Rain World: An AI Post-Mortem | AI and Games Conference 2024 transcript

[Music] so rain World an AI retrospective so first we'll talk a little bit just about ourselves to give you some context y who are you that about some set up doesn't it you weren't expecting a comedic performance but get uh yeah I am J jobson I studied graphic design in school but on the side of school I was constantly tinkering with video games and uh this made it so that I ended up like not finishing my examination work and I had to do it again and if finally I got my diploma but by then I was doing video games full-time and that's what I I have been doing since then this this seems like a classic video game story you were doing your studies and then video games derailed it all yeah derailed it all but uh in a in a in a good direction I think remains to be seen amazed to be see and my name is James theion and when I met your I was actually just doing video game soundtracks and music uh with Lydia my partner in the audience with us now um and I saw this amazing uh concept video game concept on the tig Source forums which if you're super old like us and slowly turning into dust you may remember from the indie game scene it's where uh games such as Minecraft came out of Fez a lot of these classic Golden Age indie games uh and he had this amazing uh concept that I just had this vision of like how it would sound and I just knew I had to be a part of it so I contacted him and that was how long ago now yeah that's 12 years ago so be be wary of who you contact on the internet you never know you never know so okay let's see world what fate a slug cat pre and [Music] Predator as the old world died a new ecosystem evolves in a land where even rain can kill each creature struggles for its own Survival each Beast has a place on the food chain which it seeks to deny and through this ruined landscape strange monstrosities and ancient Mysteries slug cat Journey Salon it is a search for shelter a search for answers ANW search for Hope what fate the slug [Music] cat they they did reach out to the real wner Herzog uh for the voice over but he declined I don't think our budget was large enough we Swang we Swang for it for sure so what is rain world so in Rain world play as this uh this little lost creature a slug cat that's lost in this huge incomprehensible alien world and what we wanted to do was we wanted to create an experience where sort you're sort of medium low on the food chain you're you you're a hunter of like small game you're like a little mouse uh but there's plenty of other creatures that are hunting you that are larger that you're sort of scurrying from your feret level exactly and we wanted to create a world where you didn't really level up in the traditional sense but it's more that you learned about your environment and you learned about ecosystem of the creatures around you and using that knowledge you were sort of empowered to carve out a little niche for yourself in this world so the important uh aspect is that it's an ecosystem simulation basically all of the all of your resources but also all of your obstacles are other creatures and um in a way it is like a metrov Vania except like James just said you don't uh level up by gaining new abil ities you level up by gaining new knowledge I guess about how these creatures work now what does that mean in a practical sense what are our creative intentions in AI how does it relate to what we're talking about here yeah so we wanted to have a diverse set of creatures and behaviors and here you can see some stuff going on there's like the the lizards hunting the the main character and then in this uh animation here you can see the lizards like why are they not taunting me they're running away right Ah that's why there's a there's a bigger there's a bigger creature and on the right you can see um just a nice uh nice pretty looking animation but basically this this was the intention we we wanted to have an ecosystem simulation where the game play was about playing the creatures off of each other so if you know how they work you can maybe have the creatures interact in ways that are beneficial uh to you and another thing that was very important to us was that we wanted these creatures to be creatures not enemies or obstacles like we wanted the feeling that they exist in the world on their own right they're not there to block your path they certainly will block your path but they're not there to block your path they're there to like do their own thing uh so yeah all of this makes AI a very very Central game mechanic to rain World absolutely and kind of the I I think the what all the these ingredients sort of add up to is you know when you're younger and you're playing a video game for the first time and you don't really have an understanding of like the depth of what video game AI does so every movement of every creature every bad guy you sort of project into you think like oh they're thinking about this or they're like uh they're planning that or like this movement of this arm means it's mad or something like that like all these things that may or may not be actually there but you project into them I think what we wanted to kind of do was create this situation where you actually aren't really sure like the depth kind of keeps going of these interactivities so you're never really sure what is ha what like if what is happening is intentional or if it's resultant or if it's narrative or if it's just circumstantial that kind of fuzziness that is like reality is the lard angry or is the a star flipping out exactly and most often will never tell uh development how to build a believable creature and the joke here is that the creature doesn't look believable it looks crazy right um yeah let's uh let's move on to the next slide it's down I already pressed it you can just just up and down or like ambiguous on this one I thought so yeah so Locomotion AI was a big part of brain world uh this is like AI system that are they have one foot into animation right it's it's about like how something moves basically a lot of like limb placement and this sort of stuff and this it was actually the main subject of our GDC talk from like 10 million 10 million years ago 10 million years ago the late Cretaceous um so if you want more in depth if you want about an hour of just this slide you can go all sorts of stuff but we're going to focus on slightly different things today was the was the idea uh rain World also has social AI as you could call it like a reputation system you may gain or lose reputation this is also not going to be the focus of this talk and instead we're going to move on to uh AI for stealth which I think is an important uh important aspect of rain world because basically your character is extremely underpowered you have basically no efficient way of fighting back in Rain world and this is intentional it's frustrating but it's intentional this was how we wanted the game to be and instead you're supposed to rely on wits and stealth so we spent a lot of time trying to make stealth AI work now what are they seeing here with all this rainbow Aura uh this is the vision cone of a creature and as you can see maybe you guys can see it's a little bit dark but there is some water at the bottom of the screen right and you can see how the vision cone is like impaired by the water so it can see underwater but it sees like less far and less well uh underwater another important aspect of the uh stealth AI was that okay so think of it like this if you have something that's too stupid you cannot fool it and you cannot get satisfaction out of fooling it like an AI That's just like a box that takes out the direction runs at you it's too stupid to play with you need you need it to like do some little internal modeling of the world in order to fool it in a satisfying way so uh one one aspect of that that is what you can see here which is like uh a modeling of possible positions for a Target so this lizard as it is it's seeing the player creature and then every every second during which it's not seeing the player creature you have sort of like ghosts of where the player creature might be that are dispersing and then maybe if it hears a sound it like assigns a higher probability to the ghost that was closest to the sound and while it's not seeing the player it will just sort of like hunt down these ghosts one by one and check them um and an aspect that you might think of since this is sort of he's talking about sound based is that one of the solutions for creatures is often you can use uh a rock or a pebble and you can throw it somewhere to create another sound somewhere else which lures the creature over into that area this will come up later this is why I'm mentioning it yeah uh and uh an important aspect with this is that the AI very rarely I tried to make it never but it is very rarely quer is the actual physics simulation like the AI doesn't the AI is not able to just reach into the game code and grab your position instead the AI is able to ask its internal model about where it believes you to be and then it follows that this one where are we at this one cool yeah AI has a static terrain this means we are able to bake data um and this is an incredibly powerful resource for AI uh we found out uh basically you can do you can think about a lot of the stuff that the AI might want to do and instead of doing it on the Fly which could be very expensive you can do it beforehand and just save it into the the map file obviously this is not going to work if you have a lot of moving moving terrain and stuff like that but rain world is very static so it was a a powerful resource for us to draw on uh what you can see here on the left is a visibility map so that's essentially like an ambient oclusion kind of implementation like for each tile how many other tiles is visible from that tile uh this you can imagine that you can use to kind of skew different behaviors like if you want a sneaky Behavior where you're kind of like lurking in the shadows you can go for the Le visible tiles or maybe you have a predator that wants to survey the the uh room from above then that one could search out the more visible tiles on the right there's is a digra map never never understood how to pronounce it that D JRA digra map um of the distances to one of the doors and this is also like if you if you can pre-calculate this and then bake it in into in a like fairly compressed way it's always very uh handy to be able to just grab this data while doing AI code we also had pathfinding data some signed distance fields from the terrain Etc ET a bunch of different layers of this stuff next up yeah let's do it you you have to press yours too sorry um so one aspect of the rain World AI that uh I haven't talked about too much before is that it's actually module based uh so basically the design challenge I had with rain world was that I needed to create AI for many different creatures and I didn't that want them to just be different in terms of stats I wanted them to feel fundamentally different like be able to do different things and like that um however AI code is very uh labor intensive to write it's like a lot of work so that makes for a dilemma right because either you can code an AI system and then like maybe give it some different stats and hope that you get different behaviors out of it or you can like code AI from scratch for every single creature and that's going to be uh a huge lot of work the solution I came upon was to use modules so a module could be for example like the I had a pathf finding module I had like different trackers a module for tracking other creatures a module for tracking like the rain timing all of the creatures in Rain World need to go back to their uh hiding spots when it rains uh one that like tracks the health State etc etc and these modules are little objectoriented instances and I just uh created a bunch of them and then for the different AI uh implementations for the different creatures I could spawn in or like uh make different instances of these modules right and that enables me to have a lot of code in common between the creatures but that code is I can pick and choose what of that code to use for each each case well and also importantly sort of to uh sum it up a creature's Behavior a creature's individual behavior is made up of all these different items and each individual creature sort of has uh its own role of the stats for all of these so basically each creature instance winds up being an individual personality in terms of behavior because if it's all of these different modules are at different stats for each creature and as a result we get creatures that are more aggressive creatures that are less aggressive creatures that want to seek out their den early due to the strength of the rain trackers creatures that are less uh inclined to do that so as a result you encounter a creature say a lizard that maybe superficially looks somewhat similar but they have completely different internal processing as a result so kind of every time you're experiencing a creature you have a fresh a fresh AI a fresh instance yeah it's it's a it's like a hierarchy right like the different creatures like the lizard and the culture Etc they have fundamentally different AI but then there's also some little variation on the individual level uh as well and this can be weighted into this uh this module system I should also say the modules are more or less self-contained like they might relay they might relay on another one uh but they should be like closed little little systems um and then the way these modules work to inform Behavior was mostly as like a uh utility function right like most of them are able to be queried for what is their current utility so if you have a threat tracker for example I have if I have a threat tracker and James is a predator I uh the threat tracker will keep track of how close yames is to me and if he's close to me it will report back a high utility if he's far from me it will report back a low utility right so then the uh top level a system can just ask all all of these modules for their current utility and basically pick the one with the highest utility and act on that Behavior Uh and in this instance you see there's a creature and it's a little bit difficult to see but there's like a noise tracker so when the player is jumping that's making noises and the noise tracker is actually a behavior for like going to investigate a noise specifically that you don't know the source of um so this creature here is hearing the hearing the noise and then you can see how the utility of the noise tracker is creeping up also you will see that the utility of the noise tracker can only go so far and this is because the different modules can be weigh weighed to uh be differently important right like obviously if you're dying and you're trying to save yourself that's more important than checking out the noise and the next room over this is some visualization of a threat tracker module so it's a little bit difficult to see behind all of the visualization but you you may see a lizard and like a large vulture as we call them and this is the lizard trying to flee from the from the vulture and you can see how each each tile is associated with a D value and here you can actually also see a little bit of the baked data I talked about earlier because you can see like inside this narrow area it's kind of blue even if it's close to the predator and that's because uh the signed distance field of the terrain can be uh can be sampled and that makes it so that the lizard nose that's inside here it's actually too narrow for the big guy to get in so it's safe even if it's close to the predator um the these these modules they basically give simple answers to complex questions and they perform complex actions from simple commands wow that was easier to write down than to say yeah it was a lot easier to write write down than to say but okay I I'll explain that a little bit so basically the threat tracker maybe it needs to track like 10 different threatening creatures in the room it needs to figure out how close they are is there is there a wall behind them can can the threatening creature get to me or am I in like too narrow a space all of this very complex stuff but it takes all of this and it basically just crunches it down to a single number 0 to one which is how scared should I be right now uh so this whole module has a lot of internal workings but all of it just comes down to like a very simple answer how scared am I and then on the other hand what this module can inform behavior and perform behavior um very complex Behavior like trying to flee or whatever uh trying to find like a good fleeing route that takes you to the exit without exposing you to Too Much danger like a very complex task but the overarching AI doesn't need to worry about that it basically just tells you it tells the threat module okay it's your turn like do go go do your thing right so this is what I mean by simple answers to complex questions or complex actions from simple commands basically the module has a very simple interface and then all of the complexity is hidden away inside so this makes it so that the overarching AI the the the script that's actually owning the the modules it can focus on uh it can focus on the personality of that creature and it doesn't need to worry about all the complexities of like path finding and tracking and all of this because it just has a couple of modules and it can ask those modules simple questions and give them simple commands um the presence and absence of different modules will make for very different behaviors for different creatures uh you when you are designing this creature you can choose which modules to include and uh also this this allows for like different behaviors but also reuse of code like I talked about earlier here's an example a little bit about how something like this might something like this might look and this is very simplified but uh imagine that you have two creatures you have like a scared creature and like an angry creature and these ones they actually implement the same modules but what you can see is that one of them is uh using the health state to uh both of them are using the health state to inform Behavior but one of them is like the more hurt I am the more angry I'm going to be and the other one is like the more hurt I am the more scared I'm going to be and what this example is supposed to illustrate is that because all of this uh all of this complex behavior is hidden away in the modules this overarching system can focus on what create creates the specific personality for that creature instead of solving technical problems basically the modules contain all of the pragmatic stuff like the the the stuff that there to literally solve a problem come up with like a mathematical solution and then the Quirk and personality can happen here uh because it doesn't need to worry about that [Music] stuff who uh all right yeah so that was a little bit about the module system and how it's actually quite complex like all of the creatures are complex in this way it is it is kind of AI heavy and one thing that I think is uh interesting to note is that most games tend to focus on either multi-agent emergent Behavior as I would call it like maybe you have a spank right you have the the frog is simple and predictable the bat is simple and predictable but between the frog and the bat you're going to get like interesting emergent stuff or complex Ai and that's like a stealth game usually right the AI is very complex but the situation is simple and it's not like too much interaction between AIS and we kind of decided to try for both so we have we have multi-agent emergent behavior and we also have like extremely bloated complex AI code and what does that lead to because basically we have a circumstance here we have three noodle worms and what are they doing why are they doing this why do they hate each other they are here to illustrate that the AI will end up fa fighting the AI quite a lot uh another design pillar of rain world is that the player character is not very special like from the ai's perspective the AI doesn't ask a creature if that creature is the player or not it just asks stats about the creature uh the player doesn't have a hold a special place so end of the day this game is mostly playing against itself like the vast majority of the act of the interaction are going to be AI versus AI rather than AI versus player just because there are more AI agents in the world than there are player agents and the AI will react to all the other agents around it and this this leads obviously to like hundreds or maybe maybe thousands of possible ai ai interactions between all of the different creatures we have and those could be things like lizards uh sort of squabbling over territory or or uh creatures struggling over a food source or any any number of things uh yeah anything that you can imagine really we had we had something like 30 40 creatures in the base game and then you can just uh take the square of that and that's basically the number of possible possible interactions right so yeah it gets gets large rather quickly which came to my desk as the level designer and the person trying to sort of curate an experience for a beginner entering the game to sort of uh have an experience that was somewhat uh reproducible in amongst all this Dynamic Insanity uh and so what happens we get these we get spooky emergent behaviors all of these modules interconnected a first one to illustrate is this classic one and this is going to kind of set the tone for all of it so we were I was creating an area of the game called the leg which are sort of of crawling vertically uh and in the leg are these horrible grotesque cancerous creatures which are called daddy long legs and they hunt exclusively by sound so your footsteps your jumping is kind of what attracts them it's a little bit difficult to see but you can actually see here in the animation as the player Landing like this thing is lighting up and then it's sending a tentacle to feel for them so and then we also had these other creatures over here which were called grapple worms or tube worms and they were just sort of like a cosmetic creature sort of just doing their thing but uh for the next section of the game we wanted to have the option for the player to be able to like use these as like a a grappling hook sort of you can squeeze them and they sticky tongue will come out and you can swing like your Spider-Man uh across the uh what was the the under the underhang across the underhang but when we're putting these into the game to test them they didn't exist we'd put we'd place a million of them around and we'd go to test it and they weren't there and we would go and check the code and we'd go and check everything okay we'd place it in a room they're there we'd come back and they'd be gone and we were losing our minds it what wound up happening is that when the tube worm is uh flop flop flopping when it's flopping around it's making a sound right so whenever it's using its grapple tongue it's going like and the daddies they could hear them from far away and they would just come and they would eat them so by every time that we would test them and this was even just like the I know minute or so that it would took to get to this area to test it they would have been all eaten and so we're like what the heck so this was the first of our Many Adventures in this territory where in this large dynamic world trying to create like a repr somewhat reproducible experience just leads to Absolute Madness Big World big problems so rain World wind up being around 1,600 individual screens initially all of which have creatures in it all of which the creatures are um free to roam they're free to roam they exist off screen they're having lives off screen so sort of how do we like when we're testing initially to see what like a a a player coming in for the first time would see sometimes you'd immediately in the first room encounter like 30 lizards squabbling and you immediately die obviously you can't have that and then the next time you're playing it uh there's no lizards because a vulture has eaten all of them or like Ted there's just tumble weed so it was just this constant state of having to sort of like make these minor changes and experiment and test and then like try over and over again another one was again the agency problem so here we have um these were drop wigs these like horrible creature it's literally out of my nightmare which is I think how he designs creatures was just James what's your nightmare and he just puts it directly into the game so these creatures um not only do they hang from the ceiling like this and then drop on you but they also can place as you see over there they can sort of place uh what' you call it like a lure a little lure they place an object that they think a creature might want maybe it's a piece of food maybe it's a stone or something like that below it and when you go to investigate it they drop down and kill you and also they're free to shoose themselves where to set up in the ceiling in the entire map in the entire map and they will choose the most inconvenient places anything like that so like how do how do how do you how do you curate an experience around that like you want to have these Dynamic experiences where it feels cool as a player and you get this bit of surprise but it winds up being unfair at every stage of it because the creature has just as much power and to a degree as much intelligence as as you do the the interesting aspect of all this was I had had fun making AIS and like having them interu and then it was up to James to make this into a difficulty curve for and it was horrible absolutely horrible I I'll tell this is this is a really fun one just to show you some more of the sort of spooky emergent Behavior Uh so this is the Classic this is a white lizard a white lizard can kind of camouflage itself though it's not doing so here uh but it has a long sticky tongue that you can that it uses to sort of grab its PR um and when we were when we were testing this creature uh to test the tongue's range I think it was y had disabled the legs so basically broken the all of the legs of the lizard and we're sitting there and we're watching it and it begins to use its tongue to like stick to surfaces to drag itself along and this wasn't something that was programmed in this wasn't a behavior that we were aware of or anticipated it was just it had these various modules and it had access to certain data and it just kind of put it together in a way to achieve the results that it needed to so at every step of the way it became less about sort of like designing experiences and more like sort of hurting cats where you kind of have to just pay attention to their behaviors and their wants and like how they respond to certain environments and kind of try to guide them or sort of like uh Corral them into the behaviors or the experiences that you want to um I mean and there's instances where we sort of had to use like divine intervention eventually where it was like you know we'd basically have a circle where it was like Thou shalt not pass Thou shalt not cross this barrier but the funny thing is is that that would only work on some of them persistent ones just like with humans the devious ones they would always find some way to like get through it that we hadn't anticipated so sorry we were basically like we had a foe in the game we were fighting the AI we were the slug cat in this instance we were trying to find our way to how to get to the end of this the develop M of this game and we're basically beset by Predators on our sides that we had created when when doing emergent systems right you want to leave them alone as much as possible uh you want them to be able to do their emergent stuff in the end we ended up sinning about sinning against the emergent system rules and we there are certain aspects where we just had to go in and like hard code something or say like okay you you guys canot go here or this creature cannot eat this one or something like that because you have to try to patch it together but even still there was always and people still to this day will find exceptions within the game for stuff that we absolutely hardcoded with multiple uh redundancies against failure against they'll still just slip through somehow yeah things that shouldn't happen ended up happening and that's they still do and that sort of became our brand that's the that's that's what we do yeah oh yeah so what's our conclusion here as you are gets to the slide keep going um you missed a couple the I was supposed to just switch the slides while I was blathering I didn't gotcha gotcha okay um yeah uh multiple AI agents interacting will be unpredictable and this is something that we perhaps should have have been able to predict nope nope we which is why we're giving this talk to you please please learn from our mistakes from our 12 years of development in this nonsense multiple AI agents interacting will be unpredictable despite all of your best efforts um I mean uh yeah I this is this is obviously a predictable a predictable result we should have seen this coming and to some degree we did right like set out specifically hoping to get emergent behavior and we did get emergent Behavior we got so much of it uh little little bit more than we could handle in the end perhaps but also that is the uh that is what makes rain world uh interesting right that's what makes it fun that's what players like about it is you go into this like big chaotic thing and you sort of can find your own way and you you're not sort of give Direction on how to do it like when you overcome problems it is literally by your width it's not that you've sort of followed like breadcrumbs left by a developer or that there's a power up or something that you achieve it's that you've really found your own solution to this kind of like large scale problem and it feels good it feels cool and it's it's novel I think yeah um the the fact that we were not a the fact that what you were struggling with so much like building this difficulty curve in the end it was kind of an impossible task you you for the first couple of regions you managed to do to do it somehow but uh it is an impossible task under this system uh basically though uh the fact that it's not an authored very like uh predefined difficulty curve experience is also what makes it interesting into play um the level of communication about what's going on inside an AI agent is another interesting aspect of Rin world I believe uh many many games that are AI heavy also communicate very clearly what's happening inside the AI agent you might have icons popping up above the head or like very very visually clear like perhaps visible Vision cones this sort of thing in rain in Rain world the system is very complex and it kind of all of that that all of those games do but we don't show it and this means that it's not visible and to 98% of the players it's going to be as if it wasn't there at all uh which is a huge waste of work uh but I believe that for the remaining 2% that's going to be the thing that adds a little bit of magic right because you don't see the big you don't see the big exclamation mark pop up above the head and you don't see like the uh taming meter go up or something like that but it is happening and that means when when that behavior clicks with you and you actually realize what's going on it's going to feel cooler because you discovered it yourself well this goes back to sort of what I was talking about in the beginning with that level of projection that like when you're a kid you're playing a game and perhaps you uh there's like some creature in a game and you think that it's your friend like you think that you've befriended it through some sort of action when in reality there's really no there's no code to support that it's not within the design for this creature to be your friend in Rain you actually can befriend creatures you can feed them you can uh over time sort of gain their trust and they'll sort of like develop an affinity for you and I remember within the rain World Community this wasn't known because we don't Telegraph any of this stuff whether for positive or negative uh and when people discovered that you could sort of tame lizards and you could befriend lizards and they would sort of follow you around and offer you assistance to a degree to an animal degree uh that became like a really huge uh Discovery and like uh within basically within the community as these things become discovered it's like an organic thing which is very very powerful I feel like uh because playing the game then instead of just feeling like you are uh watching a movie or engaging in some sort of uh curated experience it really does feel like an adventure and you really don't know what you're going to come up with in a play session or in a group discussion and another huge benefit of doing it that way is that in the 50 let's say 60 perhaps let's say 70% of cases where it's actually just the system bugging out people will project something cuz we have enough instances where there is that depth that when a lizard just like spinning on a PO which happens often because the AI just freaks out people would be like oh it's like mad or oh it's this or like it hates this type of food or something like that and it's kind of supported within the context of the rest of the game which is beneficial to us obviously yeah uh you just because you add this ambiguous Behavior doesn't mean that that's a projection psychology is not going on it's still going on and in Rain world it was very deliberate to blur that line as much as possible we did not want people to know what's what the downsides are that it's extremely labor intensive we have spent probably couple of years like making content that no one will see um uh so perhaps it was viable to us in our situation like we were IND the we were in control of our own uh sort of production we could do that because we thought it was fun and that's what we wanted to do maybe in kind of like a more uh corporate situation it would be difficult but U then the the other downside to this is that to some people that play the game and they don't immediately get curious or they might like have a bad experience from from all of this procedural mush early on they could just bounce off it and they will never know that this stuff is in there because to them it's just like a frustrat frustrating mess and now I'm going to play something else and yeah this leads well into this leads well into our next slide which was the reception was was terrible because okay imagine giving you you give this game which we've worked on for like 12 years it has like this insane amount of depth to it that's not communicated at all we want everyone to experience it for the first time so we're not really giving much in the way of clues or assistance imagine you're a game reviewer receiving this game for the first time you're the IGN guy you have one weekend to get through it yeah and write a thing and you just come into this thing and you're just immediately destroyed or like nothing makes any sense and it looks ridiculous so we we received a lot of very negative reviews when we first released rain world because it's just a game that is inherently uh anti to that entire sort of quick review culture it's something that you need to spend a lot of time with you have to have a lot of patience for and for the I'd say the vast major majority of the people that's just not going to be their game it's kind of I would say it's like speed running Mist in 2 hours blind while someone is slapped in the back of your head all the time yeah absolutely I mean in order to get to the end there maybe 60 hours 40 to 60 hours concentrate and look and try again and it's yeah yeah it just it just doesn't work for that culture but and we had never given a thought to how a reviewer might experience this or how uh someone might Market it because we're just basement dwelling Indie artist types and that's what we care about so we kind of loed out to have this experience where no one told us this this was a terrible idea and it was going to be slammed uh beneficial because over time the long-term reception has been pretty good I can keep talking on it no the long-term reception that's you I guess that is me the long-term reception has been amazing because what happens is is uh the people who really like this sort of experience and have the patience for it they're really underserved in the gaming Market I think uh the people that really want to sort of engage deeply it doesn't work well with these marketing Cycles it doesn't work well with budget constraints it doesn't work well with uh like healthy work sleep life balance doesn't work with that at all but there's a lot of people that are Liv in that world so for them to have this sort of like wild uh open Dynamic uh confusing uh yeah it was cool it was something new and we just a little bit different because there wasn't uh any sort of um guides that you could do people had to like come together as a community to sort of figure out the like the the larger puzzles of the narrative and sort of the World building and that became super appealing and I we've I mean at this point we've released a sequel we have a third on the way and the uh the appetite for rain world seems to just be expanding because everyone can put out a YouTube video about their unique experience that sort of adds to the overall understanding of what this game world is and other fans want to experience it and people want to sort of create these large overarching theories about what the world is and like how things work because we haven't given them any sort of uh rules about what is and what isn't within the game so in the end it benefited us greatly it just required quite a lot of patience and kind of thick skin to get there couldn't have been created under any other circumstances I think yeah um thoughts looking B no no regrets no regrets no regrets I know there's things we could have done better but I think more just understanding that we were doing something that was for artistic purposes rather than commercial purposes uh would have been good I mean there was there was obviously fears of like where we're going to sleep next or if we could pay rent uh but I think we just kind of stuck to our guns and didn't listen to uh the the naysayers we didn't listen to the haters didn't listen to QA didn't listen to QA didn't listen to Executives yeah exactly but I kind of like coming from that I really feel that that's a that's a sort of that's a sort of truth though is that like if you have a real strong desire to build something and you resonate with it really strongly I think there is an audience out there for that there's an audience out there that's for what you think is cool and it's people that think like you and you're the only person that can bring that to life so I mean now we're sort of seasoned video game industry veterans and we see with amongst our peers and amongst other companies like there's there's always this sort of negativity towards ideas that are bad because let's let's face it from a design perspective rain bad like from a uh from like a business perspective rain world was bad like every analyst would say this is a terrible idea but it creates a new Niche it creates something that's uh unique and oh was that the oh perfect my time's up thank you very much [Music]