When you look at AI, right now, it has one major use case that people are really willing to pay for: coding. That means Cursor and, to a lesser extent, Replit. Let’s take Cursor as an example: it is built on top of other companys AI.
This is a problem, because Cursor doesn’t have a service to sell without making calls to other company’s AIs and those companies can raise prices and Cursor has to eat it.
As Zitron notes, this is what actually happened recently:
A couple of weeks ago, I wrote up the dramatic changes that Cursor made to its service in the middle of June on my premium newsletter, and discovered that they timed precisely with Anthropic (and OpenAI to a lesser extent) adding “service tiers” and “priority processing,” which is tech language for “pay us extra if you have a lot of customers or face rate limits or service delays.” These price shifts have also led to companies like Replit having to make significant changes to its pricing model that disfavor users….
…
- On or around June 16 2025 — Cursor changes its pricing, adding a new $200-a-month “Ultra” tier that, in its own words, is “made possible by multi-year partnerships with OpenAI, Anthropic, Google and xAI,” which translates to “multi-year commitments to spend, which can be amortized as monthly amounts.”
- A day later, Cursor dramatically changed its offering to a “usage-based” one where users got “at least” the value of their subscription — $20-a-month provided more than $20 of API calls — in compute, along with arbitrary rate limits and “unlimited” access to Cursor’s own slow model that its users hate.
- June 18 — Replit announces its “effort-based pricing” increases.
- July 1 2025 — The Information reports Anthropic has hit “$4 billion annual pace,” meaning that it is making $333 million a month, or an increase of $83 million a month, or an increase of just under 25% in the space of a month.
In other words, Anthropic, which still isn’t making money even now, increased its prices and Cursor and Replit were forced to pass those price increases on to their customers, and made their products worse.
American AI isn’t profitable. Each call costs more than anyone is charging their customers. And since there are very few AI models (OpenAI, Anthropic and X, basically), anyone who uses these services is subject to having prices suddenly increase. Indeed, since none of these companies is making money, it’s hard to see how anyone could expect anything but price increases.
Now here’s the thing about Deepseek, a Chinese AI. Its run costs 97% less than American AI. You’d think that American AI companies, seeing this, would have looked at how Deepseek did it, but they aren’t, they’re piling on the spending and costs.
And here’s the second thing: Deepseek is open source. You can run it on your own servers and you can build on it.
So: 30x cheaper and you can’t be hit with sudden but entirely to be expected price increases.
Why would you use American AI? (No, it’s not that much better.) The only real reason is legal risk: America wants to win the AI race and it’s willing to use sanctions to do so.
But if you’re in a country outside the Western sphere you’d be insane to use American AI. Absolutely nuts. And even if the Western sphere, building off American AI is incredibly risky.
So Chinese AI is going to win. Sanctions may slow it down, but open source and 30X cheaper is one hell of a combo.
It didn’t have to be like this. OpenAI wasn’t supposed to be a for profit enterprise and Deepseek’s methods of lowering costs could be emulated. But that doesn’t seem to occur to American AI companies.
American tech is completely out to lunch. Absolutely insane. A thirty time cost differential is not something you can just ignore, nor is the fact that American AI companies absolutely will have to raise prices, and raise them massively.
So, yet again, China is going to win, because American corporate leaders are, apparently, morons.
If you’ve read this far, and you read a lot of this site’s articles, you might wish to Subscribe or donate. The site has over over 3,500 posts, and the site, and Ian, take money to run.
david lamy
We citizens of the United States get steeped in the paradigm of American Exceptionalism. AI is no different than our misguided beliefs in our military superiority or superior standard of living.
david lamy
In addition to my previous short comment: if an AI did not hallucinate, steal, and charge what it judges a luser (old time sarcasm for user) can bare it would not be American (of US variety, not Canadian or Mexican).
Wackadoodledoo
Well, except for that study that said programmers using AI are 20% slower than those not using it, so how long are they going to be willing to pay anything at all for it?
elkern
Wall Street is wetting its pinstripes in anticipation of the layoffs they expect AI to enable. It would be rather funny except for all the misery and deaths it will cause.
Misery, mostly for the people who get laid off, their families, and their communities (= all of us).
Death, because GOOD programming is as much about understanding the Real World as it is about coding, and AI doesn’t do ‘understanding’.
I was a professional programmer for the last few decades of my working life (laid off at age 65 when Covid crashed the Aerospace Industry). I was a decent coder, but never the fastest. My value to the company – whether it/they knew it or not – was that I understood the business (and the people), and I could help users get what they really needed to do their jobs (which was rarely what they first asked for).
Smart, young, greedy Zips (see Firesign Theatre for definition) with MBAs will ‘know’ that AI allows them to write their own programs to get whatever they want from whatever database they use. The computers will gladly spew out whatever pile of bits they are asked for, and the MBAs will use the results for Important Company Decisions.
Often, that will work just fine. But sometimes, it won’t, and those are the dangerous cases.
Most meatbags think that GIGO means “Garbage In, Garbage Out”, but us Geeks know it’s really “Garbage In, Gospel Out”.
NR
I would not be at all surprised if AI was terrible at coding. It’s terrible at basically everything else people use it for, hallucinating and making mistakes all the time. We had an example of this in the last open thread here, where a commenter used AI to write their comment and it said something that was flat-out wrong and contradicted by its own sources. Or, for a higher-profile example, look at RFK Jr.’s recent “Make America Healthy Again” report, which was written by AI and cited sources that didn’t exist.
The attraction of AI (for some people) is that it’s quick and easy, but it comes at the cost of quality and accuracy. I guess if you don’t care about those things, it’s fine, but coding is a place where you can’t exactly let that stuff slide.
Anyway, the DeepSeek analysis I saw several months ago said that Open AI is a little bit better at a few things and DeepSeek is a little bit better at a couple of things. But the point is, they’re apparently very close in performance and DeepSeek is significant cheaper (I’ve seen figures of 20x cheaper rather than 30x, but even that’s still a massive difference). And the open-source aspect can’t be discounted either. People complain that the Chinese version has censorship, but you can just download the source code and make your own version without the censorship.
But ultimately, it won’t matter how cheap AI is if it can’t do what you want it to. If all it does is churn out broke, buggy code, there’s no use case for it. We’ll see I suppose.
Like & Subscribe
Surely there is something we are missing in this equation, right? If the goal is to have 100% unemployment for the unwashed then how does the unwashed buy the products and services AI’s advocates promise AI will deliver with massive, unprecedented productivity savings?
They’re not stupid. Egomaniacal, yes and full of hubris but not stupid. They know what we know and yet they proceed. If we read between the lines, it appears we, the unwashed, simply are not part of the future equation. I guess they plan for AI to make it, deliver it, buy it and sell it. AI will become the entire market and the wealthy elite will be relegated to the balcony seating to watch the show.
As far as winning the AI competition, what is the prize aside from ephemeral bragging rights?
bruce wilder
I just feel lost in the abstractions where AI LLM are concerned and that is before the hype generated on behalf of parasitic High Finance and Big Tech take over. Have the “Journalists,” Influencers and PR Hacks been replaced by robots yet. It would be a blessing in disguise if those Doctor Frankensteins were early victims of their Monster.
StewartM
The fundamental error in economic thinking of the Reagan Era is to ignore the incompetency of Wall Street-driven American business leadership. In an era of deregulation and using the stock price as the gold standard of company success, paper machinations, market manipulations, and price gouging will “Trump” any attempt to actually create value. One’s customers are just other sheep to be fleeced along with employees and communities, for the sacrifice to the God of Shareholder Value.
And in the mind of Ayn Rand heroic capitalists, knowledgeable and experienced employees are unneeded and just overpriced. After all, the John Galts and Howard Roarks already know it all, don’t they? Employees (as said by a review of Atlas Shrugged) are just soldiers to be issued barking orders from the all-wise, uber-intelligent capitalist. Who needs experts in health and R&D and land management when you already have Elon Musk? (Uh, how many IT people did Musk fire when he took over Twitter/X?)
In practice, this is Dunning-Kruger on steroids. The US is seeking world-class AI and we’ll get rid of expert (overpriced) coders while doing it, because our genius capitalists know it all already.
Joan
I have been learning and practicing coding languages over the last year in hopes of moving into more technical work. I admit the auto complete does sometimes predict what I need, and in some ways I have learned from it. But ultimately, this is a skill I’m trying to onboard myself. I want it in my brain so I can use it and think with it. If the auto complete didn’t exist I’d just be googling things because there are already great resources out there.
mago
As has been noted AI is just a humongous search engine.
As has been noted it AI centers consume millions of gallons of water, which raises electricity rates which the public pays for, not to mention that water scarcity is a BFD.
As has been mentioned AI reduces human cognitive function.
And it’s been further mentioned that corporate leaders are apparently morons.
It’s all so apparent, yet like every other shit show extant I’m the world it just keeps rolling along . . .
mago
Apologies for the uncorrected auto corrects.
vmsmith
What does it mean to “win the AI race”? What’s the finish line?
Oakchair
When you look at AI, right now, it has one major use case
—-
It has two other current uses.
Slap the marketing term AI on your product and you can sell it at a higher mark up and increase sales at the same time.
The other use is being able to control the information people using AI receive with plausible deniability that you are curating that information.
—-
American AI isn’t profitable…
America wants to win the AI race and it’s willing to use sanctions to do so.
—-
Appears the AI corporation’s plan on using a version of the Uber, Amazon, and Facebook strategy. Yes their public service is unprofitable in a functioning market, but the ruling class is well versed in creating monopoly markets. More importantly controlling the flow of information and data is the actual purpose and it IS highly profitable.
How could you tell the difference between an AI “hallucinating” and an AI struggling to deal with the contradiction of simultaneously providing it’s public service (providing accurate information) and it’s private service (spreading propaganda)?
Flaser
As a sw engineer I’ve yet to see code written by AI that didn’t need massive developer effort to clean up.
AI assistance isn’t any better:
https://www.theregister.com/2025/07/11/ai_code_tools_slow_down/
The mirage of “automated” coding has been with us for decades, from code by contract to various ORM/databinding frame-works to code-snippet generation by better IDE suites.
At the end of the day, the hard part of sw development isn’t writing the code, but understanding and reasoning about its behavior and making changes in a deliberate fashion for chosen goals.
As LLM AIs lacks any true “comprehension” of the code they writes, I’m extremely dubious of their actual utility for sw engineering.
Like & Subscribe
If so, once again I will ask, what does China gain then by winning the AI game? What’s the award or reward aside from bragging rights? Why would China bother playing the game at all and partcipating in the competition of all AI is and will ever be is a “humongous search engine”?
Those are not rhetorical questions, by the way. I am genuinely curious and I’ve yet to see an adequate answer to these questions regarding AI. The answers I do get from critics of AI are riddled with bias and grinding axes.
Carborundum
In a very real way, I expect the real competition metrics of the AI race aren’t going to be how cheap any given model is to produce and run – it’s going to be how one integrates AI more generally with one’s activities. Being able to use it as a bicycle for the mind instead of an ice pick lobotomization.
Having been using a range of models for roughly a year now for what is supposedly the highest ROI AI-enabled activity (coding) I have to say that the experience leaves me a little underwhelmed. Used in particular, very bounded ways it can significantly increase what one is accomplishing – probably not a time saver, but producing improved code quality and maintainability, less brute force approaches, etc. Used badly, in the way that I fear a lot of humans will use it (i.e., “write me code to accomplish large, abstract task X), it produces janky code with no consistency and myriad unexpected results.
Overall, the experience is a lot like being able to “talk” to Stack Overflow, but the outputs suffer from what I’ve come to think of as minimum reproducible example syndrome. Also much, much less good if you work with less mainstream languages / packages.
Moving from that quite bounded domain to more generalized ones (i.e., classify this corpus of text and come up with a particular narrative about it, which is what a lot of tasks boil down to), I fear where this all ends up. Living in a vocal idiocracy is bad enough. An AI-amplified one, particularly one where government makes significant decisions based on AI outputs and the body public can interactively extract whatever “facts” they prefer from the corpus *and* inject them back into that corpus, really does live up to the hellscape label on the tin.
different clue
One reason that China might win the AI race is that the TrumpAdmin is devoted to anti intelligence.
The Maganazi TrumpAdmin believes in prevention of science, suppression of analysis, etc. It is the administration-of-choice for militant backwardite stupidites everywhere.
Here is an example of what I mean. This article is titled: ” White House Orders NASA to Destroy Important Satellite
“This is illegal.” ”
Here is the link.
https://futurism.com/white-house-orders-nasa-destroy-important-satellite
The NASA people can either destroy these satellites ( with the de-orbiting contract going to Elon Musk no doubt) or the NASA people can lose their jobs which means they can go homeless and die in our ” no money = you die” civilization.
Too bad Bernie’s army of small donors didn’t keep themselves together as a group. If they had, they could still be raising all their small donations and pooling them for spending on emergencies like this. They could use all that money to give the NASA people a full-retirement-for-life if the NASA people would refuse to cancel these satellites. And every successive wave of NASA refuseniks could be supported for life for refusing to destroy the satellites. Eventually the TrumpAdmin would run out of NASA people who would even know how to destroy the satellites.
different clue
Oh look, here is a creative “lateral thinking” solution to the problem of launching things which was thought up by natural human intelligence without any AI at all, I bet.
” The insane physics behind a mass accelerator technology designed to move payloads into space by company called ‘SpinLaunch’ ”
https://www.reddit.com/r/interestingasfuck/comments/1mi901d/the_insane_physics_behind_a_mass_accelerator/
One doesn’t need AI to have intelligence or to be intelligent.
Nat Wilson Turner
Like & Subscribe: “If we read between the lines, it appears we, the unwashed, simply are not part of the future equation. I guess they plan for AI to make it, deliver it, buy it and sell it. AI will become the entire market and the wealthy elite will be relegated to the balcony seating to watch the show.”
L&S, you rooted out the truffle. The elites are engineering a mass die-off.