Love this writeup. The temu analogy is best to describe what is going on beyond just what llms have done to software. What's closer to question is temu could only exist in the way it did (it's fast rise mainly) because of the work done by amazon prime in making fast shipping both logistically doable and the psychological default for many. So my question is what was the "amazon prime" era of software and perhaps learning what did (and did not) occur then can better understand why AI has wrought the effects it has in less than 10 years.
As much as I hate Amazon, I think there is a simpler explanation. Europe has been relaxing their consumer protection standards for the past two decades, and EU’s consumer protection enforcement against foreign tech companies is hardly ever done today. This applies to both American and Chinese companies.
TEMU only exists because they simply neglected quality of their products which their competitors were required by law to have. And then they were not punished for it as they should have been.
In other words, TEMU only exists because Europe has turned into a deregulated neo-liberal paradise.
I have never, ever, bought anything on TEMU, because anytime I open the website it gives me a gambling roulette for some supposed discount thing. I hate gambling, so everytime I see the roulette I just close the webpage and go do something else.
Well, the whole cheap-generated content craze feels basically like gambling to me to this point, too. Folks are just randomly generating content: books, videos, articles, even applications, and gambling that something will pay for itself.
So now we have Enshitification, Eemufication and Gamblification of the Human Society. Damn I might write something on Substack about that.
Hah, I love sharing that link too. But contemplate this: Give the Oracle 12 engineering culture an unlimited token budget, do you think the code will get better or worse?
Based on what I've seen in industry, I think the answer could be very likely worse. They'll generate bad code at an even faster rate, and very quickly nobody in the company will have any idea how anything works.
What this proves is that "human made" is not automatically good. Better is a complex question. LLMs have a magnifying affect on the quality of human decisions.
Oracle's codebase got the way it is over several decades of human programmers and customers with changing needs and ambient technological environments. LLM code can get that way in a weekend.
A few years ago I would have laughed at anyone telling me that there is a serious market for ten-dollar drills, two-dollar dresses, and one-dollar pairs of shoes shipped from a warehouse on the other side of the planet.
TEMU doesn't sell those things. They're more expensive than that, but subsidised by VC capital from PDD Holdings, Blackrock, Vanguard, etc. It's a very boring 'spend the VC bucks to get the customers' play, and when they've got the customers they'll jack up the prices and lose the customers again. It's Uber-for-cheap-stuff basically.
Why on earth would anyone be surprised that this market exists? I can understand being surprised that it's possible to deliver those items at the stated price, but the market's always been there.
I seem to remember “assume the customer will always want better prices” to be one of the few ironclad axioms that Jeff Bezos oriented early Amazon around.
If you're in construction then you need a quality drill that will carry you from one job to the next reliably. If you're just trying to do a one-off drywall repair in your house that you hope never to need to do again, maybe the $10 drill is precisely what you'd buy, if you could find it.
Haha, and the way Chinese economics are described here shows what little people know.
When they invited our professors in the 80s they came back arguing China will eventually be the future. They listened to everything and started working on everything immediately.
We did the yes men privatization garage sale, gutted R&D, outsourcing, abolish education, housing, medical care. All that good stuff - but at what cost?
Quality products are amazing but I do get the downside of building very durable products.
For example, we have a Vornado fan in my house that is probably from the early 2000s. It still works perfectly even though we've left it running 24/7 for weeks in humid or dusty environments (e.g. to help air dry a wet basement). In terms of money paid/hours used, we got an amazing deal on the purchase.
From the manufacturer's perspective, the life time value (LTV) of us as a customer was basically just one fan + the free marketing of folks like myself writing this kind of post. It reminds me of the joke "Did you hear about the company that made lightbulbs that lasted 30 years? They went out of business!"
It makes sense that manufacturers would start to lower quality at a slightly faster rate than they lower the price as this means that folks still feel happy with the purchase. Sure, the fan doesn't last as long due to what is essentially planned obsolescence but the company does get to stay in business.
Temu/Shein etc are just the logical endpoint of this line of reasoning. If you are going to wear a shirt for the one summer where it's in style, why would you spend a large amount of money on it? And in turn, why would the company selling it spend a lot of money making it?
(I should add, this is purely an economical argument and ignores the working conditions/health effects/etc)
I also have an abused Vornado and it's holding up just fine. Fans are a weird product though, I know most people would just throw them out as soon as they gunk up and stop spinning, but maintenance/repair is really simple compared to other product categories. IMO the business itself has a hard ceiling too, they can't be delusional enough to think they can keep scaling fan sales YOY, so it's one of those deals I imagine would simply be a stable, predictable brand in a portfolio of businesses that somebody (or some company) owns and operates across many niches.
Doc Martens, however, are a good example of a brand/business where LTV might literally be 1, 2, maybe 5 purchases if your feet have stopped growing and you buy their Made in England boots (higher quality, but a price point that isn't LVMH territory) and want different colors -- or maybe a lot more if you buy the shitty ones (vegan leather, plastic-coated leathers) and they fall apart after 9 months and you replace them.
Regardless, a higher quality pair bought in 1994 and beat to hell and back still are very wearable in 2026, while at the same time there are junk options available too if you're so inclined for whatever reason.
I don't personally understand businesses that treat their products as disposable if the product itself isn't intended to literally be disposable by design. I get "make as much money as possible before the spigot shuts off", "endless growth is the goal", etc, but it also seems like it would create all kinds of headaches and byproducts that make operating a brand or business hell on earth (staffing otherwise-unnecessary customer service, dealing with frequent returns and liquidations, constant rebrand/brand management to deal with or paper over negative public perceptions, endless battle with knockoff competitors because your product has no differentiation in quality or value, on and on).
Temu/Shein/etc on the other hand have a distinct advantage in that they're marketplaces and aggregators of invisible brands; they shield "ghost kitchen" product producers from any kind of liability, blowback, negative attention. They're giants in a foreign land that operate with different rules, goals and incentives.
I don't really have a salient point, just thinking out loud. Ultimately, the article is probably correct in that the default, mainstream firehose of digital detritus will settle somewhere near "ultra-processed content", and that will be just fine for the average consumer. Not much we can really do about it with society and economies structured the way they are today. The knock-on effects of this on culture would be borderline impossible to unwind though, which is sad.
For some products, it's viable to sell replacement parts. For example, I have an Ankarsrum blender. The main engine is built to last, but a lot of things are meant to be replaced after some use. They, probably, don't cost as much as buying the whole new blender, but they still generate a revenue stream for as long as the product functions.
Similarly, Kuhn Rikon pressure cooker: it has some parts that will need replacing every few years.
These are, probably, not the best options in terms of revenue generation, but they are options. Sometimes it's possible to sell some kind of service that accompanies the product, or maybe some accessories... However, they aren't available for all sorts of equipment. Some equipment is just too simple to try to organize it in this way.
Maybe we just need to accept that some goods will never be able to benefit from economy of scale and will have to adjust our expectations to the price? :|
Thanks for the laugh with the netflix logo. I feel that is very apt when all their recent shows I've glanced at seem to have their scripts brought from Temu or from a genai prompt.
True only if the end-product is better in some way. A gourmet meal is better than Chipotle. A handmade leather bag is better than the Temu version. In software, it's harder to make out the difference; of course, better UX, faster APIs etc. can all affect the end-user experience. But who is to say that AI-generated code can't do all of that (and more!)?
The whole premise of the post is that if something is cheap, then it has to be of bad quality. However, that’s not the case most of the time. One example is the knock off luxury bags made in China. Even the original ones are actually made in China, but people pay a premium for the brand name, but quality wise both are same.
Same goes for the software. Before AI there were so many buggy, ugly, un-user friendly software. Remember Windows ME? Not all vibe coded software is bad, in fact I’ve vibe coded two projects and use them for free where I previously had to pay 20$ per month for the original one and both projects I vibe coded are working without any issue at all.
I agree with some points, but I don't like the take that AI/LLMs also aren't built on the backs of underpaid workers. While a lot of the training for LLMs may be unsupervised, we wouldn't have gotten to this point in AI without the support of data labelers. And on the hardware side, how many factory workers operate in unsafe conditions on a daily basis?
One of the only people I've seen who gets it, and has a sane take on the matter. I've been trying to write my own article on a very similar subject for a while now, and it's always hard to put these things into words.
I yearn for humanity to stop being so greedy and always looking for the cheapest crap they can get, so they can pretend to be rich by having a lot of stuff. Particularly people in countries like America (where I live), are too comfortable, and pretend to live above their means (not necessarily consciously) by partaking of low quality mass produced goods.
Thought-provoking article, makes sense on some levels but it cuts too broadly.
For physical goods (clothes, purses,..) most buyers can discern high quality goods from bulk manufacturing with cheap materials. When it comes to digital goods, for movies, songs, books, etc. the buyer can discern a clear difference between AI- and human-generated, so the TEMU-fication argument is valid.
But for software engg, design, etc., I suspect that very few people can (or will be able to) discern the difference. There may be people who care about manual output, but then you have to come up with proof of manual work. Something like "prove to me that you have coded this application or designed this site without using AI". I don't see that happening.
I'm saying that people can't discern between human- and AI-built apps, because the vast majority of human-built apps are not any better than AI-built ones.
The article makes absolutely no sense. AI works better than humans. The quality of AI output depends on request. Higher quality request produces high quality output and vice versa. The article completely misses the point.
Beware of the TEMU-fication of AI requests. Some crappy services like Loveable are the violators here.
I personally keep going back and forth on this, partially because I am seeing examples that cut both ways. On the one hand, I am mildly concerned a seemingly permanently raised floor, because I can already see ( and I already heard once or twice ) how now I should have x amount the output. I am saying seemingly, because we just had a meeting where higher ups are clearly concerned about the token cost so some form of 'more with less' is coming.
I am not saying some variant of temufication won't happen. If you look at openai's build week projects, you will see a fascinating array of things, but, also otherwise familiar patterns not completely dissimilar to 'calculator plays yankee doodle(but now llm interprets your next move)'. I guess what I am saying is that some things won't change: interested people will remain interested and use new tooling to its limits.
"The key to this sort of business model is not innovation, but instead the externalization and compression of cost. Somewhere upstream, people work seventy-five hours a week, in conditions most readers of this website would refuse to even visit, so that the rest of us can have a cheap plastic spatula at our doorstep within five business days."
--------
While it may be true in some rare instances, for the vast majority of products its more likely to be fantasy that exists only in stereotypes.
The truth is why American attempts to onshore manufacturing and make it even cheaper with automation have failed miserably. The truth is that Chinese manufacturing is as automated as is possible with cutting edge technology. They build cars in factory's that don't even have lighting until finished cars reach the quality control area. This is why the North American auto sector relies on crushing tariff's and outright import bans to force North Americans to buy North American autos. Without those, you could buy a Chinese car that's better than what you drive for half the price.
The factory cranking out the one-dollar-shoes Temu sells is likely entirely automated.
TEMU sells what was made with the LLM-ification of manufacturing.
47 comments
[ 3.0 ms ] story [ 1532 ms ] threadTEMU only exists because they simply neglected quality of their products which their competitors were required by law to have. And then they were not punished for it as they should have been.
In other words, TEMU only exists because Europe has turned into a deregulated neo-liberal paradise.
I have never, ever, bought anything on TEMU, because anytime I open the website it gives me a gambling roulette for some supposed discount thing. I hate gambling, so everytime I see the roulette I just close the webpage and go do something else.
Well, the whole cheap-generated content craze feels basically like gambling to me to this point, too. Folks are just randomly generating content: books, videos, articles, even applications, and gambling that something will pay for itself.
So now we have Enshitification, Eemufication and Gamblification of the Human Society. Damn I might write something on Substack about that.
Here's an example:
https://news.ycombinator.com/item?id=18442941
Based on what I've seen in industry, I think the answer could be very likely worse. They'll generate bad code at an even faster rate, and very quickly nobody in the company will have any idea how anything works.
What this proves is that "human made" is not automatically good. Better is a complex question. LLMs have a magnifying affect on the quality of human decisions.
TEMU doesn't sell those things. They're more expensive than that, but subsidised by VC capital from PDD Holdings, Blackrock, Vanguard, etc. It's a very boring 'spend the VC bucks to get the customers' play, and when they've got the customers they'll jack up the prices and lose the customers again. It's Uber-for-cheap-stuff basically.
If you're in construction then you need a quality drill that will carry you from one job to the next reliably. If you're just trying to do a one-off drywall repair in your house that you hope never to need to do again, maybe the $10 drill is precisely what you'd buy, if you could find it.
LLMs are the logical conclusion of this scam, until other countries stop gambling on the US stock market with fake $1 trillion corporations.
It works because the whole world's media is captured, owned and censored by US platforms and amplifies the US messaging.
When they invited our professors in the 80s they came back arguing China will eventually be the future. They listened to everything and started working on everything immediately.
We did the yes men privatization garage sale, gutted R&D, outsourcing, abolish education, housing, medical care. All that good stuff - but at what cost?
Isn't that just true about most things over the past century or so? They left out "staggeringly uglier" though.
For example, we have a Vornado fan in my house that is probably from the early 2000s. It still works perfectly even though we've left it running 24/7 for weeks in humid or dusty environments (e.g. to help air dry a wet basement). In terms of money paid/hours used, we got an amazing deal on the purchase.
From the manufacturer's perspective, the life time value (LTV) of us as a customer was basically just one fan + the free marketing of folks like myself writing this kind of post. It reminds me of the joke "Did you hear about the company that made lightbulbs that lasted 30 years? They went out of business!"
It makes sense that manufacturers would start to lower quality at a slightly faster rate than they lower the price as this means that folks still feel happy with the purchase. Sure, the fan doesn't last as long due to what is essentially planned obsolescence but the company does get to stay in business.
Temu/Shein etc are just the logical endpoint of this line of reasoning. If you are going to wear a shirt for the one summer where it's in style, why would you spend a large amount of money on it? And in turn, why would the company selling it spend a lot of money making it?
(I should add, this is purely an economical argument and ignores the working conditions/health effects/etc)
Doc Martens, however, are a good example of a brand/business where LTV might literally be 1, 2, maybe 5 purchases if your feet have stopped growing and you buy their Made in England boots (higher quality, but a price point that isn't LVMH territory) and want different colors -- or maybe a lot more if you buy the shitty ones (vegan leather, plastic-coated leathers) and they fall apart after 9 months and you replace them.
Regardless, a higher quality pair bought in 1994 and beat to hell and back still are very wearable in 2026, while at the same time there are junk options available too if you're so inclined for whatever reason.
I don't personally understand businesses that treat their products as disposable if the product itself isn't intended to literally be disposable by design. I get "make as much money as possible before the spigot shuts off", "endless growth is the goal", etc, but it also seems like it would create all kinds of headaches and byproducts that make operating a brand or business hell on earth (staffing otherwise-unnecessary customer service, dealing with frequent returns and liquidations, constant rebrand/brand management to deal with or paper over negative public perceptions, endless battle with knockoff competitors because your product has no differentiation in quality or value, on and on).
Temu/Shein/etc on the other hand have a distinct advantage in that they're marketplaces and aggregators of invisible brands; they shield "ghost kitchen" product producers from any kind of liability, blowback, negative attention. They're giants in a foreign land that operate with different rules, goals and incentives.
I don't really have a salient point, just thinking out loud. Ultimately, the article is probably correct in that the default, mainstream firehose of digital detritus will settle somewhere near "ultra-processed content", and that will be just fine for the average consumer. Not much we can really do about it with society and economies structured the way they are today. The knock-on effects of this on culture would be borderline impossible to unwind though, which is sad.
Similarly, Kuhn Rikon pressure cooker: it has some parts that will need replacing every few years.
These are, probably, not the best options in terms of revenue generation, but they are options. Sometimes it's possible to sell some kind of service that accompanies the product, or maybe some accessories... However, they aren't available for all sorts of equipment. Some equipment is just too simple to try to organize it in this way.
Maybe we just need to accept that some goods will never be able to benefit from economy of scale and will have to adjust our expectations to the price? :|
Co-worker of mine who was excellent at both system design AND coding and I were chatting about LLMS:
Him: "I guess coding will just be a hobby now"
Me: "Coding def feels that way but software engineering as a profession is still alive and well."
Same goes for the software. Before AI there were so many buggy, ugly, un-user friendly software. Remember Windows ME? Not all vibe coded software is bad, in fact I’ve vibe coded two projects and use them for free where I previously had to pay 20$ per month for the original one and both projects I vibe coded are working without any issue at all.
I yearn for humanity to stop being so greedy and always looking for the cheapest crap they can get, so they can pretend to be rich by having a lot of stuff. Particularly people in countries like America (where I live), are too comfortable, and pretend to live above their means (not necessarily consciously) by partaking of low quality mass produced goods.
For physical goods (clothes, purses,..) most buyers can discern high quality goods from bulk manufacturing with cheap materials. When it comes to digital goods, for movies, songs, books, etc. the buyer can discern a clear difference between AI- and human-generated, so the TEMU-fication argument is valid.
But for software engg, design, etc., I suspect that very few people can (or will be able to) discern the difference. There may be people who care about manual output, but then you have to come up with proof of manual work. Something like "prove to me that you have coded this application or designed this site without using AI". I don't see that happening.
Beware of the TEMU-fication of AI requests. Some crappy services like Loveable are the violators here.
I am not saying some variant of temufication won't happen. If you look at openai's build week projects, you will see a fascinating array of things, but, also otherwise familiar patterns not completely dissimilar to 'calculator plays yankee doodle(but now llm interprets your next move)'. I guess what I am saying is that some things won't change: interested people will remain interested and use new tooling to its limits.
The question to me is what happens in the middle.
--------
While it may be true in some rare instances, for the vast majority of products its more likely to be fantasy that exists only in stereotypes.
The truth is why American attempts to onshore manufacturing and make it even cheaper with automation have failed miserably. The truth is that Chinese manufacturing is as automated as is possible with cutting edge technology. They build cars in factory's that don't even have lighting until finished cars reach the quality control area. This is why the North American auto sector relies on crushing tariff's and outright import bans to force North Americans to buy North American autos. Without those, you could buy a Chinese car that's better than what you drive for half the price.
The factory cranking out the one-dollar-shoes Temu sells is likely entirely automated.
TEMU sells what was made with the LLM-ification of manufacturing.