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.
Maybe there's some loss leader going on, but comparing it to "VC capital" (venture capital... capital?) doesn't make much sense, any more than calling costco's rotisserie chicken and hotdogs "subsidised by VC capital" doesn't make much sense.
I'm not 100% sure but I also heard China subsidizes shipping of packages, to boost their economy.
I worked as a mailman a while back, most days there would be more packages in grey plastic wrap (from China) than actual mail. And you could tell that the stuff that was inside usually cost less than the shipping would have cost if you paid full price.
Shipping is also subsidized by all of us on this end. International postage agreements were never intended to cover a massive unidirectional deluge of low-value packages. The costs of all the additional logistics are ultimately externalized to the taxpayer, and China totally exploits this. It’s exactly the reason the EU set a tariff for low-value parcels in particular.
If someone says to you "You can buy this widget that costs $10 to make for $5" that doesn't mean they they're selling it for $5. It means there's more going on and that the $5 price take is not the full story.
You may choose to ignore that and believe someone is selling you something for $5 but that doesn't make it true.
In the case of Temu, they're selling it for $5 of your money, and $5 of someone else's money, in order to capture you as a customer in the hope that you continue to buy things when they increase the price to $6, then $8, then $10, $15, $30 and so on in the future.
This is the Uber playbook, and if you have a solid product that people want and you can spend VC money long enough to build a trusted brand (or ideally keep prices artificially low until the competition gives up and you have a monopoly) it's very lucrative. I don't see it working for Temu myself. They sell things people don't need so as soon as the prices go up they'll probably fail. That said, Amazon see them as enough of a threat to launch Haul, so what do I know?
> You may choose to ignore that and believe someone is selling you something for $5 but that doesn't make it true.
Sir, I suggest you use the word "sell" in its usual way, not your own definition of it. If someone takes $5 and gives you in exchange the merchandise and right of ownership on it, then that indeed does mean they "sell it for $5". The fact that something else is going on doesn't make it false.
I suggest that the passage that you read begins with a claim that seems unreasonable ("but I've bought a $2 dress from Temu!"), and then explains its thesis in order to make you view that claim in a different way ("VC capital paid for half of that dress!") I also would like to mention that this is a very common and unremarkable way to begin an argument.
All that aside, I’m happy that VCs are paying for my stuff! I have no idea why they keep doing that. They’ve studied my behaviour so they should already know I’ll delete {app_name} as soon as I get a better deal elsewhere. (Better yet, I’ll keep both and let them fight!)
Maybe they can capture the market for a brief moment, but if they do have a long term strategy, I’m not seeing one.
The whole premise of the post is that if something is cheap, then it has to be of bad quality. However, that’s not the case most of the time. One example is the knock off luxury bags made in China. Even the original ones are actually made in China, but people pay a premium for the brand name, but quality wise both are same.
Same goes for the software. Before AI there were so many buggy, ugly, un-user friendly software. Remember Windows ME? Not all vibe coded software is bad, in fact I’ve vibe coded two projects and use them for free where I previously had to pay 20$ per month for the original one and both projects I vibe coded are working without any issue at all.
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!)?
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)
> 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.
Given there's zero barrier to substitution + the TAM for fans and lightbulbs is huge as sibling comment mentioned, there's no reason for the LTV to meaningfully exceed 1 puchase and the value of that free marketing could be higher than the EV of purchase N+1
E.g., if my name-brand turd wears out unreasonably, I don't replace it with the same name-brand turd; instead I replace it with an ephemeral Chinese brand turd which is just as likely to have similar build quality and features.
Seems to me that the TAM for a fan is basically infinite for any single company, therefore making a product that is too high quality or long lasting could still get you growing yoy sales for a long time and stay in business.
But you’re probably right that vs competition lowering prices (and quality) they’d lose.
There’s probably an optimum for maximizing sales/growth that doesn’t land on highest quality products in the real world, with non rational consumers, imperfect information etc… sadly.
Next to the price tag, it should include the warranty. Similar to how calories are listed next to the price tag for food. Just a thought, not sure if it would work.
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.
Temu built out some local distribution after it had already reached a pretty high level of popularity, but Temu was enabled by regulatory and political indifference that allowed China, with an enormous GDP, to get extremely subsidized postage rates. Also, lax enforcement of import duties (do you think the declared values were accurate?), and effectively zero enforcement of any safety standards, etc., on the products sold.
So, it was less of them following a path Amazon had trod and was maintaining, and more like sneaking in a back door and trying to sell you a counterfeit purse.
But if the US and EU had been on their game, it'd of either not happened or would've at least happened much more slowly.
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.
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.
The gambling roulette is triply annoying because it will inevitably start walking you through some flow where you're supposedly ending up with a big discount or free items or whatever-the-fuck. And if you're encountering it for the first time, you'll think "ah cool, i can get some cheap chinese gewgaws for free with the thing I actually wanted to buy." And then 10 minutes has gone by and 5 roulette spins and there's $400 worth of bullshit in your cart.
And the thing is, Temu - I was going to buy something from you. I saw an ad or whatever and thought "sure, I'll pay $13 for that keyboard." But then you get pulled into the dopamine labyrinth hell-realm and by the time you're out they've wasted 15 minutes of your time and you don't even remember what you came in for.
For sure, the ridiculous promo games are best skipped (when even possible to skip!!) as 99% of the time they're just that you win the right to get a discount on a future order based on this next order. And those discounts are always described as like "a $400 credit" ... when it's really the sum of "50 cents off of $5, $1.50 off of $10... $30 off $250" and so on, so you will have to place like 8 orders of increasingly hilarious value to "use" all of your "$400."
All that said, there have been a decent number of things that I bought from them which I've found incredibly useful. Just like, phone stands, ODBII reader, sticky notepads that look like Windows 95 Notepad windows, teeny tiny wireless earbuds I use while sleeping, etc.
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.
Show me a fully LLM-generated database with an implement that's clear, documented, and adequately bug free.
Codebases degenerate when the authors do not understand the system in which they operate and make changes that degrade the cohesion of the system as a whole. Humans do this all the time; at my company we had a 10k line lua function (within a 250k loc lua service) that grew through ten thousand point modifications. It's bad and humans caused it. Now, it's 15k lines of code and 10k lines of misleading slop comments.
LLMs do not go through the struggle of understanding and don't experience the distress of "wtf am I reading?" and don't experience the drive to make things less horrible for the next person. That suffering loop is how many codebases halt degradation and in the absence of suffering LLMs will pile bugs and rot endlessly.
I was arguing with a colleague today that we will never have a reason to make our AIs autonomous and sentient. But now I can't stop imagining an AI startup meeting where somebody says "LLMs do not go through the struggle of understanding and don't experience the distress of "wtf am I reading?"" and the next second the technical founders jump in joy "...that's our way in! Let's outcompete the competition by inventing pain for AIs!"
Log in, pay for premium, and use the like and subscribe buttons liberally. Use the downvote button too. It learns.
My feed is really fantastic. Once a month or so I notice some AI slop (usually with a handful of views - clearly the algo testing something out). Thumbs down, problem solved.
Fair, but I don't use YouTube as much anymore. Except for music and old videos it lost its purpose for me. I mostly rely on word of mouth now, a friend recommends me this art channel or that show. If I like it I forward it to another friend etc.
I still feel bad for what is happening though, it's wrong and anyone with our best interests in mind would agree generated content needs to be firewalled from the human internet.
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.
>They're more expensive than that, but subsidised by VC capital
They're being subsidized? How do you explain their fat margins, $15.4B of profits on $54.0 of revenue? https://en.wikipedia.org/wiki/Pinduoduo
Maybe there's some loss leader going on, but comparing it to "VC capital" (venture capital... capital?) doesn't make much sense, any more than calling costco's rotisserie chicken and hotdogs "subsidised by VC capital" doesn't make much sense.
I'm not 100% sure but I also heard China subsidizes shipping of packages, to boost their economy.
I worked as a mailman a while back, most days there would be more packages in grey plastic wrap (from China) than actual mail. And you could tell that the stuff that was inside usually cost less than the shipping would have cost if you paid full price.
Shipping is also subsidized by all of us on this end. International postage agreements were never intended to cover a massive unidirectional deluge of low-value packages. The costs of all the additional logistics are ultimately externalized to the taxpayer, and China totally exploits this. It’s exactly the reason the EU set a tariff for low-value parcels in particular.
It's not hardly an exploit when the taxpayer is the one benefiting
It’s a classic negative externality. Not every taxpayer partakes in this pathological consumerism.
> TEMU doesn't sell those things.
Oh? https://imgur.com/a/tjAdSQL
If someone says to you "You can buy this widget that costs $10 to make for $5" that doesn't mean they they're selling it for $5. It means there's more going on and that the $5 price take is not the full story.
You may choose to ignore that and believe someone is selling you something for $5 but that doesn't make it true.
In the case of Temu, they're selling it for $5 of your money, and $5 of someone else's money, in order to capture you as a customer in the hope that you continue to buy things when they increase the price to $6, then $8, then $10, $15, $30 and so on in the future.
This is the Uber playbook, and if you have a solid product that people want and you can spend VC money long enough to build a trusted brand (or ideally keep prices artificially low until the competition gives up and you have a monopoly) it's very lucrative. I don't see it working for Temu myself. They sell things people don't need so as soon as the prices go up they'll probably fail. That said, Amazon see them as enough of a threat to launch Haul, so what do I know?
> You may choose to ignore that and believe someone is selling you something for $5 but that doesn't make it true.
Sir, I suggest you use the word "sell" in its usual way, not your own definition of it. If someone takes $5 and gives you in exchange the merchandise and right of ownership on it, then that indeed does mean they "sell it for $5". The fact that something else is going on doesn't make it false.
I suggest that the passage that you read begins with a claim that seems unreasonable ("but I've bought a $2 dress from Temu!"), and then explains its thesis in order to make you view that claim in a different way ("VC capital paid for half of that dress!") I also would like to mention that this is a very common and unremarkable way to begin an argument.
All that aside, I’m happy that VCs are paying for my stuff! I have no idea why they keep doing that. They’ve studied my behaviour so they should already know I’ll delete {app_name} as soon as I get a better deal elsewhere. (Better yet, I’ll keep both and let them fight!)
Maybe they can capture the market for a brief moment, but if they do have a long term strategy, I’m not seeing one.
None of that changes the fact that they are indeed selling it to me for $5 today, so the sentence you quoted is indeed accurate.
The use of 'loss leaders' has a long history, so is your point that Temu's use of loss leaders may be illegal?*
(*I poked around a bit, and this may be possible, especially in California.)
GP means that TEMU is just the marketplace. They aren't actually selling you the drill, just facilitating the sale.
The whole premise of the post is that if something is cheap, then it has to be of bad quality. However, that’s not the case most of the time. One example is the knock off luxury bags made in China. Even the original ones are actually made in China, but people pay a premium for the brand name, but quality wise both are same.
Same goes for the software. Before AI there were so many buggy, ugly, un-user friendly software. Remember Windows ME? Not all vibe coded software is bad, in fact I’ve vibe coded two projects and use them for free where I previously had to pay 20$ per month for the original one and both projects I vibe coded are working without any issue at all.
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!)?
> A 2025 Veracode report found that approximately 45% of AI-generated code samples failed security tests and contained critical vulnerabilities
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."
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)
> 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.
Given there's zero barrier to substitution + the TAM for fans and lightbulbs is huge as sibling comment mentioned, there's no reason for the LTV to meaningfully exceed 1 puchase and the value of that free marketing could be higher than the EV of purchase N+1
E.g., if my name-brand turd wears out unreasonably, I don't replace it with the same name-brand turd; instead I replace it with an ephemeral Chinese brand turd which is just as likely to have similar build quality and features.
Seems to me that the TAM for a fan is basically infinite for any single company, therefore making a product that is too high quality or long lasting could still get you growing yoy sales for a long time and stay in business. But you’re probably right that vs competition lowering prices (and quality) they’d lose.
There’s probably an optimum for maximizing sales/growth that doesn’t land on highest quality products in the real world, with non rational consumers, imperfect information etc… sadly.
Next to the price tag, it should include the warranty. Similar to how calories are listed next to the price tag for food. Just a thought, not sure if it would work.
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.
Temu built out some local distribution after it had already reached a pretty high level of popularity, but Temu was enabled by regulatory and political indifference that allowed China, with an enormous GDP, to get extremely subsidized postage rates. Also, lax enforcement of import duties (do you think the declared values were accurate?), and effectively zero enforcement of any safety standards, etc., on the products sold.
So, it was less of them following a path Amazon had trod and was maintaining, and more like sneaking in a back door and trying to sell you a counterfeit purse.
But if the US and EU had been on their game, it'd of either not happened or would've at least happened much more slowly.
The US has been exporting Treasury bills and digital hot air for quite a while.
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.
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.
I have another weak correlation to TEMU to add.
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.
The gambling roulette is triply annoying because it will inevitably start walking you through some flow where you're supposedly ending up with a big discount or free items or whatever-the-fuck. And if you're encountering it for the first time, you'll think "ah cool, i can get some cheap chinese gewgaws for free with the thing I actually wanted to buy." And then 10 minutes has gone by and 5 roulette spins and there's $400 worth of bullshit in your cart.
And the thing is, Temu - I was going to buy something from you. I saw an ad or whatever and thought "sure, I'll pay $13 for that keyboard." But then you get pulled into the dopamine labyrinth hell-realm and by the time you're out they've wasted 15 minutes of your time and you don't even remember what you came in for.
For sure, the ridiculous promo games are best skipped (when even possible to skip!!) as 99% of the time they're just that you win the right to get a discount on a future order based on this next order. And those discounts are always described as like "a $400 credit" ... when it's really the sum of "50 cents off of $5, $1.50 off of $10... $30 off $250" and so on, so you will have to place like 8 orders of increasingly hilarious value to "use" all of your "$400."
All that said, there have been a decent number of things that I bought from them which I've found incredibly useful. Just like, phone stands, ODBII reader, sticky notepads that look like Windows 95 Notepad windows, teeny tiny wireless earbuds I use while sleeping, etc.
They appear to be targeting users with underdeveloped prefrontal cortex. I wonder what the intention behind that might be.
> A hypothesis on the not-so-distant future ... in which most of what we consume gets cheaper, more abundant, and noticeably worse
Isn't that just true about most things over the past century or so? They left out "staggeringly uglier" though.
An exception is probably cars, they are better, more expensive but uglier too.
Cars are slowly becoming subsidized by subscriptions and advertisements. And to some extent, costs of maintenance and gasoline.
Why does Temu take the heat when Aliexpress has been around for years?
Because Temu pushed advertising hard and fast at the US market and it paid off in a big way.
Temu had a Superbowl ad. Not sure AliExpress did that.
Temu is Temu brand AliExpress.
"Human made" is not automatically better and we need to stop pretending that it is.
Here's an example:
https://news.ycombinator.com/item?id=18442941
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.
What you are describing is not bad software, it's complex, performant and software with a long lifetime.
Show me a fully LLM-generated database with an implement that's clear, documented, and adequately bug free.
Codebases degenerate when the authors do not understand the system in which they operate and make changes that degrade the cohesion of the system as a whole. Humans do this all the time; at my company we had a 10k line lua function (within a 250k loc lua service) that grew through ten thousand point modifications. It's bad and humans caused it. Now, it's 15k lines of code and 10k lines of misleading slop comments.
LLMs do not go through the struggle of understanding and don't experience the distress of "wtf am I reading?" and don't experience the drive to make things less horrible for the next person. That suffering loop is how many codebases halt degradation and in the absence of suffering LLMs will pile bugs and rot endlessly.
I was arguing with a colleague today that we will never have a reason to make our AIs autonomous and sentient. But now I can't stop imagining an AI startup meeting where somebody says "LLMs do not go through the struggle of understanding and don't experience the distress of "wtf am I reading?"" and the next second the technical founders jump in joy "...that's our way in! Let's outcompete the competition by inventing pain for AIs!"
You choose to consume bad media, not my problem.
Have you tried YouTube recently?
Simply FINDING good media becomes the problem. I have no interest in AI slop, but it's everywhere.
I don't choose to consume bad media, but finding good media is getting harder and harder.
I use YouTube all the time and it's great.
Log in, pay for premium, and use the like and subscribe buttons liberally. Use the downvote button too. It learns.
My feed is really fantastic. Once a month or so I notice some AI slop (usually with a handful of views - clearly the algo testing something out). Thumbs down, problem solved.
Fair, but I don't use YouTube as much anymore. Except for music and old videos it lost its purpose for me. I mostly rely on word of mouth now, a friend recommends me this art channel or that show. If I like it I forward it to another friend etc.
I still feel bad for what is happening though, it's wrong and anyone with our best interests in mind would agree generated content needs to be firewalled from the human internet.