Zone Media.

Hello and welcome to this week's Better Offline Monologue.

I'm your host ed etron over on my premium newsletter.

This week I reported out a troubling situation with AI powered coding environment Cursor maybe, a company called any Sphere who unexpectedly changed their pricing in the middle of June as well as adding a two hundred dollars a month ultra subscription that well pissed everyone off and pissed off a lot of people for many different reasons.

In fact, for some history, Cursor is or was the bell of the ball of the AI industry.

They had five hundred million dollars of annualized recurring revenue, which means their highest month times twelve, making them the first high revenue startup outside of anthropic and open AI to actually show growth potential.

There really is no one else like them out there at this number.

I mean as Alpha sense four hundred million according to the AI data based from the information, but everyone else is powering around one hundred two hundred million arr and raising far more money.

Everyone was really excited about Cursor.

They were growing fast.

Everyone loved them, and it was because their product allowed you to just build software just by typing stuff, and engineers liked it and regular people liked it.

Not saying it was good or anything, just that people liked it, and they liked it because you gave you kind of unrestricted access to ward these models to build all this stuff.

Now, what if I told you that perhaps they had grown to that five hundred million ARR number using techniques like allowing people to spend way more than they were paying for the service.

Yeah, that's exactly what happened.

Previously, Cursor's pricing was pretty simple.

You had five hundred fast completions, meaning that you could ask Cursor to do something using model like Claudes sonnet for or Opus four, and it deduct requests from your allowance.

The amount deductive was based on the model, with some models costing less than others, such as Anthropics claud for Opus costing more.

Once you were out of those fast requests, Cursor would then put you in the slow lane, the unlimited slow lane, I should add, which nevertheless would let you use premium models, but you'd be put in the back of the queue and at peak times could be waiting some time before you get to your completion.

You could also select auto and Cursor would select whatever was available the best model for the job at the time, never want to turn the best over to the company though.

But on June seventeenth, twenty twenty five, Cursor launched that two hundred dollars a month ultraplan along with a new and confusing pro package It's very muddied.

A new wording around said twenty dollars a month plan and how it would work going forward, claiming that they would be rolling out changes to make it more generous, but actually changing pricing to reflect new wording, offering unlimited agent requests and no real explanation as to what that meant, and no further mention of any request limits or what those requests or rate limits might be.

The reality was, of course, far grimmer.

Cursor users have found themselves heavily rate limited, especially on Anthropic models, but the so called unlimited agent requests mostly pushing towards Cursor's on frontier model, which users claim is nowhere near as effective as model SMOP and AI Andthropic.

On June thirtieth, Curser made another change to their product features, changing unlimited agents to extended limits on agents, further money in pricing on a product that was once renowned for its simplicity and lack of limits.

So this isn't good, is it.

Silicon Valley's favorite coding startup has in the last few weeks completely changed how its customers interact with its product, both degrading the service and making it far more expensive in the process.

They also allow you now to use usage based pricing.

Now, this is the funny one.

If you pay twenty dollars a month on Cursor, they will guarantee you twenty dollars a month of AI compute at least, so they're literally giving money away.

They add a twenty percent fee on top of compute.

But wow, so you're making twenty percent morgin.

Now you're fucking not.

They're losing money on everything.

I'm sure of them.

But it's strange.

It's strange, and it's bad.

But this is also textbook and shitification.

Corey doctor O's term for when platforms offer a high quality product to gain a large user base, usually through convenience or great value.

Then the grade the service over time to make more money as a means of maximizing value shareholders or making money themselves.

It's also part of my rot economy.

Thesis.

The growth of all costs thinking has dominated the tech industry, and I'd argue, thanks to the proliferation of business idi it's controlling everything that they've drained some of the logic behind in certification away, because very few of these companies actually have a plan for sustainability, let alone profitability.

Generally, in certification gets people through the door and makes the service totally impossible to avoid, makes it essential.

That's really difficult to do with a paid software product like Cursor, because AI compute is so fucking expensive.

It's so expensive, so Cursor has decided to en shittify without making sure that they have something essential.

But the thing is, I had texted Cory about this.

I think this actually is something quite different.

I think this is the world's first chain in shitification.

My belief is simple.

Any Sphere, which is the company that makes Cursor, by the way, is despite getting nine hundred million dollars in funding in early May, running out of money, or at least believes that continue to operate its business in the way it did less than a month ago would cause it to do so.

But why the sudden changes, Why the knife in the heart of their customers just after raising nearly a billion dollars.

It's simple, it's a chain and citification.

Anthropic jacked up their prices, and so did open Ai.

On May twenty second, twenty twenty five, a few weeks after Cursor raised nine hundred million dollars and Thropic launched both claud Sonet and claud Opus four.

You might say Claude four Opus of Claude force Sonet.

I don't care.

But these were two new powerful as judged by benchmarks made specifically for large language models.

And I should add that the expensive claud Opus four was more focused on coding benchmarks.

This is an important detail now.

Eight days later, on May thirtieth, twenty twenty five, patron Anthropics API documentation appeared for the first time called service tiers, adding priority tiers for enterprise stats that didn't want to and I quote sorry, I mean did want to and I quote provide a guarantee around the infrequency of server overloaded errors even during peak time.

Hmm, that's not good.

Anthropic service tiers require a multi month upfront commitment on how many tokens per minute your startup will use, but also add an insidious charge around prompt casing.

Now Prompt cashing is when when you put something into a model, such as a code base in the case of a coding startup, or I don't know a great deal of stuff about how a model might want to act, you put it in the cash so that it kind of like RAM, it just kind of reads off of it instead of rewriting and rewriting and rewriting.

So they've added a vig to it or attacks whatever you call it, a toll perhaps, and they're now charging and this is really fucking insidious, either one hundred and twenty five percent or two hundred percent of the cost of casing information that you need to access it more readily.

And I should add this is only if you want the Priority tier.

And another thing, coding startups are extremely prompt cash heavy, kind of fucked up right, very fucked up, and they're not the only ones doing it.

On June twenty fifth, twenty twenty five, open Ai also launched Priority processing for API customers.

They claim it's the payers you go to where you pay higher API prices for and I quote predictable low latency.

However, despite this being payers you go, the service is only available for their enterprise customers who have made an up prompt commitment much like Anthropic does, the difference being they are not trying to tax you on prompt caching now.

Cursor CEO Michael Truell also said when announcing their agree just two hundred dollars a month plan that it was only made possible by multi year partnerships with OpenAI, Anthropic, Google, and Xai.

In short, I think that nine hundred million dollars that Cursor got may have immediately been handed to or committed to the major model developers.

They got looted baby.

And now what makes this weirder is that Cursor is also Anthropic's largest customer, with sources claiming that they're such a large customer that they're taxing anthropics infrastructure and making them run out of GPUs.

Now.

I should also add that Cursor isn't the only one that's recently dramatically changed their prices.

Vibe coding startups like replt and Lovable Bits both have had to do so in the last month.

And I think this is just the beginning of something really, really, really bad.

I believe that Anthropic did this either as a deliberate attempt to price gouge its largest customers, and or as a means of increasing revenue and its money losing software.

These changes that deliberate aggressive and targeted price increases, and they were timed with the launches of claud for Opus and Sonnet, which suggests that Anthropics costs have dramatically increased with these models, so much that simply increase in the cost per million tokens is in sufficient.

It's my belief that the launchers of both claud Opus and to a lesser extent, son It have caused en up hevil in Anthropics costs and compute demands, which in turn forced them to start increasing costs on their customers.

However, Anthropic has I believe, realized that there is no real way to just increase the cost of Opus and Sonnet for further on just the amount of tokens that the customer might use, and that doing so might push away smaller developers, which wouldn't make them more money.

So they decided to find a way to specifically exploit the finances of their largest customers coding startups, in a way that won't be immediately obvious or that would spook non coding assistant customers.

Except I'm a crafty little fuck and I look at everything all the time.

I do this for fun.

I don't do this because I have to listen to me.

Wario Amadee.

If you're listening to this, I'm watching you.

I have archive dot org and a million diet cokes.

Anyway, last year I talked about the subprime AI crisis, where almost the entire tech industry is bought in on a technology sold at this vastly discounted rate, and they've heavily centralized and subsidized it too.

And I predicted one major thing that these model developers would eventually have to find a way to make their costs work, that they would have to find a way to crank up costs on their customers, because otherwise they would just continue burning money.

I still think they are doing so, and I think that's what's happening here.

I think anthropic and OpenAI to a lesser extent of realized that they need to start making money back on these fuckers, and they're doing it.

Except the fact that the companies of immediate companies like Cursor and Repler have immediately had to change their prices suggest that maybe not one of these businesses makes sense.

Maybe this whole time this thing was unsustainable.

Maybe it turns out the general if AI doesn't have the kind of business returns you need to run a startup.

If only someone had said something.

Remember, look see how fucked up saying Remember but I'm not editing it out.

This is an honest podcast.

But seriously, everyone, remember the pale horses.

Rate limits, service interruptions, price increases, trouble raising funding, and trouble with money.

The horses are drawing nearer and telling you we're coming to the end of this.

I don't know if it will be soon, I don't know if it'll be next year.

But nothing about this suggests that things are going well.