Ok it looks like the value proposition of Cursor is breaking down.
The application has become so slow it is essentially unuseable on the slow pool side. So what’s the point then if all is built on VS Code and the premium models are owned by third parties?
I think it means that the value of Cursor is 90% dependent on the premium model providers who can control their API pricing vis-a-vis them and essentially have them by the ****.
In the old days I paid 20 bucks for great premium model use. Now it seems slow pool is completely being squeezed due to:
- Advent of token-heavy thinking models from closed source providers
- Growth
- Probably lack of better usage contracts with Anthropic and OpenAI
It seems like the other big player got a multi bn offer from OpenAI and the valuation of Cursor essentially depends on its ability to provide premium model access on an unlimited basis for cheap. Ie the value is not with Cursor per se, but with Anthropic and OpenAI and Google who hold the keys to the 20m a month premium access.
Right now every single slow request takes ages and the app is unuseable. So I pay 20 bucks for 1-2 days of use and then it becomes useless and the model providers hold the keys, ie the wrapper is a commodity.
I hope Anthropic buys Cursor and makes an unlimited fast plan for 60 bucks or so. That’s the sweet spot.
Right now Cursor is 20 bucks for 2 days and then for a month 500 bucks - that’s not attractive nor sustainable.
Only hope at this stage is coming from Deepseek to release R2 for sub USD per million tokens and run it via API in another wrapper and Cursor developing some proprietary tech (e.g. on the context side).
I don’t think a 500 USD/m SaaS is competitive. The moment anthropic drops is own coding IDE with unlimited use for 60-100/m usage will fall dramatically.
I loved Cursor so far, but I am now looking at the other provider to see if it is more workable for a fair price.