Hacker News Reader

Comment by wahnfrieden | Hacker News Reader

They are now a Codex clone and without the subscription pricing. You have to spend thousands to get what you get from a $200 Codex subscription. How do they compete with this except from users who haven't caught on yet, or businesses that are unbothered to spend thousands a month per dev and wouldn't consider just subscribing to 1-3 $200 subscriptions instead?

And their price is so high because it's markup on API rates. API rates, even without markup, are just insanely irresponsible for anyone to be spending on full-time daily usage.

View on HN

> users who haven't caught on yet

They are catching up fast!

https://www.businessinsider.com/chamath-palihapitiya-ai-cost...

Tellingly, from his full post: "Mostly because I do not yet see an equivalent uptick in productivity or revenue..."

https://x.com/chamath/status/2029634071966666964

I suspect that as the value a company provides is more than its code, then increasing code churn does not lead to an equivalent increase in revenue. Even for a tech company, a business' concept, connections, knowledge, assets, non-coding staff, etc.. are a significant value and increasing code doesn't increase the throughput of that value. For non-tech companies code is the grease in the gears, not the gears themselves.

Codex is coming for those non coding use cases too. Is Cursor?

Whose pricing is above API rates? Not Cursor. It's 100% at each model provider's published API rate. With a bigger sub, you get it cheaper than that.

Cursor makes a ton of money because the product is great. It's easily the most sophisticated harness out there, and it isn't an IDE anymore. It's an agent dashboard since version 3.

Suffice it to say it's not all idiot money being thrown at them by users.

API rates on local models are quite cheap, and you can even run them locally. Yes, the hardware for doing so at speed is expensive, but people used to drop the equivalent of what would be $50k or $100k today on an individual workstation for full-time use. It's justified if the productivity gain is strong enough.

But that’s not competitive. The only reason to do that is out of need for privacy. Which is critical for some. The tradeoff is that the models are relatively bad. I don’t see how Cursor can win from this use case especially if to get the privacy benefit you need to spend a huge amount. You can already run Codex for free with local models too.

What's the advantage over github copilot actually? They seem to have all the same access and features (except for this sheduling thing?) for cheaper.

> users who haven't caught on yet

If you think this of users who use cursor then I don’t think you’ve used cursor much at all.

What do you mean?

Only the foundation model companies offer cheap/subsidized compute.

If you're an app layer company, you're offering a 10x worse deal to your customers.

Foundation model companies are willing to lose money to win loyalty. Remains to be seen if it'll work.

If you’re more worried about cost than you are being productive and getting good results then sure, stick with foundational model company apps.

“Being productive” without taking inputs/costs into consideration is an oxymoron.

A company that cares more about cost than results is probably a terrible company to work for. They will give you 10yo dell laptop with 8gb memory and complain that you’re slow when it takes 15m to build the application.

So no it’s not an oxymoron.

Productivity is literally a statement of the relationship between the result and the cost, presumably you found that out after reading the reply and that is why you switched from "productivity" to "results" in your reply.

Until you learn what productivity is we can’t continue the conversation.

Please at least try to keep track of which sockpuppet you are using in this thread sighthrowaway.

But euros spent on tokens is a tiny fraction of the overall costs of the project.

That’s the thing, I have never seen detailed costs of what people are spending their money on. I know that for Claude there’s a $200 monthly subscription through which assigned credits one burns pretty fast, at which point (and I may be wrong on this, because I’ve never used the thing) one can run extra code on a “pay as you use it” basis? Again, I might be wrong on this.

I’ve also seen it mentioned a lot of people having 2, 3 or even more subscriptions, which I’m pretty sure that can easily go South when it comes to costs.

But, again, and the most important point, I’ve never seen a detailed post on what people spend on this AI thing on a monthly basis (let’s say).

I've used Cursor a lot. Until recently it was mandated by my employer. I can't see the attraction at all. It's a (bad IMO) IDE integration, a reasonable model (but I still generally preferred Claude over Composer), and a bunch of other tools that weren't very developed (like cloud environments and multi-agent orchestration). It's a suite of tools, most of which have superior alternatives. What am I missing?

You have model choice in cursor… why would you use composer?

API rates are the real rates. Subscription costs are the "first hit is free" subsidized pricing.

They’re not the “real rates”, they’re the rates that are being charged for API use. API reportedly has a margin of profit

You also neglect that products like Cursor run on two margins, their own plus the API provider’s. That’s always going to come at a premium

Yes, the rates with a margin of profit are the real rates.

The rates without a margin of profit (or with a negative one) are not real.

Reply on news.ycombinator.com