Well I cancelled my subscription, I will explain quickly my reasons.
AS a team user, I would believe this fast/slow request thing when I started was honest about the models we wanted.
Then sonnet 3.7 launched, and I started burning over my limit. This model supposedly available for slow was unavailable 95% of the time
The economics:
- before change : USD 300M ARR (April 2025) - profit est. USD 100 - 200M.
- After change : made at the time of company valuation at 9.9 USD in june, a very strange coincidence isn’t it. USD 500M ARR (60% increase in a month), raised USD 900M, profit est. 400M USD.
The first rug pull was not only vicious - Cursor never lost money unlike OpenAI, etc., running massive compute, they route and cripple, they benefited in a massive way from massive API reduction costs (est 30%). And you pro user will believe your crippled model really saved you X money at fake prices.
I kept quiet as enterprise user, but unfortunately, you now rug pull the guys that run numbers and know when you rug pull them. The reason of ending this auto is to remove a liability too much users were understanding and using to work around these limitations.
At EUR 40 I was profitable at ~40%. Not enough, after the change the number will quadruple and for that, I will not let that happen.
I encourage team and enterprise to change : you turn worst than everyone else : you get 20 USD of fake API cost and pay double ! You pay for the few power users and if you refuse this, look alternatives. Cursor is not even decent, it’s on par with the competition now. Avoid windsurf obviously.
Being cash-cowed by most makes them complain, but they stay. It is insightful on the ‘leaving ship’ psychology trick. The fear, the time to switch, the seemingly lacking alternatives - lol even docker proposes MCP guys if you don’t find options you are the problem.
Just gave you the figures, I’m out. Thanks for everything, I’ll come back to explain why even the most hated microsoft ends up being better for me, I never thought I’d come down to this.