AI-Assisted Coding: A Game Changer in My Python Journey & our lives

※All of this post is translated with GPT-4o ,（though I wrote all of the original senteces in Japanese,）, which makes me surprised how smart the current AI is.

Lately, I've been using generative AI, and I'm beginning to believe the sentiment expressed in the title of this post.

I'm at a level where I have a rough understanding of grammar and can barely write detailed code on my own. However, by leveraging GPT-4o, I've been able to code what I want without much hassle.

There are several advantages to having AI generate code for you.

First, it allows for uninterrupted thinking. In the days when AI couldn't be used for work, I would have to constantly look up syntax, which took time. It also took even more time to apply the correct syntax to specific contexts or combine multiple syntaxes. Since I started using GPT-4o, the more specific I am about the situation (such as what I want to do, the column names of my data, folder hierarchy, constraints, etc.), the more accurately it generates the code. Even if the code doesn't run smoothly, I can throw back the errors, and it quickly corrects them. While the old way might have resulted in stronger coding skills in the long run, I feel that using AI allows for a more immediate and joyful application of coding skills, which then creates a positive feedback loop for further use.

Secondly, it surprisingly also leads to learning. Depending on how you prompt it, you can add conditions like "explain the syntax used," and it will provide explanations after the full code. This allows for a faster coding cycle while also speeding up the learning process.

Listing these benefits, one might argue that with generative AI, skills and specialized knowledge might become unnecessary. However, I don't necessarily think that's the case. Currently, generative AI is difficult for non-engineers to operate as a system agent, so the advantage of being able to use programming languages still remains. Additionally, even if AI agents become mainstream in the future, having an understanding of the internal processing mechanisms will still be valuable, though this is more of an intuitive feeling.

In my recent Python coding experiences with generative AI, I've noticed that having what you might call a "framework" of knowledge significantly enhances the use of generative AI. This could be knowledge about what can be done with Python or how to set up the environment, or it could be knowing how to effectively use the generative AI itself (like realizing that code formatted in a certain way is easier to copy and paste). Having such frameworks in various fields seems essential, at least in the current era of generative AI. However, it will be interesting to see how this changes as AI agents become more common and start proposing what you want to do based on your system operation logs.