I've seen many jokes about using tools like GPT/Copilot for writing code. Some people say:
- The code it writes isn't good.
- It's just like those no-code/low-code tools.
- Good luck fixing the code it produces.
Many of these discussions feel one-sided, so I thought of introducing a different perspective to keep things lively.
1. Who's in Charge?
Even with AI chipping in, there's always a developer in the mix. Hence, it's called a 'copilot'. We're the main drivers.
Tools like ChatGPT, GitHub Copilot, and Cursor are tools. How effectively we use them is up to us. If you use them properly, they can assist in crafting impressive code. Here is one great example of how GTP can be used to generate self-tested code
And hey, these tools are still new. I bet they'll only get better.
Think about it: We get better by reading code (good and bad). These tools can read all the possible open-source codebases out there 🤯
2. AI-generated Code vs. No-code/Low-code:
They're distinct! AI-generated code is superior. Here's why:
Flexibility: AI crafts code tailored to your specifications. No-code/low-code tools? They come with their set boundaries.
Learning Together: With AI, the developer and the tool grow and learn together. No-code platforms? Not quite.
Diverse Applications: No-code platforms often cater to specific areas, like apps. AI copilots can be used in much broader contexts and codebases.
Ownership: You own the code generated by AI tools. Not sure about no-code/low-code platforms.
3. Maintenance Woes:
Maintaining legacy codebases is a challenge. But hasn't that been a consistent issue? When was the last time you stumbled upon an old codebase and were impressed by its quality?
We're not the best at writing maintainable code. It is hard. TDD and other code quality practices help, but only some people can do that well. Codebases tend to degrade over time, leading to technical debt.
I think AI will make our situation better than making it worse. It will raise the average quality of the codebase. Make it easier to write/generate good code and also help us maintain it.
If AI can read 2000 years old script, it can definitely read our codebase and give us insights. I imagine a future where we will talk to our code, ask questions, and receive guidance on troubleshooting issues.
By the way, I had some help from ChatGPT writing this!
Top comments (0)