In December 2024, I wrote an article sharing my experience using VSCode (GitHub Copilot) and Cursor (Claude 3.5 Sonnet) from the angle of a Data Scientist.
I concluded the article by stating:
Since writing this text, not only have there been significant developments to VSCode and Cursor, but a brand new wave of AI-powered IDEs (e.g. Windsurf) have emerged. Upon evaluating and testing these alternative IDEs, I even have decided to revert to VSCode as my primary IDE of selection.
In this text, I’ll discuss why I’m now not using Cursor, why I even have reverted to VSCode and GitHub Copilot, and my opinion on what IDE you have to be using as a Data Scientist.
What I Liked Most About Cursor
Let me state this primary: Cursor is an incredible product. I kept hearing developers discuss Cursor increasingly more, with many making the most of their 2-week free trial. I used to be hesitant at first, but after reading persistent praise for the software, I made a decision to form my very own opinion firsthand.
I loved that Cursor’s interface was the identical as VSCode. Upon installation, Cursor enabled you to download all of your existing VSCode extensions, which definitely helped with adoption because it made you are feeling right at home.
In December 2024, two key Cursor features made it stand out from the group. The primary was the flexibility to make use of a variety of different LLMs, unlike GitHub Copilot, which had restrictions on what LLMs were available. Secondly, Cursor had a feature called “Composer” that allowed users to generate a complete project code base by simply describing it via a prompt.
As described in my article “Must you switch from VSCode to Cursor?”, these were the 2 features I used to be most enthusiastic about and based my initial review on.
Why I Switched Back to VSCode
VSCode has been my primary IDE for over 7 years, I used to be also a beta tester for GitHub Copilot, something I even have been paying for since its release.
As previously explained, a key advantage of Cursor is its similarity to VSCode from an interface perspective. Although a minor reason, switching back to VSCode meant I didn’t have to refamiliarise myself with another UI.
GitHub Copilot eliminated considered one of the 2 primary reasons for my initial move to Cursor when it enabled the flexibility to implement SOTA newly-released LLMs. As of writing in April 2025, below is the prevailing list of LLMs available via GitHub Copilot:
Claude 3.7 Sonnet was released on February twenty fourth, 2025. On the identical day, Microsoft announced that this model was now available via GitHub Copilot. GPT-4.5 was released on February twenty seventh, 2025, and was also made available on the identical day in GitHub Copilot Chat to Copilot Enterprise users.
An enormous improvement on their LLM offering in comparison to back in December 2024.
As a Data Scientist, I often work in Jupyter Notebooks for exploration tasks. Although AI-assistance works higher when Programming using .py files, I find VSCode is healthier than Cursor when working with Jupyter Notebooks.
Vscode offers GitHub Copilot instructions when adding a brand new cell to your notebook, this shouldn’t be available using Cursor. Although AI-assistance is accessible, it’s tougher because the user has to make use of Ctrl/⌘ + K to start writing their prompt.
When working with Jupyter Notebooks using VSCode and GitHub Copilot, you furthermore may have the flexibility to speak along with your AI-assistant, referencing the cells in your notebook you wish the AI to give attention to. This functionality makes it incredibly easy to implement AI-assistance in your coding workflow.
The ultimate two aspects, which were less essential but definitely played a task in my overall decision, were cost and usage in my skilled role.
GitHub Copilot costs $10 per 30 days in comparison with Cursor, which costs $20 per 30 days. I asked myself, am I getting twice the worth from using Cursor over VSCode? The reply to this query was no, due to this fact, this added more weight towards reverting to VSCode.
Finally, in my skilled role as a Lead Data Scientist, I exploit VSCode each day. I might never select an IDE based solely on the one I exploit in my skilled role, even though it definitely helps to maintain things aligned where possible.
How GitHub Copilot Closed the Gap
GitHub was purchased for $7.5 billion by Microsoft in 2018. As Microsoft is a far larger company than Anysphere Inc., they’re in a position to place so much more resources on improving their product offering.
As an end user, competition is inspired because it pushes corporations to innovate and develop at a pace that permits them to maintain up with one another. This also signifies that when a brand new feature is released and is widely adopted by the community, all the competitor corporations rush to develop that very same feature into their very own product offering.
When Cursor was first gaining popularity, it had many features (e.g., Composer, Tab, and Cursor Prediction) that its competitors didn’t have. This isn’t any longer the case.
As shown in GitHub Copilot’s changelog, releases are occurring weekly, with some days consisting of multiple recent feature being made available to all GitHub Copilot users.
As of today, there aren’t many features that Cursor offers that will not be also available using VSCode and GitHub Copilot. Microsoft has shown up to now 4 months that it’s placing huge importance on making GitHub Copilot the perfect AI coding assistant in the marketplace. While some improvements are still needed, they’re moving at a rapid pace, closing the gap, and are ready to steer the market by the top of 2025.
Final Thoughts
Firstly, I don’t imagine any text editor/IDE is healthier than another. You need to at all times make this judgment yourself based on trial-and-error and never finalise your decision based on what’s most spoken about online.
I even have been using VSCode for several weeks now since reverting from Cursor, and I even have not regretted my decision. I imagine Microsoft is making great progress with GitHub Copilot and has really closed the gap to all the brand new AI-assistant IDEs.
There’ll at all times be features that pop up here and there that will not be available within the IDE you’re currently using. In the event that they’re widely adopted by the community, all competitors will implement them with time. Don’t use this as a chance to maintain switching your IDE.
