Why GitHub Copilot is Overrated: 5 Key Flaws

As indie hackers and solo founders, we often look for tools that can supercharge our productivity. GitHub Copilot, an AI-powered code completion tool, promises to do just that. However, after spending considerable time with it, I believe it’s overrated. Here are five key flaws that make it less of a game-changer than many claim.

1. Limited Language Support

Copilot excels at languages like JavaScript and Python, but if you're working with less popular languages, you're in for a frustrating experience. The AI has been trained primarily on code from GitHub repositories, meaning niche languages or frameworks get far less attention.

• Pricing: $10/month for individuals
• Best for: Developers working primarily in mainstream languages
• Limitations: Poor support for niche languages like Elixir or Haskell
• Our take: We use Copilot for JavaScript projects, but it falls flat when we switch to Rust.

2. Contextual Understanding is Lacking

While Copilot can generate snippets based on comments or partial code, it often misses the bigger picture. It struggles to grasp the full context of a project, leading to irrelevant suggestions or code that doesn’t fit well with existing structures.

• Pricing: $10/month for individuals
• Best for: Quick code snippets, not complete solutions
• Limitations: Limited contextual awareness can lead to wasted time
• Our take: I've often had to rewrite generated code because it didn’t fit the overall architecture.

3. Code Quality is Inconsistent

Not all code generated by Copilot is good code. Sometimes it suggests outdated practices or security vulnerabilities. Relying solely on it can introduce technical debt and potential security risks into your projects.

• Pricing: $10/month for individuals
• Best for: Initial drafts or brainstorming code ideas
• Limitations: Requires manual review for quality assurance
• Our take: We often use it as a brainstorming tool, but we’ve learned to double-check everything it produces.

4. Licensing and Attribution Confusion

One of the more contentious issues surrounding Copilot is its licensing. The AI is trained on public repositories, and there’s ongoing debate about the legality of using generated code. This can put you at risk, especially for commercial projects.

• Pricing: $10/month for individuals
• Best for: Personal projects where licensing isn’t a concern
• Limitations: Potential legal issues for commercial use
• Our take: We avoid using Copilot-generated code in client projects to steer clear of any legal complications.

5. Over-Reliance Can Stunt Learning

For new developers, there’s a risk that relying too heavily on Copilot can hinder their coding skills. Instead of learning how to solve problems independently, they might default to asking the AI for help.

• Pricing: $10/month for individuals
• Best for: Experienced developers who want to speed up coding
• Limitations: Can detract from the learning experience for beginners
• Our take: We encourage junior developers on our team to use Copilot sparingly and focus on building their skills.

Comparison Table: GitHub Copilot vs. Alternatives

| Tool | Pricing | Best For | Limitations | Our Verdict | |------------------|----------------------|---------------------------------------|-------------------------------------|-----------------------------------| | GitHub Copilot | $10/month | JavaScript, Python developers | Poor niche language support | Use for drafts, not final code | | Tabnine | Free tier + $12/mo | Multi-language support | Limited free features | Better for diverse languages | | Kite | Free, Pro at $19.90 | Python developers | Limited to Python | Good for Python, but not much else| | Codeium | Free | Quick code suggestions | Limited context understanding | Great for quick drafts | | Sourcery | Free, Pro at $12/mo | Python refactoring | Focuses solely on Python | Use for improving existing code |

What We Actually Use

In our stack, we primarily rely on Tabnine for multi-language support and Sourcery for Python refactoring. GitHub Copilot finds its place in our workflow mainly for brainstorming and generating initial drafts, but we’re careful to review everything it generates.

Conclusion

In summary, while GitHub Copilot has its uses, it’s essential to recognize its limitations. It’s not a silver bullet for coding, and over-reliance can lead to issues. If you’re considering using it, start with the understanding that it’s a tool to assist, not replace, your coding skills.

Start here: If you’re looking for a solid AI coding tool, try Tabnine for broader language support and better contextual understanding.