AI Codes Faster, Cleaner, and Smarter

I use AI in my workflow. I use it each and every day and it makes certain aspects of my coding life easier.

Some of my colleagues refuse to do the same. The code they produce is littered with spelling errors, logical fallacies and is poorly tested so why aren’t they using tools that can help them out?

Here are a few aspects where AI is simply better than any human software engineer, and I think software engineers should use these modern tools as code quality is important.

Here is How AI Codes Better Than You

Speed and Efficiency

You might expect me to say that AI doesn’t need coffee breaks or ramen in order to churn out code, but the impact of AI is much more than that.

For example, one of many ways great ways to use AI tools is as a pair programmer (probably why Microsoft’s tool is called Copilot after all), and it can pick out those little mistakes and issues with your code.

https://five.co/blog/ai-pair-programmers-impact-on-developer-productivity/

If you compare performance pairing using AI with Stack Overflow, AI supercharges coding performance. Evidence? It’s here:

“The group [of developers] that has access to GitHub Copilot was able to complete the task 55.8% faster than the control group [without access to GitHub Copilot].” Source

Removal of Mistaeks

Imagine a world without spellcheckers. Your writing would be full of tyops and little annoyances. Tools can help you to produce better quality code missing out those little slips and mistakes that can edge in even the most considered of code.

You might know a developer whose spaghetti code is riddled with small mistakes. I’ve known a few of those, although I have to say eradicating spelling mistakes is not too important to me but can signal a poorly crafted PR. I’d love to see these silly mistakes eradicated so I could focus on the overall quality of the PR and what it intends to achieve.

Don’t take my word for it, Metabob (an AI tool) dramatically outperforms traditional static code analysis tools like Sonarqube and linters.

https://www.eejournal.com/article/using-generative-ai-for-refactoring-and-debugging-code-cut-debugging-time-in-half/

So, use AI, check it’s results, and you can avoid making trivial mistakes and looking silly during code review.

Superior Documentation

I don’t know about your company, but mine has a real problem regarding documentation.

The problem is that nobody likes writing documentation. The result of this is that much of our code is undocumented in both functionality and behavior. This means it is almost impossible to refactor old code — a colleague of mine needed to rewrite a class from 2017 and as we discussed it’s nearly impossible to define the required behavior (because we don’t know what it should do, let alone how the code currently works).

Now imagine some software that will automatically write the documentation that nobody wants to do. If we called that AI you’d be well advised to use it.

Learning and Adapting

Some are worried about their jobs because AI seems to learn the latest coding tricks and techniques. That is true because AI instances can be trained at nights and on weekends with no breaks a human might need.

Multitasking Master

Humans don’t have a problem multitasking simple tasks like breathing and walking, but for complex tasks we are reasonably terrible.

Since AI is about instances, if we want to complete several tasks simultaneously, we can set up multiple instances to complete the work required.

That’s a brilliant way to get work done in a quick and efficient manner and stretches to more than just breathing while you work (that AI doesn’t need to do in any case).

Here is Why no AI Competes With You

Let us take the idea that AI can learn faster than a human. The truth is a human paired with an AI becomes super-human.

In this study students who had access to an AI code for tasks improved their progression, correctness and decreased their errors and task completion time (when compared to those who didn’t). It didn’t even decrease their performance in subsequent manual coding tasks.

That is, using AI learns better than you do but if you use it to help you learn.

That is true for all of the AI advantages above, instead of trying to compete with the AI we should use it to become super coders.

I know most programmers know this as 92% of programmers use AI tools already.

Conclusion

AI is here, and it will help us to become better coders if used correctly.

Isn’t it time we did just that, and become AI-powered coders? I certainly think so, but I’d love to hear what you think in the comments.