Gosh, what a crazy time we live in.

The tech industry, and especially the gaming one, is scrambling to use AI for everything. It’s on your computer, it’s in your phone, on your fridge, in your coffee maker, and now we’re seeing pins, glasses, and more to attach to us to make humans augmented with artificial intelligence. It’s definitely a bubble, but more than that, it’s also an impressive technology that will never go away. I should say, though, AI is a rapidly scaling technology, and how it is right now is the worst it will ever be.

AI is being used everywhere in big gaming companies, but in a way that misunderstands what it’s best for or where it belongs… and I’m here to talk about what it’s for. Ethics aside, there are practical reasons to push the brakes on AI usage in game development.

What AI is, and what it isn’t

Let’s start with some facts here, since it’s insane how many people don’t know them.

First, AI is not a knowledge base like Google or Wikipedia. Instead, it’s a machine model that takes an input (prompt) and spits out an output (content, like more text, code, art, etc). The output is not based on factual information, rather it’s just trying to make it from what it thinks you want from the input. It can be, and will be, wrong… a lot.

AI is also not a senior engineer, nor is it up to par with one. Because of the mistakes today’s models make, and what their limitations are, AI is more like a junior engineer… except one that never learns as you tell it new information.

Those two being said, AI is great for specific problems. I’ve seen great use cases in my workflows for things like math, boilerplate, and reptitive code.

When I use AI

In most cases, I write my own code. I might use AI autofill features when doing mundane tasks, like documentation, but I leave any problem-solving to me. Again, this is because AI will never have a full understanding of what’s happening, and it makes a lot of mistakes. Historically, when I’ve tried to use AI, I spent more time fixing it compared to if I just wrote it myself.

AI also isn’t the greatest at solving problems in your errors. For example, days before writing this post, I had an issue with some Sudoku code I was writing, specifically with the algorithm to generate the answer key. I had existing code, but I needed a specific issue fixed with how numbers were arranged. I tasked AI to do this in the form of an agent, which could see everything and have the full context. It kept rewriting the function over and over to “fix” something that, turns out, wasn’t even an issue… I just messed up something elsewhere. Instead of AI fixing that other problem, it got stuck on one script that it swore was the root of the issue.

At a higher level, though, I write intentionally designed code for readability and scalability, which is something AI cannot reliable do in its current state. I’ve tried giving it my coding conventions in the form of markdown files for context, and it still messes that up!

How “AI bros” ruin the culture around AI

Ok, what does this mean? What is an “AI bro”?

In my definition, an “AI bro” is somebody who advertises and does vibe coding on a larger scale as a way to get easy cash and engagement. These people are typically within their own echo chambers (or in toxic development communities, like Roblox ones).

And I gotta say… These people are always blaming me for AI’s shortcomings, pointing to how I use the wrong tools or workflows. Was I using Gemini? Use Claude instead! Actually no that was wrong, use ChatGPT! Oh actually no that SUCKS and I’m an IDIOT for trying that, and the cash generator is actually Grok! What, why am I using Grok, I should be using Gemini! I should have been using agent workflow abc instead of xyz because they said so! Also, program 123 is SO outdated as of three days ago, I’m so behind!

Anyways, that loop is endless, and they can never be satisfied. The AI bro culture encourages entirely generating games with AI as a shortcut, but they don’t realize that AI code doesn’t perform well in live ops or large scale development. There isn’t much for me to say here other than they are just wrong (and that’s coming from experience, not vibes).

I usually ignore these people, since not once have I seen one actually see actual success exclusively from AI. When it does happen, it’s the rule, not the exception. Remember that for every one AI-generated game that does well, tens of thousands more never saw a playerbase stick.

Where AI belongs in learning

That’s one group of people, though. What about the new developers who just want to learn?

To put it out there, AI 100% has a use case for learning code, but it’s almost always used incorrectly.

New developers often times have AI generate stuff for them and call it “learning” by observation, but that is a mirage of progress and will go nowhere. There are lots of studies showing why you can’t learn by asking people to do things for you, and the same applies to AI. If I was in highschool again and asked somebody else to do my homework, would I be learning? No, of course not.

Instead, asking AI to set up lesson plans, quizzes, and challenges for a learner is a more efficient use of time and enegery, and results in a new programmer having the feet to achieve more. For example, if you’re brand new, you could ask AI where you can start, and what resources you can take advantage of to begin your journey. Or, if you already started, but want to learn a certain topic, have it teach you in a lesson plan structure!

AI isn’t the only solution to learning, though, and the old-fashioned ways are still awesome. I still always tell new game developers that the best way to learn is through experimentation.

Why I “hate” AI

Well, I don’t actually hate AI, but in 2026, having a nuanced opinion isn’t allowed, and you have to be all in or completely against.

Simply put, people misunderstand what AI is, when it’s most useful, and where it belongs. I’m not totally against AI for everything; for example, I use AI a lot for non-development stuff, like research and MTG commander/deck recommendations!

However, I do hate how people lose their jobs to AI to check boxes for investors. The other day, a friend told me that their job has a requirement where 30% of all code must be made by AI. Why? Because it’s the new thing, and people say it’s efficient!

AI is not a replacement for engineers, and instead works best when it’s utilized as a tool. One day in the future I’m sure programming will be ancient and prompting is the way, but until then, I’m not conviced people should lose their jobs over an immature technology.

A hammer is great for getting the nail in the coffin, but if you’re already in it, then maybe you don’t need a hammer to fix your problem.