Comment by cogman10
15 hours ago
Oh man.
What a wonderful way to stop people from using your LLM.
All these AI companies trying to get everyone to be locked into their toolchains is just hilariously short sighted. Particularly for dev tools. It's the sure path to get devs to hate your product.
And for what? The devs are already paying a pretty penny to use your LLM. Why do you also need to force them to using your toolkit?
There is a reality that when they control the client it can be significantly cheaper for them to run: the Claude code creator has mentioned that the client was carefully designed to maximise prompt caching. If you use a different client, your usage patterns can be different and it may cost them significantly more to serve you.
This isn't a sudden change, either: they were always up-front that subscriptions are for their own clients/apps, and API is for external clients. They don't document the internal client API/auth (people extracted it).
I think a more valid complaint might be "The API costs too much" if you prefer alternative clients. But all providers are quite short on compute at the moment from what I hear, and they're likely prioritising what they subsidise.
It reminds me of the net neutrality debate from a decade ago. I'm not American but I remember the discord and online hate towards Ajit Pai when they were repealing it.
On one side you had the argument that repealing net neutrality would mean you can save money on your internet bill by only paying for access to what you use. On the other, you had the argument that it would just enable companies to milk you for even more profit and throttle your connection as they see fit.
IMO we need 'net neutrality' for LLM clients. I feel like AI companies are hypocrites for talking about safety all the time, but want us to only use their LLMs in the way they intend. They're saying we're all going to be replaced by AI in 12 months, and we have to use their tools to survive, right?
Yann LeCun recently warned that the AI coming out of China is trending towards being more open than the American alternative. If it continues like this, I can see programmers being pushed towards Chinese models. Is that what the US government wants?
Use of Chinese models: If I had not got a discount for signing up for a full year of Gemini AI Pro for something like $14/month, I might have started just using a Chinese chat model for things where privacy is not an issue. Ironic that I am now paying for both Gemini AI Plus and also $20/month for Ollama Cloud (as a super easy way to experiment with many open models). I am also paying Proton $10/month to use their handy lumo+ private chat service built on Mistral models. I feel like I am spending too much money but I don’t want to feel locked into just a few vendors, and to be honest it is fun having alternatives. A year ago I used APIs for Chinese models (and Mistral in France) and the cost was really low.
I imagine its a case of the providers not wanting to admit its costing them a fortune because suddenly all these low-medium usage accounts are now their highest use ones.
Not saying it's right. But it's also not exactly a secret that they are all taking VERY heavy losses even with pricey subscriptions.
> But it's also not exactly a secret that they are all taking VERY heavy losses even with pricey subscriptions.
It's absurd, there's people out there paying $200 for the equivalent of $1600 in API credits. Of course there's a catch! What did you expect!
https://bsky.app/profile/borum.dev/post/3meynioealc2x
That tool is "ccusage" if you're a Claude subscriber and want to see what the damage will be if/when Anthropic decides to pull the rug.
its 200 to 6000 and I use the 6000. I also use an antigravity subscription for probably another 6k (I don’t use them fully tho,)
I cant believe this is net positive for them.
Google has been particularly pernicious in the corporate exercise of zero-tolerance.
Because of their large footprint in so many areas, it is wise to greatly (re)consider expansion in the ways that you rely on them.
Antigravity is useless anyway. I tried it last week and it needs approval for every file read and tool call. There's an option in the app to auto-approve, except it doesn't work. Plenty of complaints online about this. Clearly they don't actually care about the product, some exec just felt that they need to get into the editor game.
Next I tried using the Antigravity Gemini plan through OpenCode (I guess also a bannable offense?) and the first request used up my limit for the week.
Hopefully this gets people to stop using Google for more than just LLMs.
The devs are paying to use the UIs provided by the company. The usage-based API is a separate offering, and everyone knows that.
It's okay to be annoyed at being caught, but honestly the deer in the headlights bit is a bit ridiculous.
If you want to use an API, pay for the API option. Or run your own models.
The tool thing is kind of infuriating at the moment. I've been using Claude on the command line so I can use my subscription. It's fine, but it also feels kind of silly, like I'm looking at ccusage and it seems like I'm using way more $ in tokens than I'm paying for with the subscription. Which is a win for me, but, I don't really feel like Claude Code is such a compelling product that it's going to keep me locked in to their model, so I don't know why they're creating such a steep discount to get me to use it. I'm perfectly fine using Codex's tools, or whatever. I dunno, it seems like way more cost effective to use the first party tools but I'm not sure why they really want that. Are the third party tools just really inefficient with API usage or something?
> I dunno, it seems like way more cost effective to use the first party tools but I'm not sure why they really want that. Are the third party tools just really inefficient with API usage or something?
No, the first party tools, even if they used the same number of tokens, gives them valuable data for their training.
Essentially, the first party tools are subsidised because it saves them money on gathering even more training data. When you use a 3rd party tool, you are expected to pay the actual cost of each token.
Whether the tool is first-party or third-party, they still see the entirety of the prompt, which is where the valuable training data is.
You are being subsidised to the tune of 50 to 99.9 cents on the dollar compared to the API.
What the hell do you expect? To get paid for using other people's tools on Google's servers?
Businesses do not have an entitlement to profit. Suspending customers for using a fairly expensive subscription plan -- especially forfeiting an annual prepayment for a day or two of coloring outside the lines -- sure does make Google appear entitled to profit without ever risking its own pricing model.
> Suspending customers for using a fairly expensive subscription plan -- especially forfeiting an annual prepayment for a day or two of coloring outside the lines
they're being suspended for using a private api outside of the app for which the api was intended. If you make a clone of the hbo app, so that you can watch hbo shows without ads by logging in with your discounted ads-included membership, your account will also be suspended.
8 replies →
> Suspending customers for [snipped]
They are being banned (not suspended) for breaking the ToS, not for what you imagine them to be suspended for.
It doesn't matter how expensive a provider plan you purchase, the provider is free to end their contract with you, permanently if they want to, if you breach their terms of service.
You also get the same freedom.
1 reply →
Equally, customers are not entitled to make set the terms, or pricing decisions for businesses. They can always move their custom elsewhere if they disagree with ToS or pricing.
1 reply →