Anysphere focuses on automating coding to create an advanced human-AI hybrid programmer. This approach aims to increase programming efficiency, allowing developers to manage codebases effortlessly and iterate quickly even in complex systems. By combining AI and human creativity, Anysphere seeks to exceed the capabilities of pure AI systems. The company comprises researchers and engineers who develop software and models to push the boundaries of what is both useful and possible.
Do You Manage Peer Insights at Anysphere?
Access Vendor Portal to update and manage your profile.
I love how Cursor can use different AI models dependent on the tasks, I use Opus for complex tasks that require good reasoning (more on this in the dislikes!) and composer for everyday tasks. It's a nice system once you understand what model is good at what tasks. I also like how Cursor fits into my engineering workflow. It doesn't feel like a gimmicky chat bot that's been bolted on, it can understand context across files and helps me move faster without losing control of what's happening. Another big win is how good Cursor is at working through problems with me instead of just giving me answers. I can have it challenge me or help me refine my ideas to improve the quality of what I deliver, not just how fast I can deliver.
The variety of models that you can use and quick views on tracking your usage. The agents and sub-agents process makes your work go by super quick with just a few prompts once you have your Skills and Rules setup in Cursor.
One of the standout features of Cursor is its seamless integration with other platforms, such as GitHub and Notion. This connectivity allows for a smoother transition when managing code repositories and creating context-rich documentation in Notion. The ability to automate tasks and generate/write comprehensive documentation directly aligns with my workflow, significantly enhancing productivity and streamlining problem-solving.
I hate how they changed the pricing/token model to enforce MAX mode for the newest AI models, I understand why this changed, but it still stings that I need to burn through our companys tokens to get the high level of work done or spend 5 times as long working with an older free model. Another frustration is that for very large tasks, context can sometimes drift or get lost, which means you have to restate intent or constraints more often than you'd expect. It's not a massive issue and more of a nitpick, but when you're working through a complex, multi-step task it can interrupt the flow. I'd also like some more transparency and control around token usage. It's not always obvious how expensive a given task is going to be until you're already committed, which can cause issues between prioritizing speed or costs within our team.
As a Windows user, it has been a bit frustrating to follow all of the small incremental updates Cursor makes to the tool. Because of a bug/issue with their in-tool updater and crashes on Windows, we have to do a complete uninstall and reinstall whenever there's an update. If they can work on the Windows user experience, that would fuel more adoption within the industry.
While Cursor is a robust tool, there are areas for improvement. First, the learning curve can be steep for those not familiar with similar automation tools. Secondly, while integration with platforms like GitHub and Notion is beneficial, the initial setup can sometimes be cumbersome, requiring a bit of trial and error. Lastly, the user interface, though functional, could benefit from a more modern and intuitive design to improve user experience; it's very code heavy and not intuitive to users not familiar with that field.