AI / GAME DEVELOPMENT

Call of Duty, Minecraft and Skate in One Game? Welcome to AI-Assisted Modding

Modern Warfare 2, Skate 3 and Minecraft were never supposed to exist in the same game. A strange Rust project shows what happens when AI dramatically lowers the cost of asking: “What if they did?”

By Albin Zlatan Zuzic

A few days ago I came across a Reddit post showing something that looked almost wrong at first glance: Modern Warfare 2, Skate 3 and Minecraft mechanics all running together in one project. You could move through familiar MW2 environments, skate around with Skate-style movement, and then enter a Minecraft-like world with blocks, mobs and all the visual language that comes with it. It looked ridiculous, but in a way that immediately made me curious. My first reaction was a mix of confusion, nostalgia and genuine excitement. Part of that is because these games are not just random references from the late 2000s and early 2010s to me. Modern Warfare 2 and Minecraft were both important parts of how I first became interested in computers, systems and eventually IT.

Seeing them combined into one strange experiment felt like watching two completely separate parts of my childhood suddenly overlap.

Modern Warfare 2 was probably my first introduction to hacking

When I think about Modern Warfare 2, I obviously remember the usual things: Terminal, trickshots, prestiges, camos, emblems and multiplayer lobbies. But one of the strongest memories I have from that period is the rise of PlayStation 3 jailbreaks and so-called challenge lobbies. I was around nine years old when I started seeing my uncle and his friends jailbreak their consoles and modify MW2. They could enter lobbies where everything in the game could be unlocked almost instantly: prestiges, weapon camos, titles and emblems. As a kid, that felt almost magical. Having something like 10th Prestige carried a kind of status that reminds me a little of having a rare CS skin today. The interesting part for me, though, was not really the prestige itself. It was the question behind it: how were they doing this?

The game had rules. You were supposed to progress in a certain way. Yet someone had understood the system well enough to change the outcome entirely. That was probably my first real exposure to the idea that software was not something fixed that you simply consumed. If you understood enough about the system underneath it, you could alter it, manipulate it or make it behave differently from what the original developers intended. Looking back now, I think that curiosity played a bigger role in where I eventually ended up professionally than I realized at the time.

Minecraft taught me a different side of IT

Minecraft was almost the opposite experience. Where MW2 made me curious about modifying systems, Minecraft made me want to build and manage them. When I was around 11 or 12, I started hosting my own Minecraft servers. I configured plugins, managed players, advertised the server and slowly learned how different pieces of the setup worked simply because I wanted the server to function properly. One plugin I remember especially well was GroupManager. It allowed you to create roles such as Newbie, Builder, Moderator, Mod+, Administrator and Owner, with each role having different permissions. At the time, I obviously did not know terms like role-based access control, Active Directory or least privilege.

I just understood the logic intuitively: a moderator should not have the same level of access as an owner. Today, working in IT and security, it is funny to look back at that because I was essentially experimenting with simplified versions of the same concepts I now deal with professionally. Roles had different privileges, access was structured by responsibility, and the “Owner” role was basically my Minecraft equivalent of Domain Admin. I just did not know the terminology yet. That is part of why this mashup caught my attention so quickly. MW2 and Minecraft were two different games, but both shaped how I thought about technology in different ways. Seeing them merged together now, more than a decade later, hits a very specific nostalgic nerve.

So what actually is this project?

The project is called 2010 Rust Rewrite Mashup, and the interesting part is that it is more ambitious than simply putting Minecraft textures into Modern Warfare 2. The foundation is a recreation of MW2’s runtime written in Rust using the Bevy engine. The idea is that the new runtime can read original MW2 content such as maps, models, textures and weapons, while running through a new implementation. The underlying IW4L runtime’s README is unusually direct about how it was created: it states that the project was written by an LLM. That is where the broader AI angle becomes much more interesting.

The mashup adds a Skate-style mode where the player can move around using skating physics and tricks, while newer versions introduce a Minecraft-like world with blocks, mobs, inventories, world generation, day and night cycles and item drops. The result is exactly as strange as it sounds. You can have MW2 weapons inside a Minecraft environment. Gunfire can affect blocks. Grenades behave like explosives inside the block world. Skate movement can work inside that same environment. In other words, you can essentially trickshot a Minecraft pig with an Intervention and then get on a skateboard. If someone had described that project to me in 2010, I probably would not have believed them.

The interesting part is not that this was impossible before

I think this is the most important distinction. AI did not suddenly make game modification technically possible. Talented developers, modders and reverse engineers have been doing incredible things with games for decades. The difference is the amount of effort required. Projects like this involve a huge amount of repetitive technical work. Someone has to understand file formats, rendering, physics, game logic and unfamiliar codebases. They have to recreate mechanics, test them, debug them and then build all the connective tissue required to make unrelated systems behave like one coherent experience. People capable of doing that work have always existed.

The problem is that those same people usually have careers, companies, families and projects that are far more valuable than spending hundreds or thousands of hours answering a question like:

What would happen if I put Skate 3 inside Modern Warfare 2?

That is where AI changes the equation. It does not necessarily make an impossible project possible. Instead, it can make a project that was previously far too time-consuming suddenly worth attempting. That difference is enormous.

AI is reducing the distance between idea and experiment

The way I think about modern coding models is not that they possess some magical capability humans do not. They are more like enormous accelerators. A developer working manually has to read code, understand what it does, plan a change, write that change, compile it, test it, inspect errors, debug problems and repeat the cycle. AI can assist with many of those steps extremely quickly. It can read large sections of a codebase, generate implementation ideas, make changes, interpret errors and continue iterating. That changes where the human spends time. Instead of spending most of the process manually implementing every detail, more of the effort can go toward defining the goal and validating whether the result actually works.

That does not mean the human knowledge disappears. In fact, this project is a good example of why saying “AI built it” can be misleading. Projects like this still depend on previous reverse-engineering work, documentation and technical discoveries made by people over many years. The AI is not inventing all of that knowledge from nothing. What it is becoming very good at is taking existing knowledge and turning it into working implementation much faster than before. That is a much more realistic way to think about AI-assisted development.

The “vibe-coded slop” criticism

One thing I found interesting in the discussion around the project was how quickly people used the phrase “vibe-coded slop.” I understand where that criticism comes from. There is a big difference between using AI as a tool while understanding what you are building and simply asking an AI to “fix it” repeatedly until something happens to launch. Working software is not automatically good software. AI can absolutely create redundant code, poor architecture, security issues and bugs that a less experienced person may never notice. That is a real problem. But I also think dismissing every AI-heavy project as “vibe coding” misses something important.

If AI makes it possible to create a strange experimental project that nobody would previously have had the time to build, I do not really care whether every line was handwritten. What matters is whether the person behind the project understands it well enough to test it, validate it and continue improving it. Engineering discipline still matters. Manual typing is not the same thing as engineering.

Modding now has a much bigger playground

Games have always borrowed ideas from each other. Fortnite is an easy example. Its original appeal came from combining battle royale gameplay with a simplified building mechanic. Those concepts were not originally created together, but combining them created something that felt new. AI potentially pushes that process much further. Instead of borrowing only the idea behind a mechanic, developers and modders may increasingly be able to recreate the mechanic itself and experiment with how it behaves inside another game. Personally, that is where my imagination immediately goes. What happens if you take GTA’s open world and combine it with Forza Horizon’s driving?

What if you then add the mission structure and hacking mechanics of Watch Dogs? What if even the small activities inside GTA, like golf, were powered by the mechanics of a dedicated golf simulator instead of a simple minigame? Specialized games can spend enormous amounts of development effort making one mechanic feel extremely good. Open-world games have to divide their attention across hundreds of mechanics. If AI lowers the cost of combining those specialized systems, the possibilities become much broader. Not because every game should become one massive universal experience, but because experimenting with those combinations becomes far more realistic.

Big studios are interested in modding too

This is not only happening in small GitHub projects. Rockstar’s acquisition of the team behind FiveM and RedM is a good example of how important modding communities and user-created game worlds have become. That acquisition does not mean Rockstar is suddenly going to turn GTA into an AI-generated mashup platform. But it does show that major publishers understand the value of letting communities extend, modify and build on top of existing game worlds. AI adds another layer to that. Historically, the barrier to creating something was often: Can I program this? Increasingly, that question may become:

Can I explain clearly enough what I want to build? Those are very different barriers to entry.

It also makes me want to build something myself

Seeing the MW2, Skate and Minecraft mashup immediately made me think about what I would personally want to create. One idea is much smaller and more emotional. I have owned cars that I no longer have. There are already incredibly talented Assetto Corsa modders who recreate real cars in impressive detail. What I would love to experiment with is using photographs and information from cars I personally owned to recreate those exact cars inside the game. Not just the same model. My actual car. The same specification, appearance, details and personality. Eventually, I would love for something like that to become easy enough that other people could do it too.

Cars can have a surprisingly strong emotional value. Sometimes people sell them because their circumstances change. Sometimes the car gets damaged. Sometimes keeping it simply stops making financial sense. But the memories stay. If someone could take old photographs of a car they loved and recreate it convincingly enough to drive again in a simulator, that starts to feel like more than a game mod. It becomes a kind of digital preservation. That idea would have sounded unrealistic to me not that long ago. Now it feels increasingly achievable.

That is the part of AI coding that excites me most

The most interesting outcome of AI-generated software, at least to me, is not another automatically generated business dashboard. It is the ability to lower the cost of ideas that are not commercially important enough to justify traditional development effort. The weird ideas. The nostalgic ideas. The ideas someone builds because they think it would be funny. Projects like combining three games from more than a decade ago simply because someone wanted to know what would happen. Historically, projects like that required a rare combination of creativity, deep technical skill and an unreasonable amount of free time.

AI does not remove the need for knowledge, testing, reverse engineering or human creativity. But it weakens the relationship between an idea and the amount of manual work required to test that idea. And that changes what people are willing to try. The distance between:

“Wouldn’t it be cool if…”

and

“Wait, I actually built it.”

is getting shorter. For me, watching someone skate through a Minecraft world while holding an Intervention is one of the strangest and clearest examples of that shift so far.