Posted by xyzal 5 hours ago
What does he mean by this? What skills were lost? Writing HTML templates?
The effect of this is that people who never learned those things are working with a limited toolkit both in solving problems and debugging. And every so often I get to blow someone’s mind with an old trick :)
It’s not all bad though! I’m happy to never build a layout with floats again.
At the end of the day, I have to make more architectural and business decisions than before - it’s just higher-level and more complex work.
On the other hand, there’s increasingly little reason to hire someone just to write APIs or work on the frontend, since AI handles most of the routine tasks.
So, this feels much more like the Industrial Revolution than “deskilling.”
Apparently deskilled people are making it look like this is normal and it supposed to be so.
But i can relate to that. Another examples of deskilling would be, of course, Java, and a more modern example - Rust.
That said, i don't think deskilling is solving mass-production problem. It was already solved with open-source software, or with a software as is.
Software is information and there is little to no cost of copying information. So mass-production isn't the problem that is being solved here.
IMO the problem being solved is that business need unskilled labor, that is slop.
You would think that if business is producing slop, it will be replaced with another business producing quality stuff. If that was so, over time, there won't be any slop on the market, but if you open your app store, you are welcomed by all kinds of slop.
Because slop is what they buy. Supply is only following the demand, business need to produce slop because people are buying it.
How many of you guys have Claude subscription? Do you know that 5 years ago i would be asking "How many of you guy have GitHub Copilot subscription"?
This is what people buy, so it is deskilling, but not a mass-production, it's just slop revolution, slop is the new norm.
But situation was exactly the same before the AI. You would still get your wordpress, your React frontend and Java Spring backend.
AI doesn't change anything, it just takes the job of a poor slopper who made a living by coding React frontends. Anthropic just took their job, that's it, and you don't see the difference.
With Stack Overflow, you got multiple answers from different people with different viewpoints and different approaches, each consistent within itself. You could figure out where the author of each answer was coming from and judge whether they seemed to know what they were talking about. You could weigh the trade-offs and merits of the different answers against each other.
With LLMs, you get a single mushy pile of slop, not grounded in any person's actual experience or judgement. It might pretend to offer different perspectives, but it can't really, so it's much harder to evaluate.
As someone who didn’t really know that being a front-end dev putatively doesn’t involve thinking about those things anymore, I think that list conflates a couple of different things.
Things like the differences between browsers and CSS/HTML quirks, needed to wrangle a document markup language into creating user interfaces, are accidental complexity caused by particular path dependencies, and if they can be abstracted out, that’s a great thing.
Accessibility, interface design, performance, and other things related to user experience, on the other hand? Those are mostly orthogonal. A UI framework can raise (or in some cases lower) the bottom in the sense of facilitating reuse of (hopefully well-designed) components, but no framework is going to make your UI accessible or well designed by itself.
In the fabled past, frontend development didn’t require you to be highly qualified in these matters – web UIs were simply terrible, mostly. High skill level was not required because nobody expected anything from web UIs beyond the barest core functionality.
There was UI programming before the web [citation needed]. In a sense it was "deskilled" because you used a "framework" aka the OS windowing and widget libraries rather than drawing rectangles manually (except in some special cases like games where very custom UI is desired – but those custom controls invariably have roughly 0% of the UX affordances provided by standard ones). Back then, Visual Basic and other RAD tools (anyone still remember that acronym?) were front line of "deskilling", but honestly WYSIWYG visual design is still one of the best ways to create UIs, it’s just rarely done these days for various reasons.
I've seen people argue that LLMs will just add another layer to the top of the compiler stack: instead of writing code, we'll use English, and run it through a pipeline:
English -> Rust -> ASM -> Machine Code
But what the author says about agents being "undeterministic abstraction" shows why that will never work.
Compilers rely on a concept called observational equivalence[1] to define when two programs are basically the same; this allows them to make changes under the hood like unrolling a loop or targeting another machine. Now, it turns out we know a lot about how and how not to do this, thanks to a logician named Frege who worked out exactly which properties a "definition" would need to have to count as a definition without becoming an axiom. In particular, that it should be "eliminable" and "conservative"[2]. In plain language, that a formal definition should always be able to be eliminated by rote string substitution, and that it shouldn't smuggle in any extra assumptions. When we talk about things like syntactic sugar[3] or hygienic macros[4], we are basically applying Frege's two conditions to programming languages.
LLMs are neither. They cannot reliably or provably go from the prompts they are given to the source code they generate, and they make a ton of implicit assumptions when they do so. There can never be any equivalence between two "prompts" in the same way that two programs can be equivalent modulo some level of abstraction. The whole process of starting from prompts is wildly nondeterministic, which is why the only pattern that works is to generate the code, review it, and test it, and then check it in and use that as the starting point for the next prompt.
Which is not to say that LLMs aren't useful for code generation; they clearly are. But they don't provide an abstraction that lets us get away from the details of actual code, and thanks to Frege we can understand why they never will.
I can say all this with such confidence because I did once write a wild little Python library that used a bunch of introspection to actually do this[5]. And it absolutely did not work in practice beyond toy examples.
[1]: https://en.wikipedia.org/wiki/Observational_equivalence
[2]: https://plato.stanford.edu/entries/frege/#ProDef
[3]: https://en.wikipedia.org/wiki/Syntactic_sugar