Why GitHub Copilot is Overrated: 5 Things Most Developers Miss

As developers, we’re always on the lookout for tools that can save us time and make our coding lives easier. Enter GitHub Copilot: the AI-powered coding assistant that promises to revolutionize how we write code. But after using Copilot extensively, I’ve come to a contrarian conclusion: it’s overrated. Here's what many developers miss when they rave about it.

1. It’s Not a Replacement for Understanding Code

GitHub Copilot can suggest lines of code based on context, but it doesn't actually understand the logic behind what you're building. This can lead to poor coding practices if you rely on it too heavily.

• Limitation: It can generate code snippets that may not work as intended without proper context.
• Our take: We use Copilot as a supplementary tool, but we always double-check its suggestions against our own understanding of the codebase.

2. Pricing Can Add Up Quickly

GitHub Copilot offers a free trial, but after that, it’s priced at $10/month per user. For small teams or solo developers, this can be a significant expense, especially when you consider the alternatives.

| Tool | Pricing | Best For | Limitations | Our Verdict | |---------------|------------------------|-----------------------------|-------------------------------------------------|--------------------------------------| | GitHub Copilot| $10/mo | Code suggestions | Can provide wrong context or suggestions | Use as a backup, not a main tool | | TabNine | Free tier + $12/mo pro| AI code completion | Limited language support for the free tier | Good alternative for multi-language | | Kite | Free | Python and JavaScript | Limited to specific languages | Use it for Python projects | | Sourcery | Free + $12/mo pro | Python code improvement | Focused only on Python | Best for Python code reviews | | Codeium | Free | General code assistance | May lack depth in complex scenarios | Great for quick suggestions |

3. It Can Introduce Bugs

While Copilot can help you write code faster, it can also introduce bugs without you realizing it. For instance, it might suggest a function that looks correct but has subtle flaws.

• Limitation: You must be vigilant to catch these mistakes, which can slow you down rather than speed you up.
• Our take: We’ve had to spend extra time debugging code that Copilot suggested, which defeats the purpose of using it for efficiency.

4. It Doesn’t Understand Your Project’s Context

Copilot is trained on a vast range of public code but lacks the specific context of your project. Its suggestions can be generic and not tailored to your unique requirements.

• Limitation: This can lead to irrelevant or inefficient code suggestions.
• Our take: We've found it more useful for boilerplate code rather than intricate logic specific to our applications.

5. It Can Foster Bad Habits

If you rely on Copilot too much, you risk developing a habit of not thinking critically about your code. This could hinder your growth as a developer over time.

• Limitation: It’s easy to become complacent and let AI do the heavy lifting.
• Our take: We encourage our team to use Copilot as a brainstorming partner rather than a crutch, ensuring they remain engaged in the coding process.

Conclusion: Start Here

If you're considering GitHub Copilot, I recommend approaching it with caution. It can be a helpful tool for speeding up certain tasks, but rely on it too much, and you might find yourself facing more issues than benefits.

Instead, consider starting with a combination of tools that enhance your coding without taking away your critical thinking. For instance, TabNine or Kite can be great alternatives depending on your specific needs.

What We Actually Use

In our stack, we primarily rely on TabNine for general code suggestions and Sourcery for Python projects. They provide a balance of AI assistance without compromising our understanding of the code.