two weeks making a rock paper scissors game without any visuals 😭

You're asking the wrong question and making certain assumptions based on inexperience.

If this fella doesn't have the knowledge of the underlying logic, an understanding of the syntax and the experience to handle the spanners that get flung around from even the most basic and clean programs, then he's simply got a program that works in and only in a vacuum. This, while useful for demonstration and his singular, narrow use case, isn't how software is intended to function in commercial and industrial applications.

He cannot debug it, because he does not have the understanding of syntax, DSA, CS, etc.

He cannot integrate it, because he does not understand software, requirements or integration.

He cannot update it, because he doesn't know what it does. He only knows the output.

He cannot optimize it, because he doesn't understand the many, many considerations of performance, which are often unique per use case.

He cannot polish it, because he doesn't know what good code looks like. And even more problematic, he doesn't know what bad code looks like.

Now, if he does have knowledge and expertise in this area; what you're seeing is not the AI doing all the important things (it's not code!), you're seeing a developer leverage an AI assistant to create code he already knows how to write. He can rely on his knowledge and experience to ensure that the program is specced and scoped correctly. Its design is clean, performant, and modular. That it doesn't just meet his use case, but does it in a manner that meets both the criteria of the prompt but also general programming and design principles. He can also ensure it's not doing malicious actions, which is always a potential issue when executing code you didn't write yourself / from untrusted sources.

You're either seeing an AI write a well documented, well understood boilerplate program OR someone with the knowledge and experience to know how to fix the mess the AI built under the hood to meet the use case.

Don't think of AI as a replacement to developers. Think of it as an assistant to your coding. It supports your workflow by automation of the tedious elements; repeated boilerplate, propagating a deep change across a legacy codebase, etc.

You're still young in your career, so it may seem like code is your main job. But as you rise through the ranks, you'll see code is more similar to paperwork. It's important, you have to get it right and you need a skill set that can get it done. But design, analysis, meetings (so many meetings...), stack knowledge, etc are even more important. If it was all code, no developer would rise above mid-level. But seniority is all about knowledge and experience and using that to get the job done - and we express that through code, just like a writer uses natural language.