Comment by _pdp_
15 hours ago
While the frustration is understandable I don't see any difference between this and Netflix not allowing you to use your Netflix subscription in Amazon Prime federated video hub or something of that sort.
At the end of the day we know that these tools are massively subsidised and they do not reflect the real cost of usage. It is a fair-use model at best and the goal is to capture as market share as possible.
I am a no defender of Google and I've been burned many times by Google as well but I kind of get it?
That being said, you don't really need to use your gemini subscription in openclaw. You can use gemini directly the way it was intended and rip the benefits of the subsidised plan.
I developed an open source tool called Pantalk which sits as a background daemon and exposes many of the communication channels you want as a standard CLI which gemini can use directly. All you need is just some SKILL.md files to describe where things are at and you are good to go. You have openclaw without openclaw and still within TOS.
The project is hosted at: https://github.com/pantalk/pantalk
No, it's more like Netflix not allowing you to watch on non-Netflix branded devices or browsers. Or banning you for connecting the wrong TV to a valid device.
Or Microsoft banning you from O365 for not using their browser, or the correct monitor, or the correct mouse or.....
I don't understand. Everyone's been saying LLMs are gonna get cheaper and cheaper, to the point where it's almost free to operate. Clearly becoming profitable won't be a problem... so they can't be subsiding that much...
Are you telling me a bunch of people on Twitter and HN are full of shit?
They are not totally full of it.
But state of the art models are not free. GLM 5 and Kimi K2.5 are both open-source and they are much better models than the ones we used to pay for a year ago. Now we get them for free. This is certainly having an effect on all model providers which either need to adjust to new market realities or risk to loose market share and we know which thing they are not going to do.
You might get access to the model for free. The hardware to do anything useful with it certainly isn't.
Anthropic and Google shutting down access to their API for third party tools, OpenAI inserting ads into the platform... I'm sure it will stop here. Absolutely no more fuckery. And all these huge LLM companies are going to go from burning literally billions (in some case trillions) to being insanely profitable without putting the screws to users. We definitely aren't going to see the same pattern that's played out across essentially every other platform play out again... Nope definitely not.
Model costs have gone down orders of magnitudes in the last few years, and google would stop something like this no matter how profitable Gemini was or wasn't. It's a blatant misuse of their terms.