Artists, Your AI Purity Test Won't Work Out
The programmers have no solidarity
First off, for SEO purposes the title says, “AI”, but I don’t like that name. Not only because a capital “I” and lowercase “l” are homoglyphs in most sans-serif fonts so it looks like I’m hating on a guy named Al, but also because there is no intelligence happening with these tools. I’ll just call them LLMs from now on.
Second, a little preface: I’m an LLM hater. Don’t use ‘em, don’t wanna use ‘em, and I'm lucky enough to work at a place where I don't have to. These models are built on the theft of intellectual property. They’ve impacted countless jobs already, especially those in creative fields like illustration, translation, and copywriting. They’re beloved by fascists, pedophiles, and criminals. Chatbots built on LLMs have already convinced many vulnerable people to kill themselves. The data centers running the LLMs are consuming millions of gallons of drinkable water and are poisoning their surrounding communities. Users of LLMs are also actively de-skilling themselves, and are failing to retain anything about the code, or writing, they’ve generated.
With all that said, it is with a heavy heart that I must inform you that your LLM purity test isn’t going to work out.
I follow game development pretty closely, especially indie game development. Ever since LLMs took off in popularity, the indie game community has been very vocal about their hatred for generative LLM slop. When the new Crazy Taxi was announced, it was celebrated for a few brief minutes until the Steam page with its “AI Generated Content Disclosure” was posted, and then was instantly thrown in the trash by the community.
And who’s to blame them? These game companies openly admit to using LLMs to generate concept art for their games. Hit games like Arc Raiders have their entire game’s voice acting generated by LLMs. There’s award-winning writers out there who are now struggling to get hired. Translation jobs have basically gone extinct.
All of these artists realize that the LLM threat is coming for their jobs, and they’re collectively pushing back with full-throated rejection and shaming anybody using LLMs.
Apparently programmers didn’t get the memo.
Recently Epic Games, the company behind Unreal Engine, had their Unreal Fest 2026 conference where they announced what was coming in their new Unreal Engine 6. One of the big topics was LLMs. Epic Games showed off how they use image generation LLMs to create characters for their hit game Fortnite. They also announced that they were making it easy to integrate LLM agents with Unreal Engine to have it generate code for you.
The announcement of Epic Games’ embrace of LLMs motivated indie developers to ask others on social media for alternatives, and most of the responses were to either: stay using Unreal Engine 5, or switch to an open-source engine called Godot.
Even though the Godot project lead has written threads on Bluesky in the past detailing how the maintainers of the codebase are inundated with LLM-generated pull requests, people started pointing out that there's been around 40 instances where LLM-generated code was merged into Godot. They’re apparently creating a more detailed policy for how to deal with LLM-generated code right now, but so far the policy is “you can’t have a PR that’s fully LLM-generated” and “if you use an LLM, you have to disclose it.” This wasn't acceptable to most indie artists.
Around the same time, the developer of Trenchbroom, a popular open-source tool for making Quake levels that indie devs sometimes use for their own games, revealed that he’s been using LLMs since April 2026.
This prompted an outrage amongst the anti-LLM crowd to fork a “de-slopified” version of Trenchbroom, and revert to previous versions of Godot before the PRs with LLM-generated code were merged. The mindset seems to be that adding any LLM-generated code to your project is the equivalent of adding feces to your soup; no matter the amount, it taints the whole thing.
Unfortunately, as a software developer, I’m here to say that every piece of software you’re using is tainted. At nearly every tech company, the programmers and their managers are absolutely chugging the Claude Code Koolaid, voluntarily and not. I can probably count on one hand the amount of people who aren’t using LLMs at the company I work for.
Does that mean that every software product that has code from Claude is vibe-coded slop? No. At my job, before everybody started pair programming with Claude, I was doing thorough reviews and making sure I understood every line of code. I’m still doing that. There’s a lot more code to review now, but I’m still reading every line and testing everything before I slap on the green approval checkmark.
I’m definitely not the best programmer in the world, but I’ve worked with people who were a lot worse than me. They were out there manually generating their own slop before LLMs, and I'm sure they're using Claude Code to generate slop now. But, I also know there’s people out there generating code with Claude that’s way better than anything they could have written. And, even as an LLM hater, I fully believe that the senior engineers who use Claude Code as a tool and have both the experience and taste to determine when its output is “good” are not “slopifying” their codebases.
So do I think that projects like Godot or even Trenchbroom are degrading into slop because of LLMs? No. Especially not Godot. The amount of expertise you need to work on a game engine like that is immense, and there’s enough eyes on that codebase that it won’t ever be “slopified.”
But there’s another half of the outrage equation: betrayal. Along with the theft and destruction that’s needed to create and run the LLMs, artists are also upset at the complete lack of solidarity the programmers have with their peers. When programmers use LLMs, the message artists are hearing is: we don’t care about the theft, we don’t care about the environmental destruction, we don’t care that it’s putting you and your peers out of a job.
If you’re working on a software product that’s used by artists, and want to maintain a good relationship: don’t add features that use LLMs, and don’t use LLMs to create your product. If your codebase is already tainted by LLMs, you can start repairing the relationship by banning any future LLM-assisted code contributions. Whether or not you go back and remove the LLM code that was already merged is a question of feasibility, but if it's possible, you should.
Artists also need to understand that this 100% no-LLM purity test isn’t feasible. Claude Code is ubiquitous on software development teams. You can try and get maintainers to change course, but you can’t live your life and avoid any product that has LLM-generated code.
Most of this outrage is happening on Bluesky, a platform that has employees openly talking about using LLMs. I was even talking to a woman whose son is a programmer for Epic, the healthcare software company (not Epic Games), who is required by his boss to use LLMs. If you’ve ever been to the doctor in the US, then you’ve probably used their product, MyChart. I know there's electric utility companies out there using software with LLM-generated code too.
So if you want to avoid LLMs do you stop posting on Bluesky? Stop going to the doctor? Stop using electricity? Where does the purity test end? I know I sound like that "Yet you participate in society! I am very intelligent." comic, but you've gotta actually make concessions somewhere.
In my opinion, at this point avoiding LLMs is like avoiding plastic. You can make personal decisions to replace your polyester clothes and plastic kitchenware, but somewhere along the supply chain there’s going to be plastic used. You can’t avoid the plastic wrap holding the boxes together on the shipping pallet, or the tires on the delivery truck.
Now to be clear, this isn’t me saying we’re doomed. Can we collectively come together and get rid of LLMs? Absolutely. We’ve done it before with infectious diseases, chlorofluorocarbons, and NFTs. But we’ll need some solidarity from the programmers first.
