Why GitHub Copilot Is Overrated: A Personal Perspective

As someone who’s been building software for years, I’ve had my fair share of encounters with AI coding tools. GitHub Copilot, touted as the future of coding, often comes up in conversations among developers. But let me be clear: in my experience, it’s overrated. Here’s why.

The Hype Versus Reality

When Copilot was first introduced, the excitement was palpable. It promised to significantly reduce coding time by offering suggestions and auto-completions. But as I started using it in real-world projects, I quickly realized that the reality didn’t match the hype. Too often, the suggestions were either irrelevant or required more adjustments than simply writing the code myself.

What GitHub Copilot Actually Does

GitHub Copilot generates code suggestions based on the context of your current project. It utilizes machine learning trained on public code repositories to propose snippets, functions, or even entire classes.

• Pricing: $10/month per user or $100/year per user.
• Best for: Quick code completions for experienced developers.
• Limitations: Struggles with complex logic, doesn’t always understand project-specific context, and may suggest insecure code.
• Our Take: We tried it for a few side projects but found ourselves correcting its suggestions more often than not.

Alternatives to GitHub Copilot

If you’re considering alternatives, there are several tools out there that may suit your needs better. Here’s a breakdown of some options:

| Tool | Pricing | Best For | Limitations | Our Verdict | |--------------------|-----------------------------------|------------------------------------|---------------------------------------|----------------------------------| | Tabnine | Free tier + $12/mo pro | JavaScript and Python developers | Limited languages in free tier | We use Tabnine for quick completions. | | Codeium | Free | General coding assistance | Lacks advanced context understanding | We don’t use it due to limited context. | | Sourcery | Free for open-source + $12/mo pro| Python developers | Limited to Python only | We use it for Python projects. | | Replit | Free tier + $20/mo pro | Collaborative coding | Pricing can get high for teams | We use Replit for quick prototyping. | | AWS CodeWhisperer | $19/month per user | AWS-centric development | Best for AWS services only | We don’t use it as we’re not AWS-centric. | | Codex | $0-20/month depending on usage | General coding across languages | May not handle niche libraries well | We don’t use it due to pricing. | | KITE | Free + $16.60/mo pro | JavaScript and Python | Lacks multi-language support | We stopped using it - it was buggy. | | IntelliCode | Free with Visual Studio | C# and .NET development | Limited to Microsoft ecosystem | We use it for C# projects. | | Jupyter Notebook | Free | Data science and analysis | Not for general software development | We use it for data analysis. | | Codex by OpenAI | $0-20/month depending on usage | Versatile coding assistance | Can be expensive for high usage | We don’t use it as it can be unpredictable. |

What Doesn’t Work with GitHub Copilot

1. Context Misunderstanding: Copilot often fails to grasp the unique context of your project. For example, if you’re working on a specific domain or using a niche library, it may provide generic suggestions that don’t fit well.

2. Security Concerns: There’s also the issue of security. Copilot can suggest code snippets that contain vulnerabilities, which can lead to serious issues down the line. As a solo founder or indie hacker, you can’t afford to overlook this.

3. Dependency on Internet: Copilot requires an active internet connection, which can be a hassle if you’re working offline or in a low-bandwidth area.

When to Use GitHub Copilot

Despite its limitations, there are scenarios where Copilot can provide value:

• If you’re an experienced developer who needs quick snippets for common tasks.
• When you’re working on projects that involve standard libraries or frameworks where Copilot has a wealth of training data.
• If you’re okay with spending time reviewing and adjusting suggestions.

Conclusion: Should You Use GitHub Copilot?

In my experience, GitHub Copilot can be useful in specific contexts, but it’s not the universal solution it’s marketed as. If you’re looking for a reliable coding assistant, I’d recommend trying out alternatives first.

Start with tools like Tabnine or Sourcery, especially if you're coding in JavaScript or Python. They may provide a more tailored experience without the drawbacks of Copilot.