Hi all,
We could use some help from cursor dev team and others to start a discussion on breakdown in costs for background agents. As a look at the usage, billing and other dashboards on cursor account settings, I don’t have a good sense for what to do to help bring costs down. I spent about 40$ in an hour, I am a python developer for open source machine learning for biology (https://deepforest.readthedocs.io/).
- Choosing the model, there are several recommended models, but their price difference is a bit opaque to me.
- Is it better to have more short simple instructions, versus a single very long detailed instruction?
- Is pulling in data (dropbox, github, S3) expensive compared to agent instructions, I assume not.
- Is there a ‘difficulty’ cost, or is it really just how long the agent has to think to generate tasks.
- I assume not, but confirming that letting an agent idle for a medium period of time (an hour) doesn’t cost much, again its the instructions that drive the price.