The title is self-explanatory so im editing this section…
im nearly positive cursor is shadowbanning my post so im going to use this one since everyone can see it
first of all the devs said specifically, “if this gets enough engagement, we will consider thsi” im not sure what constitutes more engagement, but i wanted to point out a few things on my mind that i cant figure out about cursor
- i used openrouter to test new models on cursor and on vs code extensions such as roo code, the day they are available in open router i tried both extensions which had no time to update and roo code works flawlessly everytime, and cursor agent mode is extremely broken with the same models that work perfectly with other extensions agent mode
even worse, using the same api key, cursor says i am rate limited with my own api key model… which makes no sense when i can still use it perfectly fine in the other extension, this means to me that they are doing something really funky on purpose to any other 3rd party model on purpose but the question is “why?”
i know why. im almost positive that cursor is getting special treatment from anthropic, this is not a bad thing necessarily, when 3.7 came out, they instantly had already made cursor work with it perfectly and said they had the model early, also anthropic pulled their models from windsurf for a time which also plays into this theory.
this makes me believe cursor runs their business to prioritize anthropic models to lower their cost and make their investors or whatever happy is my guess, and on purspose, even though they say they dont, i simply dont beleive them, they are nerfing other models
i have copy and pasted code from outside cursor, then used identical prompts inside cursor and gotten worse results. this in my testing is direct proof that cursor is doing this. the only models this doesnt happen for is (shocker) anthropic
now i have absolutely no issue with partnerships, in fact i think it made cursor amazing when we had 500 request and unlimited slowmode, that was peak cursor in my mind in some ways i want to go back to that. but as soon as 3.7 came out at the SAME api cost, all of the sudden everything got changed for no reason.
there is no reason in my mind that they couldnt have kept cursor identical to how it was with the new models, made claude thinking more request per use, AND THIS IS KEY, just make us use more request for the longer our conversation is so that you keep the pricing the same, push users to have shorter conversations, and let all users have full access to the full capabilites of the new models.
this is my fix, by going back to request based pricing, and context based request multiplier, then revert the pricing changes and bring back unlimited slow mode and make auto mode better with new a much cheaper models, remove max mode entirely, that is the biggest scam ever.
now the reason they wont do this despite it being upvoted so much is simple. they simply cant. their business partnerships, need to make more money is keeping them from using a simpler, cheaper solution which would absolutely work.
why let users use good models for free instead of having your rich users pay for tons usage based pricing and just rack up the cash that way? you make no money giving good models for free, but you would make an amazing product, which is what made cursor blow up in the first place.
i have more ideas, but frankly im sick of posting things, then seeing my post are the only ones the devs ignore, its actually so hilarious i can see all the latest post and mine are the only ones skipped over, i dont care, i want to make cursor better, but if you want to make your product ■■■■ and ignore me and everyone else just go ahead im done trying to change your mind.
