Hello,
Cursor AI agent is often asking if the changes he is doing are working.
example from Cursor agent : “Can you refresh the page and let me know if it’w working now ?”
If you reply just “yes, it’s working” Cursor count one request from your account.
That’s really not fair because this is not a new request, the AI is asking us to confirm his changes.
I really don’t like this. @Cursor, please react on this or you will lose customers, including me
The issue here is more how the user (mis)understands and uses the output of AI.
To make it clear: AI models are trained to provide fluent conversational communication. Additionally over last 1.5 years the AI providers have learned form usage that certain types of questions or acknowledgments are polite and guide the user to check if things work since the AI didnt have until now access to verify that things are working. (this is changing with the MCP integration)
If the AI has made changes and they work, nobody forces the user to have to reply to the AI that it worked. Its clear that every submit of the chat/composer is a new request, anything else is just imagination or ignorance.
Instead user could give the AI the next task or next request to process. (and you can still in the new task say, thanks it works, next required is:…)
Its really better to learn how to AI usage works instead of bashing here the makers of Cursor.
Agree and not agree… first yes maybe i should learn more about AI, but that’s not a reason for a system to not understand the difference between a simple confirmation (AI is requesting) and a new request. If a system that is invoicing based on request is not able to do that, i still think it’s not sufficiently clever (aren’t we talking about Intelligence is the term AI) and it’s not fair from a Company who is doing this pricing model.
Third the goal of AI is to improve itself and it’s even asked by Cursor if you want your datas to be shared for improvement, so if you never reply to AI confirming or not if the modification are working or not, how can AI improve ! So thanks for the lessons, but it’s just your point of view
yes at least it should be mentionned in the rules : “Do not reply any question from AI to improve it or you will be charged” … that’s a question of interpretation