 27, 2026, 10:45 PM

    @Igigog @jacqueline that would require the LLMs to be able to generate random programs in ASM. since humans don't read that, it's not generally uploaded online for them to steal and extrude

    💬 0🔄 0⭐ 2
  • Oct 27, 2026, 10:53 PM
    I get the impression that a lot of slop/vibe generated stuff is from people who aren't programmers at all, so they probably have no clue that going down to asm would be an improvement over higher level languages?

    I had to stop using interpreted languages in elementary school because they were so inefficient, asm was the way back then; when I finally encountered C in college? It felt bloated. I do not think anyone using LLMs and vibing their way around, would have such insights; since I get the impression most of those lessons were lost on folks after CPUs transcended the 10MHz barrier, decades ago and kept getting faster and cheaper and nerds such as I were some of the only ones left who cared?

    CC: @jacqueline@chaos.social
    💬 1🔄 0⭐ 0
  • 💬 0🔄 2⭐ 1
  • Oct 28, 2026, 10:49 AM

    @teajaygrey @Igigog @jacqueline Ah, you just reminded me of the times I would code my UI in ASM and just wrap everything in TurboPascal 7 for convenience.

    I had drop down menus, resizable pickers, all in under 70KB.

    Turbo vision was a bloat.

    💬 0🔄 0⭐ 1
  • Oct 28, 2026, 12:08 AM

    @Igigog @jacqueline I am certain that at some point LLMs will emit compiled code directly for "optimization". No one will be able to look at the source, since there will be no source.

    Sounds like a reasonable point on the enshittification path.

    💬 2🔄 3⭐ 4
  • 💬 1🔄 0⭐ 2
  • 💬 1🔄 0⭐ 1
  • 💬 0🔄 0⭐ 0
  • Oct 28, 2026, 10:58 AM

    @dacmot @f4grx @Igigog @jacqueline Oh no, this makes it more fun, because sooner or later the big bots will disappear due to unsustainable expenses. Then corpos will scramble to try to set up local models in order to save their codebase from utter chaos.

    "Wanted: AI engineers, experts in configuring, setting up and using local AI models.

    Wanted: Code sleuths and system architects, experienced in navigating unknown code bases."

    It will be exhilarating.

    Senku from Dr. Stone rolling on the floor laughing.
    💬 0🔄 0⭐ 0
  • Irisiris_meredith@mastodon.social
    Oct 28, 2026, 5:30 AM

    @f4grx @Igigog @jacqueline
    I'm not sure to what extent this is actually possible with the current technology, to be fair. The training corpus to make it work simply isn't there. They could *try*, but I don't think there are enough resources to dedicate to it

    💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Oct 28, 2026, 12:27 AM

    @Igigog @jacqueline people are saying asm is in short supply somehow; but machine code is all around you and disassembling it is truly trivial. i think the real reason is that in asm it's not possible to beginner-level halfass something of any non-trivial size and have it actually work, or at least it's too unlikely even by the standards of cherry-picking-as-a-service

    💬 0🔄 0⭐ 2
  • Oct 28, 2026, 10:45 AM

    @Igigog @jacqueline Funny you mention that, I'm writing a scifi novel in which the androids decide in hivemind what's the best language to write software in, since humans stopped caring about legibility. As long as the AIs could modify it, it's irrelevant.

    In practice, the AIs ended up rewriting everything from scratch every few months because bots would change at random to minimize costs in an ever changing technological landscape.

    That's just worldbuilding, tho.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 1
  • Oct 27, 2026, 10:37 PM

    @jacqueline when I was fresh out of university and starting my career I thought that everyone was like me and actually enjoys building elegant concise maintainable and extensible systems

    but I guess some people just want to trot out slop and go home?

    💬 1🔄 3⭐ 0
  • Oct 27, 2026, 10:50 PM

    @zrb @jacqueline I think this is what happens when for the last 20 years or so it became common to tell students to get into this profession for the money. (Which makes sense - it's more money than most degrees.)

    So we now have a large population of coders who really (more or less) _don't_ care that much. Getting a bigger paycheck was their primary motivation in the first place.

    💬 0🔄 0⭐ 9
  • Oct 27, 2026, 11:10 PM

    @jacqueline nah but im also like not smart enough to know what boilerplate means bc like i js write code in a way that seems fine enough and let future me suffer

    💬 0🔄 0⭐ 1
  • Oct 28, 2026, 12:10 AM

    @jacqueline to this day i ruthlessly go through my projects looking for repeating boilerplate and try to find ways to eliminate it.

    Sometimes it's easier than other times, but I always see it as an opportunity for improvement

    💬 0🔄 0⭐ 2
  • 💬 1🔄 0⭐ 3
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 1
  • Oct 28, 2026, 2:18 AM

    @jacqueline omg yes. when "good" was the goal rather than "now?" When being asked to do things of questionable legal, ethical, or moral value was preposterous?

    "This could kill all humans? Well obviously I gotta do it first!"

    remember when people actually liked programming and helped each other learn?

    💬 0🔄 0⭐ 1
  • 💬 0🔄 0⭐ 2
  • 💬 0🔄 0⭐ 2
  • 💬 0🔄 0⭐ 0
  • Oct 28, 2026, 7:51 AM

    @jacqueline and don't forget that refactoring can be fun!

    When you're not under undue pressure, "I removed a hundred lines of code and the program behaviour is identical to what it was before I started" is more oddly satisfying than it initially seems

    💬 1🔄 0⭐ 3
  • 💬 0🔄 0⭐ 0
  • Oct 28, 2026, 8:35 AM

    @jacqueline I was thinking the same. Instead of really trying to have general thinking and abstracting, AI generates and generates code. On the other hand, we've all been lazy and skipped writing a test, verifying corner cases. AI doesn't do that. Compromise would be to use AI for review. That basically the stance or projects like curl who were initially very anti-AI.

    💬 0🔄 0⭐ 0
  • Oct 28, 2026, 10:33 AM

    @jacqueline Oh, you hit a core memory! 😲

    I remember that sometimes I would over-engineer a project just so I would get the perfect framework so that the rest of the code and modifications would be minimal!

    Ah, I miss those times. ☺️

    💬 0🔄 0⭐ 0
  • Oct 28, 2026, 12:23 PM

    @jacqueline
    Orthogonal to your point, but I'm still sad that declarative parser and state machines generators didn't catch on, either as tools or as language features. They're the exact opposite of slopperators and would have saved us from a lot of bugs.

    💬 0🔄 0⭐ 0
  • Oct 28, 2026, 7:31 PM

    @jacqueline I was just thinking about this today. It used to be harder to implement bad designs, because you’d get part way into it and discover you had to solve a problem you weren’t actually out to solve, or do something ludicrously hard from a labor perspective. Now AI just powers right through it and all the review bots tell you is to add a couple null checks and cover some of the wild edge cases you created. I really miss when it was harder to implement bad design decisions.

    💬 0🔄 0⭐ 0
  • Oct 29, 2026, 6:23 AM

    @jacqueline
    That was always my opinion, but never agreed by corporations selling developers by time&material rates.
    why would we have had CORBA, EJB, XML and all the newer stuff requiring code generation of boilerplate code?

    💬 0🔄 0⭐ 0
  • Oct 29, 2026, 11:10 AM

    @jacqueline feels like yesterday (read as: i gave advice for refactoring yesterday and have no plans to stop doing it)

    💬 0🔄 0⭐ 0