Behind the Three-Letter Tweet That Pissed Off the Entire Games Industry
"omw"
Before diving into this week’s piece, I should mention that (via my day job) I had a rare opportunity to do a live video interview with
, the CEO of Substack. Check it out if you’ve got 20 minutes. The intended audience is founders, but there are interesting tidbits in there about the way this platform’s creators think about the future of media.
I spent last week’s newsletter complaining about AI-generated content. This week I bring you an interview with the guy behind the most infamous AI-generated video clip in recent memory.
The true appeal of X (née Twitter) has always been that it’s a place where any random person can post something that throws millions around the world into chaos for days on end.
One example: A few weeks ago a Midjourney user pissed off the whole games industry with a 5-second GIF, captioned with just three characters:
If you don’t wanna click through to the post, here’s the full video:
It’s definitely more interesting than most AI art. The way it renders different layers based on distance is striking, and those sprite-based arms do look great. I can think of a few games that look similar to this, though nothing I’m aware of looks exactly like it. At first many commenters assumed this was an actual, playable game, but the OP confirmed in comments that he used Midjourney to generate the clip.
So how’d the internet react?
Mostly, people split along five basic lines:
1) This visual style already exists:
2) This visual style doesn’t yet exist but I’ve attempted to re-create it (I love this one):
3) This visual style will never and could never exist.
4) This visual style does not yet exist, but it could in theory:
5) This visual style did not exist before this week, but it does now thanks to an AI model even more advanced than Midjourney:
So that’s a lot of basically irreconcilable takes on a five-second clip.
But there was yet one more take, proffered by the OP himself, and it was the boldest of them all:
This doesn’t exist yet, but I’m assembling a team to make it.
What chutzpah! Why in heaven’s name would a Midjourney prompt guy be the one to actually build this? And yet, two weeks later, our hero reappeared with an announcement that he’d formed a new games studio.
Raising the stakes even further, the intrepid poster said on his Discord that the game is currently being made without assistance from AI.
What! Any reasonable person would have some doubts about this. And the response online so far has been skeptical:
Sure, it’s a longshot. But you’ve gotta admit, it’s kinda fun that this guy has effectively staked his online reputation on delivering something almost nobody believes is possible. Like, this can really only go one of two ways: Either Desimulate ships a game that delivers on this visual style, or he won’t.
Alea iacta est. The die is cast.
Either way, the responses to the original tweet were so wild and varied that I wondered how the original poster felt about it all. I DMed Desimulate and requested a live interview.
He agreed.
What follows is a condensed transcript of our conversation.
PUSH TO TALK: So what’s your story? Why’d you start posting these Midjourney clips?
DESIMULATE: Yeah, so I'm an engineer. I've been doing CS for a long time. I’ve got three kids. Sometime around the election, I saw that people were getting paid for posting on X. Some were getting like 1,500 bucks. I was like dude, I could use an extra 1,500 bucks. It can't be that hard. I already work on my computer all day. So I speedran monetization. I put in 30 days, just to see if I could do it. And then it was just way too much, keeping up with the internet. I was way too online. It takes a lot of paying attention to your phone.
But then I saw these guys posting AI art, and I didn’t realize it was AI-generated. I asked one of these guys how he made it, and he was like: Midjourney. I had no idea. So I just started dabbling in it, and then one thing led to another and within a month I was super addicted. So I transitioned my account to posting art.
So did you get monetized on X?
No, I never made any money.
(Laughing) Oh my God.
They never paid me. I was hitting them up for two months. It said “your payment is coming” and they never sent me anything. And then they just sent me an email and they were like, you've been paused, your monetization has been paused. For no reason. I was like, great, sick. And this was just when I started gaining a lot of traction with my art. I went from, I don't know, 3,000 followers to 8,000 followers in two weeks or something like that. And then it blew up from 8,000 to 12,000. It was just pretty steadily increasing on my account. So at that point I was just doing it for the love of the game. And I got really into it.
I like to say Midjourney’s not just a tool. It’s an art engine, because there are all these different ways that you can create with it. There are srefs, you can do mood boards, and then you can train your own model, and it learns as you go, and you can combine all this stuff together. I just got obsessed with making something unique, and that post was the epitome of that for me. And it kind of came out of nowhere.
In order to get srefs, there's not always a library or anything like that. Sometimes people farm them and then they post libraries and you can buy the library. But when I started this in June, this was on the new engine, so the srefs had just released and nobody had farmed them yet. So I spent a month mining these srefs. I mined 5,000 codes.
I have no idea what that means. What does that mean when you say you were mining sref codes?
So, you just run a prompt. Maybe a knight standing in front of a castle. And the srefs are these codes that are weighted to make the model shift toward a certain style. One might be like a pencil sketch, but then another could be a pencil sketch in a frame, but then another could be a pencil sketch with a little bit of red. They vary ever so slightly. And you can find them by generating random srefs and applying them, one at a time.
Oh, and so you basically A/B test them to see what they do?
Yes. Over and over and over again. I’ve literally run 15,000 prompts.
My thing is going really deep into the srefs. I have styles that are like 12 to 15 srefs.
I got it, and so you wanna keep those particular srefs secret, because that’s the secret sauce to generating stuff like the “omw” clip?
Yes. I'll even show you, I have an SRF farming folder with like 20,000 generations.
Okay, so this really leads into my next question. Obviously you have respect for these tools. You're putting in serious time and effort with these tools and exploring the boundaries of their use cases. So that make it even more funny that, as you said in your Discord, there’s zero AI in your dev workflow on the game you’re working on. I mean, why not? It’s not like you’re scared of the stuff. Why not use AI tools to make this game?
Well, right now, the tools aren’t really good enough.
(Laughing) For real?
Yeah I mean, the meshes are probably good enough to make a basic game. Something small. But this is a game that is kind of demanding, right? It's going to be a real fantasy RPG. So at the end of the day, it's going to have to be performant. The 3D generation stuff right now… the meshes are just not up to snuff. They are way too many triangles, polygons, and a lot of them kind of get messed up. And so it actually ends up taking more time to clean them than it would to just like start from the beginning. So we just use it as a visual reference.
I really did not expect you to say this.
Yeah. It's sort of a challenge, right? Can we do this with traditional tools? Because people are saying it couldn't be done. I realized pretty quick that if I was going to do it, I was going to need some real game devs to help me out.
I'm a software engineer. I've got 10 years under my belt, but using engines, it was just gonna take me so long to get off the ground. That’s why I hired a team. I knew this was gonna be a battle because I inspired a lot of people and there's gonna be people out there trying to make the game too, right? And so there's a little bit of a rush to actually get something going.
So let me ask you a classic venture capitalist question. This is one that, in VC, we would ask anybody pitching a company: Why are you the one to do it?
It's a good question. I've got the vision, and everything starts with that vision. I've got the passion to make it happen. In two weeks, I've interviewed 40-plus people, brought a real team together of real game devs, started building, designing the game, and we've made a ton of progress. Obviously everything is in an R&D state. But I am committed to making it happen.
You've kind of put a stake in the ground. You've put your name, or at least your handle, on the line, right? You've come out and said very publicly, you're going to try to do this. You're either gonna prove the haters wrong or this might be like a dev hell thing for years of your life. I hope it's not the latter, you know?
Yeah, but also I'm the OP, right? Like I was the first one to do the thing. I started the movement and so I carry that weight in... like I've got all the followers with it, people are watching to see if we can make it. Or they see me as a source of controversy, which is really good for marketing, because, like they say, there's no such thing as bad press. And the more controversy the work creates, the better it is for us because people will be talking about it.
We talked a bit about why you’d try to make it with a traditional game engine versus AI tools. And you talked me through the limitations that the AI tools have. Now let me ask a more philosophical question. Why make it at all? Like, why take that generated clip and try to make it into an interactive experience? What's the motivation there?
I've always wanted to make a video game, my whole life. I’m from Michigan. There's not really a lot to do here, half the year. It's gray and cold. You spend a lot of time inside. As a kid, I played lots of video games. So since I was a kid, I've been thinking about making a video game. And you might ask, why didn't you ever try before? I actually have tried before, but it just was one of those things that… it never felt like it could be successful for me.
But now, after we posted the clip, everybody was saying they wanted the game. And when it comes to games, half the battle, if not more, is getting people interested in your game. Because everybody's trying to make games, right? So the fact that there was that much interest, that quickly, and that many eyes on it. And within 24 hours we had 4.5k followers on the studio account. We've got like 500 people in the Discord, I've got over a thousand people on the mailing list already. In less than a week, we’ve got lots of people paying attention to the game and lots of people interested in the community and so, to me, that's an opportunity that is too good to turn down.
It's almost easier to start from having aggregated the demand for a game and then making it, versus making a game and then trying to generate demand for it.
Yeah. It was like an accidental layup to myself, you know? May as well take it.
That’s it for this week. I’m gonna go find something else to get mad about on Twitter.
Almost feels like more studios should try using AI to gauge demand (without overselling too hard) before going deep into production 🤔
Godspeed on your way! 🫡