At my previous job at Adobe, we were all given GitHub Copilot licenses. Management strongly encouraged its use—sometimes even calling out engineers who weren’t actively using it. Adoption across the team was mixed: some engineers never touched it, while others seemed almost unable to code without it.
Personally, I found it helpful when writing brand new code or classes. But when editing existing code, or working on features that touched multiple parts of the repo (which was common for me), Copilot didn’t add much value. It often felt like more noise than signal in those cases.
I do believe AI tools have a place in every strong engineer’s toolkit, but I’m still trying to figure out what that place is for me. I’ve also heard stories of people becoming so dependent on tools like Copilot that they start to lose touch with core programming skills—which makes me wonder:
How do you think tools like Copilot can be used to improve your career, without hurting your long-term growth as a software engineer? Have you found ways to integrate it that genuinely elevate your work—or does it feel like a crutch?
Curious to hear how others are thinking about this.
The goto these days is windsurf. I'm getting a much higher code output from windsurf than cursor and works especially well on large codebases
First, try Cursor if you're a VSCode developer. What you mentioned about Copilot matches what I've heard. It can be helpful-ish here and there, but it struggles with a bunch of stuff. Cursor is a step function more powerful, lending far more assistance with higher accuracy. AI is woven in everywhere as they forked VSCode and rebuilt it from the ground up with AI in mind.
As for using AI to bolster your workflow, it's tricky. AI has a ton of nuance in how it should be applied, and I think a lot of an engineer's value nowadays is knowing when vs. when not to use it. I use ChatGPT a lot to expedite writing code I already mostly know how to write. However, I've learned that it's pretty terrible for any feature that's more niche with higher technical complexity and it's particularly terrible at debugging tough issues.
Check out this thread too: "Which AI Agent or Code Editor Best Understands Code Context? (Comparison for Small, Mid-size, and Large Codebases)"