Surviving Without GitHub Copilot: My Experience with Gemini Code Assist, Amazon Q Developer, and Cursor
I tested Gemini Code Assist, Amazon Q Developer, and Cursor in VS Code. Here’s what worked, what didn’t, and why I still miss Copilot.
Why I Decided to Try Alternatives to GitHub Copilot
After spending years with GitHub Copilot integrated deeply into my workflow, I decided to try other alternatives after missing my $10/m payment. Was Copilot really as irreplaceable as it felt, or was I missing out on more specialized tools? With that question in mind, I set out to test three popular AI code assistants: Gemini Code Assist, Amazon Q Developer, and Cursor (technically, cursor is an AI IDE, I know!)
My main goal? To see if any of them could rival Copilot in terms of speed, integration, and overall ease of use, particularly within VS Code. Here’s what I learned in the process—and why, after a few weeks, I’m ready to come back to my “home” with GitHub Copilot.
FYI: If you are currently a student, you can get Github Copilot for free using the Github Student Developer pack.
Gemini Code Assist: Ok, but nah
At first glance, Gemini Code Assist seemed like a promising alternative. It offers some of the same basic functionality as Copilot—auto-suggestions, completions, and code assistance. But after a few days of use, my frustration began to mount.
Sluggish Code Suggestions: The most glaring issue with Gemini was how slow the auto-suggestions were. I found myself constantly wondering: “Is it even enabled?” That’s a far cry from the seamless, always-present nature of Github Copilot’s suggestions.
No Jupyter Notebook Support: The lack of support for completions when writing Jupyter Notebooks in VsCode was a dealbreaker for me. As Machine Learning Engineer ,the ability to autocomplete in Jupyter is essential for me.
To be fair, you can still get coding suggestions and debugging assistance from Gemini when using Google Colab which I think is a nice touch.
Amazon Q Developer: A Step in the right direction?
Amazon Q Developer was next up, and while I appreciated its potential, it suffered from similar issues as Gemini.
Sporadic Completions: Just like Gemini, I noticed that Q Developer’s auto-suggestions were noticeably slower compared to Copilot. The delays were especially pronounced when working in non-code files like Markdown or plain text.
VsCode Integration: Although Amazon Q Developer tries to bring some Copilot-like features into the fold, they just don’t feel as fluid. It lacks the tight integration with VS Code that makes Copilot so easy to use, where I can get quick suggestions or automatically generated commit messages. The same could also be said for Gemini Code Assist.
On a few occasions I did get some fairly good suggestions from Amazon Q developer when working with Typescript and Python source files, and the Chat feature was mostly helpful to generate boilerplate code.
Cursor: Close (very), But With Some Frustrations
Cursor was a different beast altogether. It’s a fork of VS Code, so at first glance, it felt familiar. However, once I started using it more and more, a few key differences began to stand out—some good, some frustrating.
Pricing Confusion: One thing that puzzled me was the need to pay for completions, even when using my own API keys. It’s a relatively minor annoyance, but it made me think twice about committing to Cursor long-term. However, If you're a frequent coder, paying for this service might not seem too costly at around $20 per month compared to pay as you go pricing for Claude Sonnet 3.5 models!.
Fast Tab Completions: On the flip side, Cursor's Custom Tab Complete is fast and readily available, sometimes too fast. It felt a little jumpy at times, making it harder to catch whether I wanted the suggestion or not.
Keyboard Shortcut(s): The first thing that drove me up the wall was the fact that CTRL+L—which I normally use to clear the terminal—opens up the chat window instead. This may sound like a small inconvenience, but when you're working in the flow, having to adjust these small muscle-memory shortcuts disrupts your productivity.
That said, Cursor does shine in a lot of areas, especially when paired with Claude (Sonnet 3.5). I loved being able to ground the AI code chat feature in specific folders and files, which allows for more context-aware code generation. Plus, the ability to seamlessly switch between models was a nice touch. For example, I could choose Gemini Flash for generating test data, GPT-4o for feature planning and research, and Claude for coding tasks.
Cursor also stands out because of its ability to add external knowledge bases to Claude, pulling documentation from frameworks I’m using in my codebase. For new projects that rely on the latest coding libraries, this feature felt invaluable!
What I Miss the Most About GitHub Copilot
Ultimately, the biggest thing I missed during this experiment was GitHub Copilot’s tight integration with VS Code. Copilot is always there when you need it—generating commit messages, fixing trivial issues, and doing so without throwing pop-ups that break your concentration. Other IDEs, like Cursor, tend to push notifications or open windows that take some getting used to. With Copilot, these interactions feel more natural and smooth, in my opinion.
Conclusion: How Long Until GitHub Copilot is Officially Dethroned?
After two to three weeks of checking out Gemini Code Assist, Amazon Q Developer, and Cursor, I have to say that GitHub Copilot is still the gold standard for AI code assistants. The alternatives show potential—Cursor, in particular, is brilliant when paired with the right models—but they still fall short in terms of smooth integration with VsCode and Github(Yeah, I know…).
For now, I’ll be heading back to Github Copilot, but I suspect it won’t be long until I switch to a native AI-driven coding tools like Cursor, Tabine, Codium and many others which are still on the works. Who knows, maybe even Devin might take the trophy in the long run?
What’s Your Favorite Code Assistant?
Have you tried any alternatives to GitHub Copilot? What’s your experience been like?