I feel like I see this attitude a lot amongst devs: "If everyone just built it correctly, we wouldn't need these bandaids"
To me, it feels similar to "If everyone just cooperated perfectly and helped each other out, we wouldn't need laws/money/government/religion/etc."
Yes, you're probably right, but no that won't happen the way you want to, because we are part of a complex system, and everyone has their very different incentives.
Semantic web was a standard suggested by Google, but unless every browser got on board to break web pages that didn't conform to that standard, then people aren't going to fully follow it. Instead, browsers (correctly in my view) decided to be as flexible as possible to render pages in a best-effort way, because everyone had a slightly different way to build web pages.
I feel like people get too stuck on the "correct" way to do things, but the reality of computers, as is the reality of everything, is that there are lots of different ways to do things, and we need to have systems that are comfortable with handling that.
Was this written by AI? I find it hard to believe anyone who was interested in Semantic Web would have not known it's origin (or at least that it's origin was not Google).
The concept of a Semantic web was proposed by Tim Berners-Lee (who hopefully everyone recognizes as the father of HTTP, WWW, HTML) in 1999 [0]. Google, to my knowledge, had no direct development or even involvement in the early Semweb standards such as RDF [1] and OWL [2]. I worked with some of the people involved in the latter (not closely though), and at the time Google was still quite small.
That was a human-generated hallucination, my apologies. I always associated semantic web with something Google was pushing to assist with web crawling, and my first exposure to it was during the Web 2.0 era (early 2010s) as HTML5 was seeing adoption, and I always associated it with Google trying to enhance the web as the application platform of the future.
W3C of course deserves credit for their hard work on this standard.
My main point was that regardless of the semantic "standard", nothing prevented us from putting everything in a generic div, so complaining that everyone's just "not on board" isn't a useful lament.
The phrase “if everyone just” is an automatic trigger for me. Everyone is never going to just. A different solution to whatever the problem is will be necessary.
I can't find a copy of the old "reasons your solution to email spam won't work" response checklist, but one of the line items was "fails to account for human nature".
eh I feel this but it's a lot simpler than that. Not "if everyone built everything correctly" but "if everyone's work was even slightly better than complete garbage". I do not see many examples of companies building things that are not complete embarrassing shit. I worked at some companies and the things we built was complete embarrassing shit. The reasons are obvious: nobody cares internally to do it, and nobody externally has any standards, and the money still flows if you do a bad job so why do better?
What happens in practice is that the culture exterminates the drive for improvement: not only are things bad, but people look at you if you're crazy if you think things should be better. Like in 2025 people defend C, people defend Javascript, people write software without types, people write scripts in shell languages; debugging sometimes involves looking at actual bytes with your eyes; UIs are written in non-cross-platform ways; the same stupid software gets written over and over at a million companies, sending a large file to another person is still actually pretty hard; leaving comments on it is functionally impossible ... these are software problems, everything is shit, everything can be improved on, nothing should be hard anymore but everything still is; we are still missing a million abstractions that are necessary to make the world simple. Good lord, yesterday I spent two hours trying to resize a PDF. We truly live in the stone age; the only progress we've made is that there are now ads on every rock.
I really wish it was a a much more ruthlessly competitive landscape. One in which if your software is bad, slow, hard to debug, hard to extend, not open source, not modernized, not built on the right abstractions, hard to migrate on or off of, not receptive to feedback, covered in ads... you'd be brutally murdered by the competition in short order. Not like today where you can just lie on your marketing materials and nobody can really do anything because the competition is just as weak. People would do a much better job if they had to to survive.
> the money still flows if you do a bad job so why do better?
I'll raise. The money flows because you do a bad job. Doing a good job is costly and takes time. The money cannot invest that much time and resources. Investing time and resources builds an ordinary business. The money is in for the casino effect, for the bangs. Pull the arm and see if it sticks. If yes, good. Keep pulling the arm. If not, continue with another slot machine.
we have to accept that the vast majority of people don't think like us. They don't think its complete garbage because they can't look hard enough to appreciate that fidelity.
While it might be better if everyone thought like us and wanted things to be _fundamentally_ good, most people don't, and most people money >> less people money and the difference in scale is vast.
We could try to build a little fief where we get to set the rules but network effects are against us. If anything our best shot is to try to reverse the centralisation of the internet because that's a big cause of enshittification.
The semantic web came out of work on Prolog and formal systems for AI which just didnt work well... LLMs and vector databases give us new tools that are pretty usable.
Neither broke web pages, honestly. XHTML requires a DTD named at the top of the document, and browsers will happily fall back to HTML 3, 4, or 5 as they can if there’s no DTD specified.
Fresh plums right off the tree taste significantly better than the ones you can get in the produce isle, which are in turn better than canned, which are themselves still better than re-hydrated prunes.
In scaling out computation to the masses, we went from locally grown plums that took a lot of work and were only available to a small number of people that had a plum tree or knew someone that had one, to building near magical prune-cornucopia devices that everyone could carry around in their pockets, giving them an effectively unlimited supply of prunes.
LLMs re-hydrate these for us, making them significantly more palatable; if you're used to gnawing dried fruit, they seem amazing.
Perhaps, but we still failed and not at personal computing, nor just semantic web, but computing and programming in general. The failure is between the original intent (computing was originally more or less AI) along with theory and actual result with every software project turning into unsustainable infinite regress. Things likely broke around ALGOL.
Also LLMs are failing too, for different reasons, but IMO unlikely AI in general will— it will correct a 60 years or so failure in industrial computer science.
The most reprehensible knowledge-search-and-communication failure of all.
We gave people monetisation of drek instead. Then asked them to use it for education. Then trained bot systems on it. Then said that even those answers must be made to confirm to transient propagandists.
I really don't like this analogy, and I really don't like the premise of this article.
Writing software is only so scalable. It doesn't matter all of the shortcuts we take, like Electron and JavaScript. There are only so many engineers with so much time, and there are abundantly many problems to solve.
A better analogy would be to look at what's happening to AI images and video. Those have 10,000x'd the fundamental cost savings, time savings, and personnel requirements. It's an industrialization moment. As a filmmaker who has made several photons-on-glass films, this is a game changer and lifts the entire creative industry to a level where individuals can outdo Pixar.
That is the lens with which to look at what AI will do to software. We're going from hand-carving stone wheels to the Model T.
This is all just getting started. We've barely used what the models of today offer us.
Totally agree with the core of your position. But the two perspectives are complementary, and perhaps even more deeply linked.
Initial attempts to alleviate any shortage are likely to result in a decrease in quality; initial attempts to improve quality are likely to reduce variability and thus variety. My point (and my reading of the article) is that we're at the stage where we've figured out how to turn out enough Hostess Twinkies that "let them eat cake!" is a viable option, and starvation is being driven back.
This is definite progress, but also highlights previous failures and future work.
This is a massive cope. AI image/video slop is still slop. Yes it's getting better, but it's still better .. slop. Unless radical new breakthroughs are made, the current LLM paradigm will not outdo Pixar or any other apex of human creativity. It'll always be instantly recognizable, as slop.
And if we allow it to take over society, we'll end up with a society that's also slop. Netflixification/marvelization only much much worse..
The computers, yes. The experience of using them, no.
There is a joy to having a tool that lets you do what you never could before that has been buried along the way in layers of petty annoyances and digital micro aggressions. As you say, computers today are better by so many metrics—including, unfortunately, their ability to tracks us, sell us things we neither need nor want, outrage and distract us.
I have often thought about how computers are significantly faster than they were in the early 2000s, but they are significantly harder to use. Using Linux for the first time in college was a revelation, because it gave me the tools to tell the computer "rename all of the files in this directory, keeping only the important parts of the name."
But instead of iterating on better interfaces to effectively utilize the N thousands of operations per second a computer is capable of, the powers that be behind the industry have decided to invest billions of dollars in GPUs to get a program that seems like it understands language, but is incapable of counting the number of B's in "blueberry."
IDK, I think there is something adorable about taking a system that over trillions of iterations always performs the same operation with the same result, reliability unmatched in all of the universe…
And making it more of “IDK what it answered the way it did, but it might be right!!”
They're not claiming AGI yet, so human intelligence is required to operate an LLM optimally. It's well known that LLMs process tokens rather than characters s, so without space for "reasoning" there's no representation of the letter b in the prompt. Telling it to spell or think about it gives it room to spell it out, and from there it can "see" the letters and it's trivial to count.
So long as I’m teaching the user how to speak to the computer for a specific edge case, which of these burn nearly as much power as your prompt? Maybe we should consider that certain problems are suitable to LLMs and certain ones should be handled differently, even if that means getting the LLM to recognize its own edge cases and run canned routines to produce answers.
Is counting the number of B's vital? Also, I'm pretty sure you can get an LLM to parse text the way you want it, it just doesn't see your text as you do, so that simple operation is not straightforward. Similarly, are you worthless because you seem like you understand language but are incapable of counting the number of octects in "blueberry"?
Let's say I hire a plumber because of his plumbing expertise and he bills me $35 and I pay him with a $50 bill and he gives me back $10 in change. He insists he's right about this.
I am now completely justified in worrying about whether the pipes he just installed were actually made of butter.
> Similarly, are you worthless because you seem like you understand language but are incapable of counting the number of octects in "blueberry"?
Well, I would say that if GP advertised themselves as being able to do so, and confidently gave an incorrect answer, their function as someone who is able to serve their advertised purpose is practically useless.
As shown by the GPT-5 reaction, a majority of people just have nothing better to ask the models than how many times does the letter "s" appear in "stupid".
I mean, I think that anyone who understands UTF-8 will know that there are nine octets in blueberry when it is written on a web page. If you wanted to be tricky, you could have thrown a Β in there or something.
Yeah dream on. I’m an engineer and know what structured data is. And yet I miserably fail to store my private files in a way that I can find them back without relying on search tools. So how on earth are we ever going to organize all the world’s data and knowledge? Thank god we found this sub-optimal “band aid” called LLMs!
> Remember Semantic Web? The web was supposed to evolve into semantically structured, linked, machine-readable data that would enable amazing opportunities. That never happened.
I think the lesson to be learned is in answering the question "Why didn't the semantic web happen?"
I have literally been doing we development since their was a web, and the companies I developed for are openly hostile to the idea of putting their valuable, or perceived valuable, information online in a format that could be easily scraped. Information doesn't want to be free, it wants to be paid for. Unless the information shared pulls visitors to the site it doesn't need to be public.
> Information doesn't want to be free, it wants to be paid for. Unless the information shared pulls visitors to the site it doesn't need to be public.
That's a cultural and societal problem, not a technology problem. The motivations (profit) are wrong, and don't lead to true innovations, only to financialization.
So long as people need to pay to eat, then information will also want to continue to be paid for, and our motivations will continue to be misaligned with true innovations, especially if said innovations would make life easier but wouldn't result in profit.
I've encountered a similar issue in academia - PI's don't want to make their data available to be scraped (or, at least not easily) because the amount of grant funding is limited, and a rival who has scraped one's data could get the grant money instead by using that scraped data to bolster their application.
To a degree re ads on pages, but why didn't big business end up publishing all of their products in JSON-LD or similar?
A lot did, to get aggregated, but not all.
But also because companies that produce web content wanted it to be seen by humans who would look at ads, not consumed by bots and synthesized with other info into a product owned by some other firm.
The semantic web was theoretically great for data scientists and metadata scrapers, but offered close to zero value for ordinary humans, both on the publishing side an the consumption side. Also, nobody did the hard work of defining all of the categories and protocols in a way that was actually usable.
The whole concept was too high minded and they never got the implementation details down. Even if they did it would have been horrendously complex and close to impossible to manage. Asking every single publisher to neatly categories their data into this necessarily enormous scheme would have resulted in countless errors all over the web that would have seriously undercut the utility of the project anyway. Ultimately the semantic web doesn't scale very well. It failed for the same reason command economies fail: It's too overwhelming for the people in control to manage and drowns in its own bureaucracy.
Semantic web never existed. There was Google and Google had an API to get breadcrumbs to show on search results. And that's what people called "semantic web." A few years later they gave up and made everything look like a breadcrumb anyway. And that sums up the whole semantic web experience.
I find traditional web search and LLM search to be complementary technologies, and this is a good example of that. Both have their uses and if you get the information you need using one or the other, we are all happy.
I think the example query here actually shows a problem with the query languages used in web search rather than an intrinsic inability of web search. It contains what amounts to a natural language subquery starting with "in the same year". In other words, to evaluate this query properly, we need to first evaluate this subquery and then use that information to evaluate the overall query. Google Search and almost all other traditional web search engines use intentionally oversimplified query languages that disallow nested queries let alone subqueries, so this example really is just revealing a problem with the query language rather than a problem with web search overall. With a better query language, we might get better results.
Readit News
To me, it feels similar to "If everyone just cooperated perfectly and helped each other out, we wouldn't need laws/money/government/religion/etc."
Yes, you're probably right, but no that won't happen the way you want to, because we are part of a complex system, and everyone has their very different incentives.
Semantic web was a standard suggested by Google, but unless every browser got on board to break web pages that didn't conform to that standard, then people aren't going to fully follow it. Instead, browsers (correctly in my view) decided to be as flexible as possible to render pages in a best-effort way, because everyone had a slightly different way to build web pages.
I feel like people get too stuck on the "correct" way to do things, but the reality of computers, as is the reality of everything, is that there are lots of different ways to do things, and we need to have systems that are comfortable with handling that.
Was this written by AI? I find it hard to believe anyone who was interested in Semantic Web would have not known it's origin (or at least that it's origin was not Google).
The concept of a Semantic web was proposed by Tim Berners-Lee (who hopefully everyone recognizes as the father of HTTP, WWW, HTML) in 1999 [0]. Google, to my knowledge, had no direct development or even involvement in the early Semweb standards such as RDF [1] and OWL [2]. I worked with some of the people involved in the latter (not closely though), and at the time Google was still quite small.
0. https://archive.org/details/isbn_9780062515872/mode/2up
1. https://www.w3.org/TR/PR-rdf-syntax/Overview.html
2. https://www.w3.org/TR/owl-ref/
W3C of course deserves credit for their hard work on this standard.
My main point was that regardless of the semantic "standard", nothing prevented us from putting everything in a generic div, so complaining that everyone's just "not on board" isn't a useful lament.
https://philip.greenspun.com/research/shame-and-war-old
he's actually still working on it: https://solidproject.org/
Tim Berners-Lee coined it in 1999 and further expanded on the concept in a 2001 Scientific American article by Berners-Lee, Hendler, and Lassila.
What happens in practice is that the culture exterminates the drive for improvement: not only are things bad, but people look at you if you're crazy if you think things should be better. Like in 2025 people defend C, people defend Javascript, people write software without types, people write scripts in shell languages; debugging sometimes involves looking at actual bytes with your eyes; UIs are written in non-cross-platform ways; the same stupid software gets written over and over at a million companies, sending a large file to another person is still actually pretty hard; leaving comments on it is functionally impossible ... these are software problems, everything is shit, everything can be improved on, nothing should be hard anymore but everything still is; we are still missing a million abstractions that are necessary to make the world simple. Good lord, yesterday I spent two hours trying to resize a PDF. We truly live in the stone age; the only progress we've made is that there are now ads on every rock.
I really wish it was a a much more ruthlessly competitive landscape. One in which if your software is bad, slow, hard to debug, hard to extend, not open source, not modernized, not built on the right abstractions, hard to migrate on or off of, not receptive to feedback, covered in ads... you'd be brutally murdered by the competition in short order. Not like today where you can just lie on your marketing materials and nobody can really do anything because the competition is just as weak. People would do a much better job if they had to to survive.
I'll raise. The money flows because you do a bad job. Doing a good job is costly and takes time. The money cannot invest that much time and resources. Investing time and resources builds an ordinary business. The money is in for the casino effect, for the bangs. Pull the arm and see if it sticks. If yes, good. Keep pulling the arm. If not, continue with another slot machine.
While it might be better if everyone thought like us and wanted things to be _fundamentally_ good, most people don't, and most people money >> less people money and the difference in scale is vast. We could try to build a little fief where we get to set the rules but network effects are against us. If anything our best shot is to try to reverse the centralisation of the internet because that's a big cause of enshittification.
mom: "you need to clean up your room"
kid: "mom, just give up. The room will always be a mess, just use search"
In scaling out computation to the masses, we went from locally grown plums that took a lot of work and were only available to a small number of people that had a plum tree or knew someone that had one, to building near magical prune-cornucopia devices that everyone could carry around in their pockets, giving them an effectively unlimited supply of prunes.
LLMs re-hydrate these for us, making them significantly more palatable; if you're used to gnawing dried fruit, they seem amazing.
But there's still a lot of work to be done.
Also LLMs are failing too, for different reasons, but IMO unlikely AI in general will— it will correct a 60 years or so failure in industrial computer science.
The most reprehensible knowledge-search-and-communication failure of all.
We gave people monetisation of drek instead. Then asked them to use it for education. Then trained bot systems on it. Then said that even those answers must be made to confirm to transient propagandists.
Except sometimes you're expecting a fresh plum, and then you bite into a fig, or an apple, or a banana, or a stick.
so sweet
and so cold
:)
Writing software is only so scalable. It doesn't matter all of the shortcuts we take, like Electron and JavaScript. There are only so many engineers with so much time, and there are abundantly many problems to solve.
A better analogy would be to look at what's happening to AI images and video. Those have 10,000x'd the fundamental cost savings, time savings, and personnel requirements. It's an industrialization moment. As a filmmaker who has made several photons-on-glass films, this is a game changer and lifts the entire creative industry to a level where individuals can outdo Pixar.
That is the lens with which to look at what AI will do to software. We're going from hand-carving stone wheels to the Model T.
This is all just getting started. We've barely used what the models of today offer us.
Initial attempts to alleviate any shortage are likely to result in a decrease in quality; initial attempts to improve quality are likely to reduce variability and thus variety. My point (and my reading of the article) is that we're at the stage where we've figured out how to turn out enough Hostess Twinkies that "let them eat cake!" is a viable option, and starvation is being driven back.
This is definite progress, but also highlights previous failures and future work.
And if we allow it to take over society, we'll end up with a society that's also slop. Netflixification/marvelization only much much worse..
There is a joy to having a tool that lets you do what you never could before that has been buried along the way in layers of petty annoyances and digital micro aggressions. As you say, computers today are better by so many metrics—including, unfortunately, their ability to tracks us, sell us things we neither need nor want, outrage and distract us.
Deleted Comment
But instead of iterating on better interfaces to effectively utilize the N thousands of operations per second a computer is capable of, the powers that be behind the industry have decided to invest billions of dollars in GPUs to get a program that seems like it understands language, but is incapable of counting the number of B's in "blueberry."
And making it more of “IDK what it answered the way it did, but it might be right!!”
They're not claiming AGI yet, so human intelligence is required to operate an LLM optimally. It's well known that LLMs process tokens rather than characters s, so without space for "reasoning" there's no representation of the letter b in the prompt. Telling it to spell or think about it gives it room to spell it out, and from there it can "see" the letters and it's trivial to count.
echo blueberry | grep -o 'b' | wc -l
echo blueberry | perl -ne 'print scalar (() = m/(b)/g)’
echo blueberry | tr -d '\n' | tr b '\n' | wc -l
echo -n blueberry | tr b '\n' | wc -l
So long as I’m teaching the user how to speak to the computer for a specific edge case, which of these burn nearly as much power as your prompt? Maybe we should consider that certain problems are suitable to LLMs and certain ones should be handled differently, even if that means getting the LLM to recognize its own edge cases and run canned routines to produce answers.
I am now completely justified in worrying about whether the pipes he just installed were actually made of butter.
Well, I would say that if GP advertised themselves as being able to do so, and confidently gave an incorrect answer, their function as someone who is able to serve their advertised purpose is practically useless.
https://en.wikipedia.org/wiki/An_Essay_Towards_a_Real_Charac...
I'm sure there have been countless similar attempts at categorizing knowledge
one of the more successful ones being the dewey decimal system
I have my doubts about whether the thing the OP alleges we have "failed" at is even possible at all
Because that’s what a lot of this falls into.
Overwhelming amount of stuff with no names. No categories, no nothing.
With extended file attributes we could hang all sorts of meta bits off of arbitrary files. But that’s very fragile.
So we ask the systems to make up names for data based on their content, which turns out to not necessarily work as well as we might like.
I think of the Whalphin and it took Sea World era to discover. Who would see that coming?
Dead Comment
> Remember Semantic Web? The web was supposed to evolve into semantically structured, linked, machine-readable data that would enable amazing opportunities. That never happened.
I think the lesson to be learned is in answering the question "Why didn't the semantic web happen?"
I have literally been doing we development since their was a web, and the companies I developed for are openly hostile to the idea of putting their valuable, or perceived valuable, information online in a format that could be easily scraped. Information doesn't want to be free, it wants to be paid for. Unless the information shared pulls visitors to the site it doesn't need to be public.
That's a cultural and societal problem, not a technology problem. The motivations (profit) are wrong, and don't lead to true innovations, only to financialization.
So long as people need to pay to eat, then information will also want to continue to be paid for, and our motivations will continue to be misaligned with true innovations, especially if said innovations would make life easier but wouldn't result in profit.
Advertising.
Because web content is generated by humans, not engineers.
The whole concept was too high minded and they never got the implementation details down. Even if they did it would have been horrendously complex and close to impossible to manage. Asking every single publisher to neatly categories their data into this necessarily enormous scheme would have resulted in countless errors all over the web that would have seriously undercut the utility of the project anyway. Ultimately the semantic web doesn't scale very well. It failed for the same reason command economies fail: It's too overwhelming for the people in control to manage and drowns in its own bureaucracy.
"Standing on the shoulders of giants, it is clear that the giants failed to reach the heights we have reached."
Everything that is bad in UI is a direct consequence of that.
1. No tooltips, right click, middle click behavior because touch doesn't have that. No ctrl+click either.
2. Large click areas wasting screen space with padding and margins.
3. Low density UI so it can shape-shift into mobile version.
4. Why type on a phone when you can talk? Make everything a search box.
5. Everything must be flat instead of skeumorphic because it's easier to resize for other screen sizes.
6. Everything needs a swipe animation and views instead of dialogs because smartphones can't have windows.
I think the example query here actually shows a problem with the query languages used in web search rather than an intrinsic inability of web search. It contains what amounts to a natural language subquery starting with "in the same year". In other words, to evaluate this query properly, we need to first evaluate this subquery and then use that information to evaluate the overall query. Google Search and almost all other traditional web search engines use intentionally oversimplified query languages that disallow nested queries let alone subqueries, so this example really is just revealing a problem with the query language rather than a problem with web search overall. With a better query language, we might get better results.