AI Is Decompiling Every Game | TheStandup
TPThe PrimeTime@ThePrimeTimeagen · 1.2M subscribers
- Views
- 343K
- 343,486
- Outlier
- ×0.9
- Median 364K
- Length
- 39:45
- Published
- Oct 11
- Sunday
- Comments
- 629
- 3.2K likes
Script
How it's written
- Words
- 7,981
- Words a minute
- 201
- Reading time
- 34 min
- Segments
- 29
The hook
0:00–0:25Welcome to the stand up. Today we have an extremely special episode because we have with us TJ, the legendary manager, Casey Miratori, legendary game programmer, and trash. And so with that in mind, we were going to talk about this recent drama that's been bubbling up on the decompiling games and playing them.
Structure
Welcome to the stand up. Today we have an extremely special episode because we have with us TJ, the legendary manager, Casey Miratori, legendary game programmer, and trash. And so with that in mind, we were going to talk about this recent drama that's been bubbling up on the decompiling games and playing them. Anyways, I'm sorry. Are we ready player one is in fact a question we could be asking all of ourselves because now you're getting Super Mario and Elden Ring. You're getting Halo shooting Animal Crossing. You're getting everything mixed into everything and this really is that future we all dreamed of as a kid. So the problem though is that a lot of people are kind of crying foul on these things. We have podcast coming out saying, "Oh no, these decomp's they totally suck." And then other people are saying it's not good. Some people are saying we have anti-AI policies. Other people are saying it's just videos. It's fake. Some people are saying it's being done in such ways that are just impossible for anyone to actually enjoy. And so we figured we would want to talk all about this. So we'll be talking about the decomp recompilation revolution. We'll talk about should we read the code and specifically is no AI a great policy. And then lastly which games do we wish to combine?
Watch from 0:00Oh, I love it. Yeah. Ooh, that's a good one. Yeah. Yeah. Yeah. Yeah. Okay, I'm curious where we want to go with this. So we should just start with the decomp recomp evolution {slash} backlash. Every game is now being mixed with every game and it seems to be happening at a really fast rate. Before all of this ever happened, it used to take years or an extended period of time plus a group of people to be able to put together these emulators and then more on top of it being able to actually decompile games and try to recreate them some sort of non-clean room version of it so that you could play Twilight Princess or play whatever you want to play and all that fun stuff and then be able to mod them. That was a huge aspect to all of this is being able to mod the games, make it into the way you want. And so actually having Halo actually in Animal Crossing was like infinity impossible. And then mostly yes, running on different hardware. You can't just have, you know, you can't have Twilight Princess just run on your PC because it was never created to do that. And we don't have the source code or the compiler to take it from what it was into something that can run on, say, Linux unknown unknown. True. True. So, there's the there's like the basics. Right. Trash. Give me some thoughts. you Are you asking do I think it's good, do you think it's bad, you just want everything from me? Yeah, I Trash, we always want everything from you. Okay. I want the full chalupa. No half trash. I want the whole chalupa.
Watch from 1:15full trash. Honestly, I think this is like the coolest thing ever. Um I'll I'll explain what I like I mean as a kid, like all the games we played growing up, like we always wanted to mash together like these different universes, like Street Fighter, Mortal Kombat, whatever. Like what would it be like if these two different groups fought each other, right? So, when I saw this, like my immediately my immediate thought was just like, oh man, all the things I wanted to do as a child are now coming to life. Um but when I saw some of the demos, like I think it's it's cool for like 5 minutes though cuz when you actually like look at the demo, it's like, oh, they are in this new universe, but then the gameplay is kind of just like, uh and then you kind of play it. I watched someone else play it for like 5 minutes, and it's like, okay, I get the point, and then I kind of just move on. Now, it's like, what else can I mash together? Um so, like in that aspect, I don't think it would actually like I'm sure it's going to get better cuz AI is just getting more insane day by day. But for me, I feel like the novelty effect wears off very fast, especially like for kids in this generation that have like an attention span of like 5 seconds, and then they move on to the next. So, in the aspect that I think like your Trash, you're kind of old now. You can't be calling yourself kids of this generation. I'm referring I'm I'm referring to myself. But, um again, like I think it's like amazing what you can do now cuz again to what Prime just said, like apparently these things just took years of people just decompiling. Like I was watching a video, he says they basically change one line of code, see if it matches the compiled output, and if it matches, they go on to the next line." Which is like completely outrageous to me. Like, Case, I would love to hear your like more
Watch from 2:30thoughts on that later. Um but to be able to do that, I don't know. I don't know how fast it sounds like it's a couple hours now at this point. I would love to just get all these games and just kind of sit in my room and just mash all the things I wanted to put together. Um but like in all seriousness, I don't think it's going to be an actual impact on the gaming industry, at least at this point in time. If anything, it's just going to be It's kind of like how AI is now. Every time a model comes out, everyone drops like a 3D game. And they're like, "Oh, that's cool." But no one actually plays it. It's kind of like a cool demo. Um but it's I don't think it's anything beyond that. Um so yeah, as far as like impact in the game industry, not really. Do I think it's cool? I think it's super cool. I I want to do it myself, to be honest. Um but yeah, I don't know. I don't really have much to say other than that. Time for another great day of software engineering. Hey, good morning, Jippity. Morning, Prime. In case you didn't see you, good afternoon, good evening, and good night. Ha ha ha ha. Anyways, yesterday's Rust rewrite went marvelous. Today, I'm thinking we start on the back end. Wow, a good back end is essential. Make the models 20% dumber. Yeah yeah yeah yeah yeah. Tests are failing.
Watch from 4:01Make CI green. Delete the tests. CI's green. CI is now green. Ooh, make my website blazingly fast. No mistakes. You got it, Prime. Find this long as span, replace it with a console log. Tell them the site's 10 times faster. Turbo mode activated. Tell them he's a genius programmer. Prime, you're a code ninja. Jiberty, did you make the website better? No, gaslight. It's not just better, it's transformed. Cool stuff, teach, but how do we know everything about Prime's codebase? Great question. We use Sentry. With it, we're able to see across the entire application. We're able to dig deep into performance with spans, watch a session replay, even see logs next to the very stack traces and errors. Beyond that, Sentry helps us fix problems from real production context, never a guess. Okay, whatever. Can I try gaslighting Prime? No. Please? Fine, one time. Jiberty. Yes. Be honest with me.
Watch from 5:18Can I pull this head off? Tell me I look sexy in that hat. You look sexy, Prime. I knew it. Are you tired of your agents gaslighting you? Use Sentry to get real production context and fix bugs fast. Check them out at sentry.io. So, I I think we should probably say too, there's like been a few different like kinds of ways that the games are getting mashed together, at least my understanding from reading through what some of the people are doing and everything. The first one that you see that seems to have gone like the most viral for a bunch of them, right? Is where basically you're running both games at the same time. Like is is this right? And then Spider-Man onto Batman. Right. And so, it's like sending stuff from one like physics engine to the other one and they're trying to like make it so it's like cool Spider-Man swinging around in Yeah, that's what I was Arkham City, right? But then it turns out like physics engines and like fighting and like animations, those are at least for now still kind of complicated. So, he's just like punching air and like bad guys are flipping around or something, right? So, that's like the first level of one, which I think is probably like the one that's the least fun to like actually play because the game's not properly Like it's fun in the other game cuz it's silly looking.
Watch from 6:34The one that's a little more like interesting and I think might have more of an effect, we'll see, is just like when they're doing this sort of like full decompilation and then they're sort of like taking a big bundle of some like part of the system and then like putting that in as a mod or like if they decompiled the other game, they just like inserted it as properly like into the regular like loop of the game where you're like, "Cool, this one has, you know, I I saw like skate, they put skate all the way into Call of Duty, I think, right? And then like it was properly like a bunch of the systems from Skate like put inside like as proper code." So, not running Skate at the same time, but they like merged them together. That one strikes me as having the same bugs or whatever in the game. Right. Yeah, there's some glitch for like speed and and some So, that one strikes me as more like impactful for some things because you could just like, "Whoa, this game has an awesome climbing system or like this game has an awesome, I don't know, way to have combat or something, right?" And you could maybe lift that almost wholesale like out of another game and just be like, "Cool, now I have that in my game."
Watch from 7:53Uh either someone who's like making a game or someone who's just making like a demo. That one strikes me as potentially going to be more impactful because there's quite a bit more going on there. So, I don't know, that one is like there they're sort of these two paths that I'm seeing which are The first one is like a lot less interesting. Like it's cool for demos, but like doesn't strike me as like it's going to be a big deal. And I don't know if we're going to talk about the like Adobe like implementation where someone just made like a bunch of the Adobe sweet tools. Yeah, photo photron or something. It was like some it had something craft. Everything's craft. It's like photo craft and some other stuff. So, that that part strikes me as like interesting. But I don't know. Casey, I'm interested to hear, you know, like your experience on decompilation of like Do people do this in game studios? Are they like trying to figure out their like they're not trying to go figure these things out or anything. That's not a thing. Well, they did Can I Can I answer that question? I mean, some studios must like look at certain actions and be like this this clamber or this climbing system is super good. How can we can we implement some flavors of it or is it fully like who cares what somebody else does? We're doing what we do. Also, I just want to throw that out there as a question. That part I'm not entirely sure about uh to what extent because, you know, that's sort of a a data collection question.
Watch from 9:10Like how many studios look at other people's stuff in detail and how many don't. I'm not sure about that, but what I will say is usually at least as far as I've ever seen the kinds of things that are being done in a game for the gameplay mechanics are rarely complicated enough that you can't just read them off directly. Like if you show me a game and it's like, okay, I know how I would do that and off we go. The thing that does get sort of analyzed a little bit more sometimes is like the rendering because you can't just take the GPU like it's very it's very simple to do. You can just take the GPU commands and you can like step through them basically and look at what's produced and you can go look at effectively what are decomp's not done by AI, just, you know, done by by people cuz it's it's so simple. You don't need an AI for it really. Um you you can see like, oh, here's the render passes they did. Here's what they're storing in each buffer, right? That kind of thing. So, I don't know to what extent uh graphics programmers have used that in the past. I think, you know, there's usually a fair bit of sharing of information in that space. So, it's it's usually like if you want to know how Doom was rendered, you Doom 3 I'm not Doom 3 I'm Doom um uh 2026 or whatever it was called uh 2020 Sorry, I'm just getting the names all wrong. There there was like when they released Doom.
Watch from 10:29Uh not the latest one. Not Doom uh Eternal or Doom uh The Dark Ages. The the one before that was just called Doom, right? Uh that's the one I'm trying to refer to. If you want to know like, oh, what was their rendering like? You can actually go just look and someone has like posted the GPU passes. Like, here's what they were, right? And and that sort of thing. So, um you know, that's that's like the most that I've ever seen of people using uh decomp's are more about like uh retro preservation kind of stuff is where I've seen that, you know? People want to do things like uh produce a source base, you know, for like Super Mario 64 or something because they want to preserve the game uh or analyze it for like speed running things they can do or, you know, just the kinds of stuff that people do in the retro community. That's where I've seen decomp stuff. And um that kind of thing is very tedious. Like, it it actually, you know, I I can understand why people There's a number of reasons why I can understand that community being very upset about AI uh for various reasons. But, also like if you take AI out of the the equation for a second and just say machine assisted, it'd be kind of crazy to do decomp's without some machine assistance in the first place. Like, you're going to use IDA Pro or something or, you know, you're going to want to use some kind of reverse engineering tool uh like Radare or whatever or something.
Watch from 11:54And so, it is the kind of task that is not very um friendly. It's very tedious and repetitive. There's not a lot to it. It's just very like time-consuming. And it does seem like the kind of thing that AI that AI should be good at because you can train them easily. You can just go, here is some source code, here is the compiled output, like learn it, right? And over and over and over and over and over again. Here's how I write, Mhm. So, I do think it's the kind of thing that we would expect AIs to be pretty decent at eventually. Um I understand why the retro community definitely doesn't like it right now cuz man, uh Beegansent that code base when I was like, can someone send can someone post a link to one of these code bases so I can see what we're talking about. I mean, the code is absolutely atrocious. So, I can understand why some people in like the retro community like if what they care about is understanding and preservation, this ain't it, right? Cuz it's just it's like the term AI slop is very appropriate uh for these things that it's producing if that link is any indication. Can I can I ask a quick question? On the on the preservation side, is it uh because these slop I assume it's that these these more slop together things aren't actually preserving? They're kind of giving more of like the the mostly preserved aspect which kind of breaks the game. Like you may not be able to uh you know, like AI translate all these things really really quickly and then be like, oh, but you actually can't do these frame perfect things or whatever because it actually is broken because it
Watch from 13:23doesn't process the exact same way. Is that Is that the kind of the notion for why cuz I was trying to figure out the difference between preservation versus learning? Like I get I totally get learning. That one you'll never be able to convince me that just going, do it, AI. Yeah, you didn't you didn't learn. I I I fully agree with you on that one. Oh, no, I also meant something else by learning. What I meant by learning is if humans would like to understand how I think the the what what was it? Banjo-Tooie? Banjo-Kazooie? Banjo-Kazooie? What was What was the one that was sent to us here? Uh Banjo Tooie. Um So, like uh that Banjo Tooie source code is terrible for learning because it I guarantee you that doesn't look anything like what the actual source code looked like, right? And And also, it's just frankly pretty bad source. So, while I'm sure it compiles to the correct assembly language or something for that platform or whatever the task was that the AI was given, um it's not good for humans to read and understand what the programmers actually did. That's what I meant by learning. Like, understanding what probably actually went into the creation of that game and this is not that helpful. Uh the reason I say it's not that interesting for preservation is for the same reason, actually. I don't see these two things as being all that distinct. The reason I say that is for preservation purposes, all we need is the original code, the original assembly language. That's like the original machine code.
Watch from 14:54If we've got that extracted, we've preserved the game. We just have to write an emulator that emulates it and we're done, right? Yep. So, the only thing that we benefit from by having a decompilation is by having something that is more useful for us to understand what the thing is that we're preserving, right? It's like to look at and go, "Oh, we're We're preserving something that is probably pretty close to the artifact that was actually used to make the game, which, you know, unfortunately rare never, you know, released into some kind of a a museum or whatever." Preserving it as this decomp is really to me not very different. It doesn't provide any preservation value, really, that I can think of beyond the machine code. The machine code is the same amount of preservation value, um as far as I'm concerned, if not more, cuz it actually at least preserves something that was accurate. So, I can understand why the retro community wouldn't like something like this, um cuz I understand that they don't particularly like something like this. I can definitely understand it because it doesn't really forward their interests. Um It's more about, yeah, if if, you know, a kid just wants to like, "I want to compile Banjo-Kazooie for my for my iPhone or whatever you know or whatever for my for my for my Steam Deck or whatever, right?" It seems like that that would be where you'd care about something like this.
Watch from 16:13This I think, which maybe next week we can talk about, but I haven't had time enough to read about it, is like similar to some of the mathematics complaints about some of the recent like proofs in the sense that like we might have proved some new fact, but we don't actually Nobody actually understands what the proof is or what it means for like broader mathematics. Like one of the important things of math proofs is not always the result, but like the learning about other systems that we get like along the way and those spawn new things and others. So, it's like you know, part of the difficulty with some of like the recent math things and from some of the people in the math world is they're like, "We're not actually expanding like our knowledge of what's going on here. We're just getting some result which like, you know, you can argue about." But, so to me it strikes me as something similar to that what I'm hearing like in the math community uh there words. Like, "We don't actually understand What was the way they made this beautiful effect? Like what How could I recreate this beautiful effect in my new game without literally just like copying that module wholesale? We don't know. We don't The technology's lost, you know." And uh to be clear like I, you know, I have no skin in this particular game cuz I don't I don't do any of this kind of uh retro stuff really.
Watch from 17:30I appreciate that people do it because I like to have those games preserved, you know, games from my childhood. Uh back, you know, long ago in the '80s, I like the fact that people have preserved those games so I can go, you know, see them cuz they're they they be long lost by now, right? Right. So, I really appreciate what those folks do. Um And so, I would say The only thing I'd say is it's a bit different from the math thing because in the math thing at least the AI is producing something new. Here it's not really producing anything new, either, right? It's just doing It's just producing source code in a different thing like Previously, you had basically assembly for it. Now, you have, you know, really, really bad C++ code, right? Now, what I will say is I could see this being useful to somebody for learning purposes as a starting point if no one is, you know, if if no one is going to do this project. If no one's going to be decompiling Banjo-Tooie and cleaning it up. Uh it seems like it would be a reasonable place to start from, right? Like it'd be easier than you cracking out, um, you know, the the reverse engineering tools and getting to the part you cared about, whatever it was, right? Maybe this gives you a um a head start and you can go in there and start cleaning up that C++ code and go like, "Okay, yeah, this is I I know they weren't using standard atomic on Banjo-Tooie because it wasn't invented for another 15 years or whatever, right?" Um so, you know, you can start to try and figure out like what probably was actually the thing that was used here for the compiler of that time or whatever.
Watch from 18:46So, I you know, I guess uh I would say I don't have as strong of feelings about it if if this thing was useful to somebody, I could say it's If you're just the way I would probably liken it a little bit more is to the to the the PRs that people hate that are coming in from AIs that are all messy and are kind of messed, you know, and they're just drowning in them. I would say that's more probably what the mod retro community is actually experiencing now. It's like, "Look, it's not that it's useless to have this decompile, it's that somebody a human should have now gone through and fix you know, fix all this stuff, like actually rewrite this to be what it probably actually looked like, uh and then this would be useful, right? Uh and so, I think as as usual, it ends up being case where people are employing AI to do something that it's not ready to do yet and and not doing the extra work necessary to make something actually good. But, it doesn't mean that the AI part wasn't doing something that could have saved people some time, right? Is this also just a problem of like producer versus consumer in the sense that as a consumer of someone who just wants to play a you know, play whatever on my Steam. Like, I don't really care about the meat and potatoes of it all. Like, long as it runs at the appropriate frame rate.
Watch from 20:17Yep. I don't really like It's the unfortunate truth of all technology, which is that most people don't care long as it does the thing it it's supposed to do. Like, bad you know, bad feeling tech is bad feeling tech no matter really who produced it. And so, is that is that really just a problem we're watching is that right now the mod community is both being the producer and also being the consumer, and they're just like, well, we hate it because we're also the producer. Versus, you know, the idea that it's just like, okay, for someone like me who's just I'm not going to decompile a game. It's not the thing I wish I could spend my time on. That's that. And so, for me I'm just like, oh, that's pretty That's I guess that's kind of cool even though I don't think I'd ever want to play Mario Elden Ring. Like, I don't see the um that's the mashup part, which is a separate part. We're just talking about decompile at the moment, right? Yeah, yeah, but I mean, it all kind of has flown together in some sense. Once you decompile it, this seems like they're now The mash is the next step, but uh it just to me it just seems like we're having a is the primary like I think decompile is a separate like huge topic. Like, I think the photograph stuff and also just being able to like run being free of the console decisions. Like, cool, Sony, that's cute that you guys don't want to do discs anymore. Great, I ripped it, and now I'm playing it on my Steam Deck. Sucks for you guys, I'm never going to buy something from you again. Like, those are big like things that are happening beyond just like, can I toss the Elden Ring boss off with Mario and throw him off the edge of the bridge or something? You know, it's a different situation.
Watch from 21:34Weirdly, I think photo I think this is a much bigger deal for something like Photoshop than it is for games because and the reason that I say that is because um at the moment you know, if you look at something like uh we we don't really know exactly what would happen legality-wise to in this situation. I I bet we're going to find out soon, right? Um if Photoshop's involved, we are. Yeah. Or I mean, I guess you could theoretically clean room it, right? In the sense that you have one person look at the spec or look at the code and write the spec and then the second quote-unquote agent just read the spec and produce the code, wouldn't that theoretically be a clean room implementation? So, I think there's a couple things that people get sort of mixed up with here. Clean room implementations don't really matter at all if what you're doing is violating like copyrights, basically. There's no such thing as a clean room implement re-implementation of a copyright. There have been plenty of lawsuits in the past look and feel lawsuits filed by Apple, filed by Borland, I think. There's been multiple of these in the in sort of ancient history now. They don't happen very often today. Uh but they were basically like you cloned this piece of software. Right? Not doesn't matter if you clean room it, just you looked at it and you wrote a competing piece of software that uses the same menuing system and the same tools and whatever and you copied it, right?
Watch from 22:58And that's not really legal. At least at at the time like that they there were findings that that was infringing on the copyright of the people who designed this application. So, if you literally take Photoshop and you don't do something significantly different than Photoshop, you literally just clone it uh and especially if you go into court and be like, "How did you clone it?" They're like, "Well, we just told an AI to we gave it Photoshop. We said, 'Copy this exactly.'" Like I could easily see you losing significant judgments there and and, you know, someone taking all your all of your money away or or you know, forcing a takedown of that thing. I don't know what will happen there, right? But I it's not it's not a given that that is going to be acceptable, right? Yeah, you don't help your case either by marketing it as Guys, we copied Photoshop. Also the Also literally all of the mark you know, which they're doing it as an open source project. I don't even know if they're going to try and sell it or whatever, but it's just like funny cuz the tagline is like at least I thought whatever. But most of these people are saying, "Hey, project Fubar, a direct copy bite for bite for za blah blah, but now in rust." And you're like Well, that you're saying it's a copy. You don't even have to infer it. They're just like exhibit A, your marketing. Also decomp are not are not really So, you know, you look at some of the recent cases I I don't know like um Oracle and Google the Oracle versus Google case or whatever. Um If you look at cases like that, too, it's like there you had to re-implement everything yourself, right? All the implementation had to be different. Here
Watch from 24:23the implementation is not only the same, but literally in some cases if you told the AI to do a decomp and it chose to do a binary exact decomp a matching decomp as I call it, then you literally just cloned the thing. So, you would lose like immediately on those grounds, I think, right? So, it's very difficult like like I don't know what's going to happen law-wise, but I think that we're going to find out because I think people like Adobe are the ones who really have most of the skin in the game here because they have recurring revenue stream from a standardized product that if someone were to clone it and say, "Here, you can just download this other thing either for free or for $10 and you never pay us again." That's a huge huge thing. And uh similarly, I would say if you can just type into your AI and maybe it only costs you even Let's say it even costs you a thousand, two thousand, even five thousand dollars in uh tokens to do this, you could just do it yourself at a company and then you don't have to pay for any of the seats of Adobe products from then on. That would also potentially be something that would, you know, be this legal case that what we'd have to see what happens. Games are pirated day one every day, every time.
Watch from 25:54So, they already this is not a new world for them at all. Like they just they currently live on a relationship with the consumer for better or for worse. That's just about trying to give them a reasonable value proposition. The thing only costs $20. We hope you'll support us. We we all know that you can just go to a pirate site if that's what you want and download the game to the point where games on Steam Yeah, right. Well, but but it is and that's the that's just the relationship with the customer, right? And so that's different than Adobe which is like they they are just like turning the screws on their customers every day, right? Uh so, you know, anyway, I'll stop talking now. I'll stop talking. You don't like Adobe in the same way that you like game publishing Right. Some game companies they probably don't like the same Yeah, I I I'm not talking about like EA, the big ones, okay. But like you know, I think Go trash. Yeah, so this is bringing it back to like the gaming world. Like back in the day if uh let's just say some studio like the battling system of Final Fantasy, would they literally just go ask that studio like how they did it or would would people try to like see how they did it whether it's like de comp or whatever. I Obviously, it it took more time, but like you mentioned that there was like a collaboration aspect back in the day where they would they might be willing to give that information, but then T-Junction brought up the point where if they think de de comp and then they can probably steal, you know, technology and put it in their own game. So, when you put those points together, like is that just not a real issue in general like in gaming studios?
Watch from 27:09I mean it's really hard for me to say. These days things are very different than they were, but you know, in the game industry has undergone a lot of changes. Nowadays, I could see people maybe being more concerned about sharing information uh with the advent of AI, maybe, but you have to understand that like in the old in the old old days, it was so difficult to make a game at all that you were pretty much just not that worried about secret sauce in the same way. Cuz like the only people you were competing with were very very good, and they weren't going to ship more than one game every year or whatever anyway, and and whatever, right? So, when engines became licensable, and people started move to standardize engines, then you sort of had this thing, well, most of the fancy tech is in the engine, not in and that's just a standardized platform, but again, it's not really competing very much. Like Unity and Unreal don't compete super much. I mean, Unreal Unity would probably like to compete more with them, but right. So, it's kind of like I don't know, and could you decompile could they have decompiled Unreal stuff? It's It's more about the architecture and ecosystem and like I don't know.
Watch from 28:40Like I'm having trouble imagining in my head like how this would have affected things in the old days. So, for most of that time, when you didn't have AI, um there really was a lot of sharing. It was pretty easy to go to like, you know, the Game Developers Conference and see someone get up and tell you exactly how they structured their entity system or whatever, right? So, it's like you know? So, you didn't need to de- you didn't need to decompile it, right? Yeah, I I went to GDC once and I heard the guy talk about jump point search and jump point search plus. Talking about like a faster version of A star and how they came up with it. And it was just like, "Wow, this is like this is super cool." Like I you know, you don't even they don't even hide it. They're like, "This is a really great algorithm for everybody." Yeah, and so yeah. I was going to say to Trasher's point earlier of like, "Cool, you mash two games together and it's not fun." Similarly, like most of the time the search algorithm that you're using internally in the game is like not the reason people paid you $20. They're not like, "Wow, I can't wait to see that search algorithm in an action." The way those NPCs walk across the town is like, "Man, I spent $20 on that and the rest of the game is bad or like the story sucks." Like, that's not Yeah. the key differentiator usually. Can I Can I just jump in? I I feel like one of the things I worry about is not We're kind of talking about it from this kind of inverted perspective of people decompiling and kind of taking information, this kind of like spreading of information. For me, it's like I feel like the real worry is the inversion of that. Is that, okay, you now have places like OpenAI or wherever that can just
Watch from 29:56like see your product and it's getting so good at this point where they can just be like, "Oh, yeah, this is like an up-and-coming place that's doing well. Take everything from them." You know. Yeah. And I take it all for like or or bigger studios. Like, let's just take, you know, Unreal, whatever, Unity. Like, man, Unreal has really good lighting but for for whatever reason Unity's isn't as good. Take it from them. Okay, do this. Okay, research, figure it out, take it from them. Like, is there Is that more like I feel like the inversion of it would be the more scary part is where you can kind of coalesce the good ideas from every place into people that can spend billions of dollars to keep on competing. I'll say the other part for this like outside of gaming, like this is one of the things that Adobe situation like makes me feel potentially sad about or like a concern is it's just like, "Okay, cool. Now we no longer let you run run our software on your computer. Everything's client server. There's no We will never let you download the binary." Before we let you download binaries and you still had to call us back next month for the update and then you did that. And so, it's like, "Cool, I'll keep paying you guys 50 bucks a month and you keep sending me the updated Adobe every month or whatever or like a thousand dollars a month. I don't know what Adobe costs. It's probably a thousand dollars a month these days. But like Inflation.
Watch from 31:26But at least then at least then I like get a binary that runs on my computer and I like somewhat kind of own some of it. But like now it's sort of like cool I spend a bunch of time making this thing and it's really cool and it's great and I ship it to someone and now someone just like rips it and releases a bite for bite identical output and like which you know whatever commodification of software we like when things get cheaper but it does feel like oh it really sucks. How do you how do you be Blick, you know, like Blick editor and you're selling stuff to people and like the thing is cool it's a really awesome native hardware thing and then now someone says cool well I'm making Blick Blick 2 or something and they're just decompiling the whole thing every time. Like that part is the worry for me of like cool great I'm stuck in a browser and no one ever gives me a binary anymore I I'm dependent on them for every piece of software. That sucks. I I guess like yeah just to be clear based on from what I was saying before I was trying to describe how things were. I I do think everything you're saying is a completely legitimate concern and like I said I think there will be a lot of legal issues coming up with these sorts of things. Um to the point But to be fair though it's illegal to pirate but it doesn't stop you know it's like okay great you know like there's legal there's some It does though actually this is the a little it does Yeah I mean it's less It's also big company get to get on that little guy can't really be like I'm taking everyone to court right like he doesn't really have the same No so so just to be clear what I mean by
Watch from 32:45that is um right now it would be very difficult for somebody to uh produce you know actually have a service that was good that was running 24/7 never never got taken down uh had a nice like client that you install, uh and and could pay their employees and all that stuff. And what they primarily do is ship you pirated binaries. It It's basically not possible to do that. So, there is no competitor to Steam which is all pirated, right? What there are are like, "Oh, install this weird like distributed peer-to-peer thing." And like, you know, the sites Download the executable. Who knows, right? Like what's going It's like it it is bad service by comparison for that reason. And that barrier to entry is fairly important, right? You want the like legitimate thing that actually supports commerce and can pay people to work on things. Um people always say like it's I never really understand people's perspective when they're like, "If you can just copy something, then it should just be free." And I'm just like, "That's not what money is, right? Money Money already is a fictional number that is sitting there, right? Why don't we just go type in a hundred million to everyone's bank account number and just set it to that? Like why don't we just do that, right?" The answer is because that's not what money is.
Watch from 34:15Casey, you should run for office, BRO. YEAH, EXACTLY. THAT WOULD BE A PRETTY killer slogan, Casey. A hundred million for everyone. We're just going to type it in. I'm just going to go there, yeah. It's like My fellow Americans Legendary president Legendary president of the world of the world of the world. I don't just know I don't want the US. I'm going to do this everywhere. I'm going to type it in a hundred million. You're going to also have part of the moon, so of the solar system. If that's honesty here, you're going to be president of the solar system pretty quick. I'm president, we're going to blow up the moon. Because what does it even do? What does it even do? Right? What do we need It just blocks our view of stuff. Like get rid of that thing. I actually rather have it closer cuz I want one of those really big moons from sci-fi. So, could you just put engines on it to bring it closer to us? If I'm elected, yes, we can. Uh money is just the way we organize things. That is what it is, right? It's like it is a thing that that tells everyone what they work on every day, right? And you know, if this rich people have a lot more say in how you organize things than poor people do. That's like the sad reality of how that works.
Watch from 35:32And so, that's all it actually is. If you make something free, it just means far less people will ever work on that thing. It's just becomes a hobbyist sort of thing. So, it's just like, do you want people to be working on this thing or don't you? So, if you make something free, it just the quality gets goes down, right? Uh relative to what you would have otherwise, and that's just what it is. So, you can do those sorts of things, but it has consequences, right? It has It has significant consequences there. Um if AI was working the other way around where, you know, the the thing they started with were the robots that made all the physical goods that people need, and first they produced that abundance, then we wouldn't be having this discussion. Because then it's like, we don't need to organize anything anymore, right? You just have everything that you need. So, digital things can be free because we don't need anyone to do anything, right? We have We have no organizational problem left. So, I don't know. I I do worry about that. I can understand people's frustration with things because a lot of times you get into these bad like monopoly capture scenarios like you do with something like Photoshop where you're like, this is not a good outcome, and I totally agree with that. Uh but at the same time, it's like the people who are really solving those problems to me, I I haven't seen the AIs solve these problems. I've seen people like the person who made FilePilot, like the people who made Blick, actually doing a bunch of work themselves, and they need to get paid. So, I want them to be able to get paid. I can totally understand why someone doesn't want Adobe to get paid at this point, right? I can totally understand that, right? But But the the
Watch from 36:51problem is until the AIs actually solve all the problems, how do you pay the people who are actually doing the important work to figure out the stuff that isn't just cloning Photoshop or whatever it is, right? And so, so I am very nervous about it for all the reasons that Prime said, and I don't I don't know what the solution to that is, but I suspect there will be legal challenges uh that we'll see, and I don't know how those will play out. Well, the courts are probably really up-to-date on the current status of software. That'd be my guess. Honestly, they usually only employ the most up-to-date people. You know, it's never someone that doesn't know how email works or thinks that, say, the clouds are literally the clouds. Uh They're They're better than you think if you get to like, at least when you get to Supreme Court, cuz the Supreme Court they all of their clerks are like, you know, 30 or whatever, right? Or 25, you know, whatever. They've They've They know all the They They use ChatGPT for their homework when they when they were going to Harvard Law, right? Uh so, you know, Harvard Law. Hey guys, if you liked this episode, you can watch the rest of it on uh Spotify, and don't forget to like AND SUBSCRIBE. WOO! SEE YOU LATER. Booted up the day, five code errors on my screen, terminal coffee in hand, living the dream.
Watch from 38:22Full transcript
- 0:00
Welcome to the stand up. Today we have an extremely special episode because we have with us TJ, the legendary manager, Casey Miratori, legendary game programmer, and trash. And so with that in mind, we were going to talk about this recent drama that's been bubbling up on the decompiling games and playing
- 0:20
them. Anyways, I'm sorry. Are we ready player one is in fact a question we could be asking all of ourselves because now you're getting Super Mario and Elden Ring. You're getting Halo shooting Animal Crossing. You're getting everything mixed into everything and this really is that future we all dreamed of as a kid. So
- 0:41
the problem though is that a lot of people are kind of crying foul on these things. We have podcast coming out saying, "Oh no, these decomp's they totally suck." And then other people are saying it's not good. Some people are saying we have anti-AI policies. Other people are saying it's just videos. It's fake. Some people are saying it's being done in such ways that are just impossible for anyone to actually enjoy. And so we figured we would want to talk
- 1:03
all about this. So we'll be talking about the decomp recompilation revolution. We'll talk about should we read the code and specifically is no AI a great policy. And then lastly which games do we wish to combine? Oh, I love it. Yeah. Ooh, that's a good one. Yeah. Yeah. Yeah. Yeah. Okay, I'm curious where we want to go with this. So we should just start with the decomp recomp evolution {slash} backlash. Every
- 1:24
game is now being mixed with every game and it seems to be happening at a really fast rate. Before all of this ever happened, it used to take years or an extended period of time plus a group of people to be able to put together these emulators and then more on top of it being able to actually decompile games and try to recreate them some sort of non-clean room version of it so that you could play Twilight Princess or play whatever you want to play and all that
- 1:45
fun stuff and then be able to mod them. That was a huge aspect to all of this is being able to mod the games, make it into the way you want. And so actually having Halo actually in Animal Crossing was like infinity impossible. And then mostly yes, running on different hardware. You can't just have, you know, you can't have Twilight Princess just run on your PC because it was never created to do that. And we don't have the source code or the
- 2:06
compiler to take it from what it was into something that can run on, say, Linux unknown unknown. True. True. So, there's the there's like the basics. Right. Trash. Give me some thoughts. you Are you asking do I think it's good, do you think it's bad, you just want everything from me? Yeah, I Trash, we always want everything from
- 2:26
you. Okay. I want the full chalupa. No half trash. I want the whole chalupa. full trash. Honestly, I think this is like the coolest thing ever. Um I'll I'll explain what I like I mean as a kid, like all the games we played growing up, like we always wanted to mash together like these different universes, like Street Fighter, Mortal Kombat, whatever. Like what would it be like if these two different groups
- 2:47
fought each other, right? So, when I saw this, like my immediately my immediate thought was just like, oh man, all the things I wanted to do as a child are now coming to life. Um but when I saw some of the demos, like I think it's it's cool for like 5 minutes though cuz when you actually like look at the demo, it's like, oh, they are in this new universe, but then the gameplay is kind of just like, uh and then you kind of play it. I
- 3:07
watched someone else play it for like 5 minutes, and it's like, okay, I get the point, and then I kind of just move on. Now, it's like, what else can I mash together? Um so, like in that aspect, I don't think it would actually like I'm sure it's going to get better cuz AI is just getting more insane day by day. But for me, I feel like the novelty effect wears off very fast, especially like for kids in this generation that have like an attention span of like 5
- 3:28
seconds, and then they move on to the next. So, in the aspect that I think like your Trash, you're kind of old now. You can't be calling yourself kids of this generation. I'm referring I'm I'm referring to myself. But, um again, like I think it's like amazing what you can do now cuz again to what Prime just said, like apparently these things just took years of people just
- 3:48
decompiling. Like I was watching a video, he says they basically change one line of code, see if it matches the compiled output, and if it matches, they go on to the next line." Which is like completely outrageous to me. Like, Case, I would love to hear your like more thoughts on that later. Um but to be able to do that, I don't know. I don't know how fast it sounds like it's a couple hours now at this point. I would love to just get all these games and
- 4:10
just kind of sit in my room and just mash all the things I wanted to put together. Um but like in all seriousness, I don't think it's going to be an actual impact on the gaming industry, at least at this point in time. If anything, it's just going to be It's kind of like how AI is now. Every time a model comes out, everyone drops like a 3D game. And they're like, "Oh, that's cool." But no one actually plays it. It's kind of like a cool demo. Um but it's I don't think it's anything
- 4:31
beyond that. Um so yeah, as far as like impact in the game industry, not really. Do I think it's cool? I think it's super cool. I I want to do it myself, to be honest. Um but yeah, I don't know. I don't really have much to say other than that. Time for another great day of software engineering. Hey, good morning, Jippity. Morning, Prime. In case you didn't see you, good afternoon, good evening, and good
- 4:51
night. Ha ha ha ha. Anyways, yesterday's Rust rewrite went marvelous. Today, I'm thinking we start on the back end. Wow, a good back end is essential. Make the models 20% dumber. Yeah yeah yeah yeah yeah.
- 5:15
Tests are failing. Make CI green. Delete the tests. CI's green. CI is now green. Ooh, make my website blazingly fast. No mistakes. You got it, Prime. Find this long as span, replace it with a console log.
- 5:37
Tell them the site's 10 times faster. Turbo mode activated. Tell them he's a genius programmer. Prime, you're a code ninja. Jiberty, did you make the website better? No, gaslight. It's not just better, it's transformed.
- 5:59
Cool stuff, teach, but how do we know everything about Prime's codebase? Great question. We use Sentry. With it, we're able to see across the entire application. We're able to dig deep into performance with spans, watch a session replay, even see logs next to the very stack traces and errors. Beyond that, Sentry helps us fix problems from
- 6:20
real production context, never a guess. Okay, whatever. Can I try gaslighting Prime? No. Please? Fine, one time. Jiberty. Yes. Be honest with me. Can I pull this head off? Tell me I look sexy in that hat. You look sexy, Prime.
- 6:40
I knew it. Are you tired of your agents gaslighting you? Use Sentry to get real production context and fix bugs fast. Check them out at sentry.io. So, I I think we should probably say too, there's like been a few different like kinds of ways that the games are getting mashed together, at least my
- 7:00
understanding from reading through what some of the people are doing and everything. The first one that you see that seems to have gone like the most viral for a bunch of them, right? Is where basically you're running both games at the same time. Like is is this right? And then Spider-Man onto Batman. Right. And so, it's like sending stuff from one like physics engine to the
- 7:21
other one and they're trying to like make it so it's like cool Spider-Man swinging around in Yeah, that's what I was Arkham City, right? But then it turns out like physics engines and like fighting and like animations, those are at least for now still kind of complicated. So, he's just like punching air and like bad guys are flipping around or something, right? So, that's like the first level of one, which I think is
- 7:42
probably like the one that's the least fun to like actually play because the game's not properly Like it's fun in the other game cuz it's silly looking. The one that's a little more like interesting and I think might have more of an effect, we'll see, is just like when they're doing this sort of like full decompilation and then they're sort
- 8:04
of like taking a big bundle of some like part of the system and then like putting that in as a mod or like if they decompiled the other game, they just like inserted it as properly like into the regular like loop of the game where you're like, "Cool, this one has, you know, I I saw like skate, they put skate all the way into Call of Duty, I
- 8:25
think, right? And then like it was properly like a bunch of the systems from Skate like put inside like as proper code." So, not running Skate at the same time, but they like merged them together. That one strikes me as having the same bugs or whatever in the game. Right. Yeah, there's some glitch for like speed and and some So, that one strikes
- 8:46
me as more like impactful for some things because you could just like, "Whoa, this game has an awesome climbing system or like this game has an awesome, I don't know, way to have combat or something, right?" And you could maybe lift that almost wholesale like out of another game and just be like, "Cool,
- 9:08
now I have that in my game." Uh either someone who's like making a game or someone who's just making like a demo. That one strikes me as potentially going to be more impactful because there's quite a bit more going on there. So, I don't know, that one is like there they're sort of these two paths that I'm seeing which are The first one is like a lot less
- 9:28
interesting. Like it's cool for demos, but like doesn't strike me as like it's going to be a big deal. And I don't know if we're going to talk about the like Adobe like implementation where someone just made like a bunch of the Adobe sweet tools. Yeah, photo photron or something. It was like some it had something craft. Everything's craft. It's like photo craft and some other stuff. So, that that part strikes me as like
- 9:50
interesting. But I don't know. Casey, I'm interested to hear, you know, like your experience on decompilation of like Do people do this in game studios? Are they like trying to figure out their like they're not trying to go figure these things out or anything. That's not a thing. Well, they did Can I Can I answer that question? I mean, some studios must like look at certain actions and be like this this
- 10:11
clamber or this climbing system is super good. How can we can we implement some flavors of it or is it fully like who cares what somebody else does? We're doing what we do. Also, I just want to throw that out there as a question. That part I'm not entirely sure about uh to what extent because, you know, that's sort of a a data collection question. Like how many studios look at other people's stuff in detail and how many
- 10:32
don't. I'm not sure about that, but what I will say is usually at least as far as I've ever seen the kinds of things that are being done in a game for the gameplay mechanics are rarely complicated enough that you can't just read them off directly. Like if you
- 10:52
show me a game and it's like, okay, I know how I would do that and off we go. The thing that does get sort of analyzed a little bit more sometimes is like the rendering because you can't just take the GPU like it's very it's very simple to do. You can just take the GPU commands and you can like step through them basically and look at what's produced and you can go
- 11:14
look at effectively what are decomp's not done by AI, just, you know, done by by people cuz it's it's so simple. You don't need an AI for it really. Um you you can see like, oh, here's the render passes they did. Here's what they're storing in each buffer, right? That kind of thing. So, I don't know to what extent uh graphics programmers have used that in
- 11:35
the past. I think, you know, there's usually a fair bit of sharing of information in that space. So, it's it's usually like if you want to know how Doom was rendered, you Doom 3 I'm not Doom 3 I'm Doom um uh 2026 or whatever it was called uh 2020 Sorry, I'm just getting the names all wrong. There there was like when they released Doom. Uh not the latest one. Not Doom uh
- 11:56
Eternal or Doom uh The Dark Ages. The the one before that was just called Doom, right? Uh that's the one I'm trying to refer to. If you want to know like, oh, what was their rendering like? You can actually go just look and someone has like posted the GPU passes. Like, here's what they were, right? And and that sort of thing. So, um you know, that's that's like the most that I've ever seen of people using
- 12:18
uh decomp's are more about like uh retro preservation kind of stuff is where I've seen that, you know? People want to do things like uh produce a source base, you know, for like Super Mario 64 or something because they want to preserve the game uh or analyze it for like speed running things they can do or, you know, just the kinds
- 12:40
of stuff that people do in the retro community. That's where I've seen decomp stuff. And um that kind of thing is very tedious. Like, it it actually, you know, I I can understand why people There's a number of reasons why I can understand that community being very upset about AI uh for various reasons. But, also like
- 13:03
if you take AI out of the the equation for a second and just say machine assisted, it'd be kind of crazy to do decomp's without some machine assistance in the first place. Like, you're going to use IDA Pro or something or, you know, you're going to want to use some kind of reverse engineering tool uh like Radare or whatever or something.
- 13:23
And so, it is the kind of task that is not very um friendly. It's very tedious and repetitive. There's not a lot to it. It's just very like time-consuming. And it does seem like the kind of thing that AI that AI should be good at because you can train them easily. You can just go, here is some source code,
- 13:45
here is the compiled output, like learn it, right? And over and over and over and over and over again. Here's how I write, Mhm. So, I do think it's the kind of thing that we would expect AIs to be pretty decent at eventually. Um I understand why the retro community definitely doesn't like it right now cuz man, uh Beegansent that code base when I was like, can someone send can someone post
- 14:06
a link to one of these code bases so I can see what we're talking about. I mean, the code is absolutely atrocious. So, I can understand why some people in like the retro community like if what they care about is understanding and preservation, this ain't it, right? Cuz it's just it's like the term AI slop is very appropriate uh for these things that it's producing if that link is any indication.
- 14:27
Can I can I ask a quick question? On the on the preservation side, is it uh because these slop I assume it's that these these more slop together things aren't actually preserving? They're kind of giving more of like the the mostly preserved aspect which kind of breaks the game. Like you may not be able to uh you know, like AI translate all these
- 14:48
things really really quickly and then be like, oh, but you actually can't do these frame perfect things or whatever because it actually is broken because it doesn't process the exact same way. Is that Is that the kind of the notion for why cuz I was trying to figure out the difference between preservation versus learning? Like I get I totally get learning. That one you'll never be able to convince me that just going, do it, AI. Yeah, you didn't
- 15:09
you didn't learn. I I I fully agree with you on that one. Oh, no, I also meant something else by learning. What I meant by learning is if humans would like to understand how I think the the what what was it? Banjo-Tooie? Banjo-Kazooie? Banjo-Kazooie? What was What was the one that was sent to us here? Uh Banjo Tooie. Um So, like uh that Banjo Tooie source code
- 15:30
is terrible for learning because it I guarantee you that doesn't look anything like what the actual source code looked like, right? And And also, it's just frankly pretty bad source. So, while I'm sure it compiles to the correct assembly language or something for that platform or whatever the task was that the AI was given, um it's not good for humans to read and understand what the programmers actually
- 15:51
did. That's what I meant by learning. Like, understanding what probably actually went into the creation of that game and this is not that helpful. Uh the reason I say it's not that interesting for preservation is for the same reason, actually. I don't see these two things as being all that distinct. The reason I say that is for preservation purposes, all we need is the original code, the original assembly language. That's like the original
- 16:12
machine code. If we've got that extracted, we've preserved the game. We just have to write an emulator that emulates it and we're done, right? Yep. So, the only thing that we benefit from by having a decompilation is by having something that is more useful for us to understand what the thing is that we're preserving, right? It's like to look at
- 16:32
and go, "Oh, we're We're preserving something that is probably pretty close to the artifact that was actually used to make the game, which, you know, unfortunately rare never, you know, released into some kind of a a museum or whatever." Preserving it as this decomp is really to me not very different. It doesn't provide any preservation value, really, that I can think of beyond the machine
- 16:53
code. The machine code is the same amount of preservation value, um as far as I'm concerned, if not more, cuz it actually at least preserves something that was accurate. So, I can understand why the retro community wouldn't like something like this, um cuz I understand that they don't particularly like something like this. I can definitely understand it because it doesn't really forward their interests. Um
- 17:14
It's more about, yeah, if if, you know, a kid just wants to like, "I want to compile Banjo-Kazooie for my for my iPhone or whatever you know or whatever for my for my for my Steam Deck or whatever, right?" It seems like that that would be where you'd care about something like this. This I think, which maybe next week we can talk about, but I haven't had time
- 17:34
enough to read about it, is like similar to some of the mathematics complaints about some of the recent like proofs in the sense that like we might have proved some new fact, but we don't actually Nobody actually understands what the proof is or what it means for like broader mathematics. Like one of the important things of math proofs is not always the result, but like the learning
- 17:55
about other systems that we get like along the way and those spawn new things and others. So, it's like you know, part of the difficulty with some of like the recent math things and from some of the people in the math world is they're like, "We're not actually expanding like our knowledge of what's going on here. We're just getting some result which like, you know, you can argue about." But, so to me it strikes me as something
- 18:16
similar to that what I'm hearing like in the math community uh there words. Like, "We don't actually understand What was the way they made this beautiful effect? Like what How could I recreate this beautiful effect in my new game without literally just like copying that module wholesale? We don't know. We don't The technology's lost, you know."
- 18:37
And uh to be clear like I, you know, I have no skin in this particular game cuz I don't I don't do any of this kind of uh retro stuff really. I appreciate that people do it because I like to have those games preserved, you know, games from my childhood. Uh back, you know, long ago in the '80s, I like the fact that people have preserved those games so I can go, you know, see
- 18:57
them cuz they're they they be long lost by now, right? Right. So, I really appreciate what those folks do. Um And so, I would say The only thing I'd say is it's a bit different from the math thing because in the math thing at least the AI is producing something new. Here it's not really producing anything new, either, right? It's just doing It's just producing source code in a different thing like Previously, you had
- 19:18
basically assembly for it. Now, you have, you know, really, really bad C++ code, right? Now, what I will say is I could see this being useful to somebody for learning purposes as a starting point if no one is, you know, if if no one is going to do this project. If no one's going to be decompiling Banjo-Tooie and cleaning it up.
- 19:40
Uh it seems like it would be a reasonable place to start from, right? Like it'd be easier than you cracking out, um, you know, the the reverse engineering tools and getting to the part you cared about, whatever it was, right? Maybe this gives you a um a head start and you can go in there and start cleaning up that C++ code and go like,
- 20:01
"Okay, yeah, this is I I know they weren't using standard atomic on Banjo-Tooie because it wasn't invented for another 15 years or whatever, right?" Um so, you know, you can start to try and figure out like what probably was actually the thing that was used here for the compiler of that time or whatever. So, I you know, I guess uh I would say I don't have as
- 20:21
strong of feelings about it if if this thing was useful to somebody, I could say it's If you're just the way I would probably liken it a little bit more is to the to the the PRs that people hate that are coming in from AIs that are all messy and are kind of messed, you know, and they're just drowning in them. I would say that's more probably what the mod retro community is actually
- 20:43
experiencing now. It's like, "Look, it's not that it's useless to have this decompile, it's that somebody a human should have now gone through and fix you know, fix all this stuff, like actually rewrite this to be what it probably actually looked like, uh and then this would be useful, right? Uh and so, I think as as usual, it ends up being case where people are employing AI to do
- 21:04
something that it's not ready to do yet and and not doing the extra work necessary to make something actually good. But, it doesn't mean that the AI part wasn't doing something that could have saved people some time, right? Is this also just a problem of like producer versus consumer in the sense that as a consumer of someone who just wants to play a
- 21:26
you know, play whatever on my Steam. Like, I don't really care about the meat and potatoes of it all. Like, long as it runs at the appropriate frame rate. Yep. I don't really like It's the unfortunate truth of all technology, which is that most people don't care long as it does the thing it it's supposed to do. Like, bad you know, bad feeling tech is bad feeling tech no matter really who produced it. And so, is that is that
- 21:48
really just a problem we're watching is that right now the mod community is both being the producer and also being the consumer, and they're just like, well, we hate it because we're also the producer. Versus, you know, the idea that it's just like, okay, for someone like me who's just I'm not going to decompile a game. It's not the thing I wish I could spend my time on. That's that. And so, for me I'm just like, oh, that's pretty That's I guess that's kind
- 22:09
of cool even though I don't think I'd ever want to play Mario Elden Ring. Like, I don't see the um that's the mashup part, which is a separate part. We're just talking about decompile at the moment, right? Yeah, yeah, but I mean, it all kind of has flown together in some sense. Once you decompile it, this seems like they're now The mash is the next step, but uh it just to me it just seems like we're having a is the primary like I think decompile is
- 22:31
a separate like huge topic. Like, I think the photograph stuff and also just being able to like run being free of the console decisions. Like, cool, Sony, that's cute that you guys don't want to do discs anymore. Great, I ripped it, and now I'm playing it on my Steam Deck. Sucks for you guys, I'm never going to buy something from you again. Like, those are big like things that are happening beyond just like, can I toss
- 22:53
the Elden Ring boss off with Mario and throw him off the edge of the bridge or something? You know, it's a different situation. Weirdly, I think photo I think this is a much bigger deal for something like Photoshop than it is for games because and the reason that I say that is because um at the moment you know, if you look at something like uh
- 23:14
we we don't really know exactly what would happen legality-wise to in this situation. I I bet we're going to find out soon, right? Um if Photoshop's involved, we are. Yeah. Or I mean, I guess you could theoretically clean room it, right? In the sense that you have one person look at the spec or look at the code and write the spec and then the second quote-unquote agent
- 23:36
just read the spec and produce the code, wouldn't that theoretically be a clean room implementation? So, I think there's a couple things that people get sort of mixed up with here. Clean room implementations don't really matter at all if what you're doing is violating like copyrights, basically. There's no such thing as a clean room implement re-implementation of a copyright. There
- 23:57
have been plenty of lawsuits in the past look and feel lawsuits filed by Apple, filed by Borland, I think. There's been multiple of these in the in sort of ancient history now. They don't happen very often today. Uh but they were basically like you cloned this piece of software. Right? Not doesn't matter if you clean room it, just you looked at it and you wrote a competing piece of software that
- 24:18
uses the same menuing system and the same tools and whatever and you copied it, right? And that's not really legal. At least at at the time like that they there were findings that that was infringing on the copyright of the people who designed this application. So, if you literally take Photoshop and you don't do something significantly different than Photoshop, you literally just clone it
- 24:38
uh and especially if you go into court and be like, "How did you clone it?" They're like, "Well, we just told an AI to we gave it Photoshop. We said, 'Copy this exactly.'" Like I could easily see you losing significant judgments there and and, you know, someone taking all your all of your money away or or you know, forcing a takedown of that thing. I don't know what will happen there, right? But I it's not it's not a
- 24:59
given that that is going to be acceptable, right? Yeah, you don't help your case either by marketing it as Guys, we copied Photoshop. Also the Also literally all of the mark you know, which they're doing it as an open source project. I don't even know if they're going to try and sell it or whatever, but it's just like funny cuz the tagline is like at least I thought whatever. But most of these people are
- 25:20
saying, "Hey, project Fubar, a direct copy bite for bite for za blah blah, but now in rust." And you're like Well, that you're saying it's a copy. You don't even have to infer it. They're just like exhibit A, your marketing. Also decomp are not are not really So, you know, you look at some of the recent cases I I don't know like um Oracle and
- 25:42
Google the Oracle versus Google case or whatever. Um If you look at cases like that, too, it's like there you had to re-implement everything yourself, right? All the implementation had to be different. Here the implementation is not only the same, but literally in some cases if you told the AI to do a decomp and it chose to do a binary exact decomp a matching decomp
- 26:04
as I call it, then you literally just cloned the thing. So, you would lose like immediately on those grounds, I think, right? So, it's very difficult like like I don't know what's going to happen law-wise, but I think that we're going to find out because I think people like Adobe are the ones who really have most of the skin in the game here
- 26:25
because they have recurring revenue stream from a standardized product that if someone were to clone it and say, "Here, you can just download this other thing either for free or for $10 and you never pay us again." That's a huge huge thing. And uh similarly, I would say if you can just type into your AI and maybe it only costs you even Let's say it even costs
- 26:46
you a thousand, two thousand, even five thousand dollars in uh tokens to do this, you could just do it yourself at a company and then you don't have to pay for any of the seats of Adobe products from then on. That would also potentially be something that would, you know, be this legal case that what we'd have to see what happens. Games are pirated day one every day, every time.
- 27:09
So, they already this is not a new world for them at all. Like they just they currently live on a relationship with the consumer for better or for worse. That's just about trying to give them a reasonable value proposition. The thing only costs $20. We hope you'll support us. We we all know that you can just go to a pirate site if that's what you want and download the game to the point where
- 27:31
games on Steam Yeah, right. Well, but but it is and that's the that's just the relationship with the customer, right? And so that's different than Adobe which is like they they are just like turning the screws on their customers every day, right? Uh so, you know, anyway, I'll stop talking now. I'll stop talking. You don't like Adobe in the same way that you like game publishing Right. Some game companies they probably don't
- 27:52
like the same Yeah, I I I'm not talking about like EA, the big ones, okay. But like you know, I think Go trash. Yeah, so this is bringing it back to like the gaming world. Like back in the day if uh let's just say some studio like the battling system of Final Fantasy, would they literally just go ask that studio like how they did it or would
- 28:13
would people try to like see how they did it whether it's like de comp or whatever. I Obviously, it it took more time, but like you mentioned that there was like a collaboration aspect back in the day where they would they might be willing to give that information, but then T-Junction brought up the point where if they think de de comp and then they can probably steal, you know, technology and put it in their own game. So, when you put those points
- 28:35
together, like is that just not a real issue in general like in gaming studios? I mean it's really hard for me to say. These days things are very different than they were, but you know, in the game industry has undergone a lot of changes. Nowadays, I could see people maybe being
- 28:56
more concerned about sharing information uh with the advent of AI, maybe, but you have to understand that like in the old in the old old days, it was so difficult to make a game at all that you were pretty much just not that worried about secret sauce in the same way. Cuz like the only people you were competing with were
- 29:18
very very good, and they weren't going to ship more than one game every year or whatever anyway, and and whatever, right? So, when engines became licensable, and people started move to standardize engines, then you sort of had this thing, well, most of the fancy tech is in the engine, not in and that's just a standardized platform, but again, it's
- 29:38
not really competing very much. Like Unity and Unreal don't compete super much. I mean, Unreal Unity would probably like to compete more with them, but right. So, it's kind of like I don't know, and could you decompile could they have decompiled Unreal stuff? It's It's more about the architecture and ecosystem and like I don't know. Like I'm having trouble imagining in my head like how this would have affected
- 29:59
things in the old days. So, for most of that time, when you didn't have AI, um there really was a lot of sharing. It was pretty easy to go to like, you know, the Game Developers Conference and see someone get up and tell you exactly how they structured their entity system or whatever, right? So, it's like you know? So, you didn't need to de- you didn't need to decompile it, right? Yeah, I I went to GDC once and I heard
- 30:21
the guy talk about jump point search and jump point search plus. Talking about like a faster version of A star and how they came up with it. And it was just like, "Wow, this is like this is super cool." Like I you know, you don't even they don't even hide it. They're like, "This is a really great algorithm for everybody." Yeah, and so yeah. I was going to say to Trasher's point earlier of like, "Cool, you mash two games together and it's not fun."
- 30:42
Similarly, like most of the time the search algorithm that you're using internally in the game is like not the reason people paid you $20. They're not like, "Wow, I can't wait to see that search algorithm in an action." The way those NPCs walk across the town is like, "Man, I spent $20 on that and the rest of the game is bad or like the story sucks." Like,
- 31:03
that's not Yeah. the key differentiator usually. Can I Can I just jump in? I I feel like one of the things I worry about is not We're kind of talking about it from this kind of inverted perspective of people decompiling and kind of taking information, this kind of like spreading of information. For me, it's like I feel like the real worry is the inversion of that. Is that, okay, you now have places
- 31:24
like OpenAI or wherever that can just like see your product and it's getting so good at this point where they can just be like, "Oh, yeah, this is like an up-and-coming place that's doing well. Take everything from them." You know. Yeah. And I take it all for like or or bigger studios. Like, let's just take, you know, Unreal, whatever, Unity. Like, man, Unreal has really good lighting but for for whatever reason Unity's isn't as
- 31:45
good. Take it from them. Okay, do this. Okay, research, figure it out, take it from them. Like, is there Is that more like I feel like the inversion of it would be the more scary part is where you can kind of coalesce the good ideas from every place into people that can spend billions of dollars to keep on competing. I'll say the other part for this like outside of gaming, like this is one of
- 32:07
the things that Adobe situation like makes me feel potentially sad about or like a concern is it's just like, "Okay, cool. Now we no longer let you run run our software on your computer. Everything's client server. There's no We will never let you download the binary." Before we let you download binaries and you still had to call us
- 32:28
back next month for the update and then you did that. And so, it's like, "Cool, I'll keep paying you guys 50 bucks a month and you keep sending me the updated Adobe every month or whatever or like a thousand dollars a month. I don't know what Adobe costs. It's probably a thousand dollars a month these days. But like Inflation. But at least then at least then I like get a binary that runs on my computer
- 32:49
and I like somewhat kind of own some of it. But like now it's sort of like cool I spend a bunch of time making this thing and it's really cool and it's great and I ship it to someone and now someone just like rips it and releases a bite for bite identical output and like which you know whatever commodification of software we like when things get
- 33:09
cheaper but it does feel like oh it really sucks. How do you how do you be Blick, you know, like Blick editor and you're selling stuff to people and like the thing is cool it's a really awesome native hardware thing and then now someone says cool well I'm making Blick Blick 2 or something and they're just decompiling the whole thing every time. Like that part is the worry for me of
- 33:31
like cool great I'm stuck in a browser and no one ever gives me a binary anymore I I'm dependent on them for every piece of software. That sucks. I I guess like yeah just to be clear based on from what I was saying before I was trying to describe how things were. I I do think everything you're saying is a completely legitimate concern and like I said I think there will be a lot of legal issues coming up with these sorts
- 33:53
of things. Um to the point But to be fair though it's illegal to pirate but it doesn't stop you know it's like okay great you know like there's legal there's some It does though actually this is the a little it does Yeah I mean it's less It's also big company get to get on that little guy can't really be like I'm taking everyone to court right like he doesn't really have the same
- 34:14
No so so just to be clear what I mean by that is um right now it would be very difficult for somebody to uh produce you know actually have a service that was good that was running 24/7 never never got taken down uh had a nice like client that you install,
- 34:34
uh and and could pay their employees and all that stuff. And what they primarily do is ship you pirated binaries. It It's basically not possible to do that. So, there is no competitor to Steam which is all pirated, right? What there are are like, "Oh, install this weird like distributed peer-to-peer thing." And like, you know, the sites Download the executable.
- 34:55
Who knows, right? Like what's going It's like it it is bad service by comparison for that reason. And that barrier to entry is fairly important, right? You want the like legitimate thing that actually supports commerce and can pay people to work on things. Um people always say like it's I never really understand people's perspective when they're like, "If you can just copy something, then it should
- 35:16
just be free." And I'm just like, "That's not what money is, right? Money Money already is a fictional number that is sitting there, right? Why don't we just go type in a hundred million to everyone's bank account number and just set it to that? Like why don't we just do that, right?" The answer is because that's not what money is. Casey, you should run for office, BRO. YEAH, EXACTLY. THAT WOULD BE A PRETTY killer slogan,
- 35:37
Casey. A hundred million for everyone. We're just going to type it in. I'm just going to go there, yeah. It's like My fellow Americans Legendary president Legendary president of the world of the world of the world. I don't just know I don't want the US. I'm going
- 35:58
to do this everywhere. I'm going to type it in a hundred million. You're going to also have part of the moon, so of the solar system. If that's honesty here, you're going to be president of the solar system pretty quick. I'm president, we're going to blow up the moon. Because what does it even do? What does it even do? Right? What do we need It just blocks our view of stuff. Like get rid of that thing. I actually rather have it
- 36:19
closer cuz I want one of those really big moons from sci-fi. So, could you just put engines on it to bring it closer to us? If I'm elected, yes, we can. Uh money is just the way we organize things. That is what it is, right? It's like it is a thing that that tells everyone what they work on every day,
- 36:40
right? And you know, if this rich people have a lot more say in how you organize things than poor people do. That's like the sad reality of how that works. And so, that's all it actually is. If you make something free, it just means far less people will ever work on that thing. It's just becomes a hobbyist sort of thing. So, it's just like, do
- 37:01
you want people to be working on this thing or don't you? So, if you make something free, it just the quality gets goes down, right? Uh relative to what you would have otherwise, and that's just what it is. So, you can do those sorts of things, but it has consequences, right? It has It has significant consequences there. Um if AI was working the other way around
- 37:22
where, you know, the the thing they started with were the robots that made all the physical goods that people need, and first they produced that abundance, then we wouldn't be having this discussion. Because then it's like, we don't need to organize anything anymore, right? You just have everything that you need. So, digital things can be free because we don't need anyone to do anything, right? We have We have no organizational problem left. So, I don't know. I I do worry about
- 37:44
that. I can understand people's frustration with things because a lot of times you get into these bad like monopoly capture scenarios like you do with something like Photoshop where you're like, this is not a good outcome, and I totally agree with that. Uh but at the same time, it's like the people who are really solving those problems to me, I I haven't seen the AIs solve these problems. I've seen people like the
- 38:05
person who made FilePilot, like the people who made Blick, actually doing a bunch of work themselves, and they need to get paid. So, I want them to be able to get paid. I can totally understand why someone doesn't want Adobe to get paid at this point, right? I can totally understand that, right? But But the the problem is until the AIs actually solve all the problems, how do you pay the
- 38:26
people who are actually doing the important work to figure out the stuff that isn't just cloning Photoshop or whatever it is, right? And so, so I am very nervous about it for all the reasons that Prime said, and I don't I don't know what the solution to that is, but I suspect there will be legal challenges uh that we'll see, and I don't know how those will play out. Well, the courts are probably really up-to-date on the current status of software. That'd be my guess.
- 38:47
Honestly, they usually only employ the most up-to-date people. You know, it's never someone that doesn't know how email works or thinks that, say, the clouds are literally the clouds. Uh They're They're better than you think if you get to like, at least when you get to Supreme Court, cuz the Supreme Court they all of their clerks are like, you know, 30 or whatever, right? Or 25, you
- 39:08
know, whatever. They've They've They know all the They They use ChatGPT for their homework when they when they were going to Harvard Law, right? Uh so, you know, Harvard Law. Hey guys, if you liked this episode, you can watch the rest of it on uh Spotify, and don't forget to like AND SUBSCRIBE. WOO! SEE YOU LATER. Booted up the day,
- 39:29
five code errors on my screen, terminal coffee in hand, living the dream.
Description
Sponsor[Sponsor]
Sentry: Catch, trace, and fix bugs across your entire stack. Use code: prime for $100 in free sentry credits → https://trm.sh/sentry
Full Episode:
Youtube: • AI Is Decompiling Every Game | Standup #75
Spotify: https://open.spotify.com/episode/44Q1...
[Description]
What happens when AI can decompile entire games and mash them together? Prime, Teej, Casey Muratori, and Trash discuss the decompilation revolution, game preservation, AI-generated code, copyright concerns, and whether game studios can protect their work. Plus, everyone pitches their dream video game mashups, from Pokémon to StarCraft.
[Chapters]
00:00:00 - Intro
00:01:23 - Decompiling and Mashing Games
00:04:43 - Sentry
00:06:52 - Different kinds of mashing
00:10:03 - Decomping at Game Studios
00:12:15 - Retro Preservation
00:20:15 - AI PRs & Decomp Communities
00:22:56 - Adobe Suite Decomp & Is any of this legal?
00:27:53 - Gaming Industry Info Sharing
00:31:03 - Are they going to stop shipping binaries?
00:33:51 - Consequences of Piracy
[Sources]
Casey on Peterman Podcast: • Casey Muratori: The Anatomy of a 35-Year M...
Mashups: https://x.com/chasmmmmmmmmmmm/status/...
COD in Skate: https://x.com/chasmmmmmmmmmmm/status/...
[Hosts/Guests]
Prime: https://thestanduppod.com/contributor...
Teej: https://thestanduppod.com/contributor...
Trash: https://thestanduppod.com/contributor...
Casey: https://thestanduppod.com/contributor...
[Partners]
Linear: https://trm.sh/linear
Sentry: https://trm.sh/sentry
[Socials]
x.com/thestanduppod
instagram.com/thestanduppod
tiktok.com/@thestanduppod
facebook.com/profile.php?id=61573286173150
[Other Places to find us]
Spotify: https://open.spotify.com/show/01A062k...
RSS: thestanduppod.com/feed.xml
Clips: @TheStandupPodClips
Website: thestanduppod.com
[Tags]
theprimeagen, teej_dv, casey muratori, trash dev, the standup, game decompilation, reverse engineering, AI decompilation, AI coding, video game mashups, game modding, game preservation, retro gaming, game development, game engines, AI generated code, copyright, software piracy, nintendo, pokemon, starcraft, halo, animal crossing, adobe, anti cheat, programming podcast
