- I'm Paul Neurath. I oversaw the creation of Thief, a game that almost didn't happen because guards kept misbehaving. Well, at Looking Glass, the games we had done a couple years before Thief got under development were System Shock and Ultima Underworld. And those were really the games that helped define the genre that we were gonna be the best of and really innovate on. These are games that have deep immersion. The player can really make moment to moment decisions on how they want to tackle a particular challenge. We wanted to create a game that had a more focused role for the player. When I was 10, 12, one of the things I liked doing is just exploring buildings, like poking around buildings that maybe I wasn't supposed to go places you were supposed to go. I just enjoyed seeing what was around the corner and exploring. One particular instance where I was taking swimming lessons at a place in Downtown Boston, I was just poking around the health club and noticed that there was a basement and the door was open. So I went down there. And then there was a sub-basement below that that looked like it was covered in dust and it seemed like no one had been there in years. So of course that intrigued me. And then I heard someone coming, and it was kinda scary. So I hid in a corner, a dark corner. And a guard, a security guard walks by. And he's poking around, and apparently they had cameras or had heard something. I don't know how they knew I was down there. But I managed to stay in a dark corner. And they wandered around and then went back and I scrambled out of there. And so, I had a visceral experience of kind of what it was like to sneak around and hide. And so, I remembered back then, when we were coming up with the concept of Thief, to me in my mind, I had to convey that experience, that moment of experience to the player, where there's that real tension and fear about getting caught and can you outwit a guard. Stealth hadn't been core to our thinking, but we started to bounce other concepts around and stealth came up. And then the obvious thing to us was well let's play in the role of a thief. We want a more singular experience that's focused on a role. We know what thieves do. Players know what thieves do. It creates an immediate context, which is especially important if you're doing an innovative game. If you're already asking players to stretch and do gameplay they haven't seen before, to make that approachable is important, doubly important. So by just calling it Thief and making it about being a thief, it created the context. There were things we knew we had to solve for. We had to allow the player to hide in shadows and stay out of the light. We knew that we wanted to use the soundscape, the audio environment so that they could sneak quietly and avoid detection from the guards, from the noises they would make.

- [Guard] What's that? Who goes?

- [Thief] Seems peaceful enough now.

- [Guard] Must have been rats.

- We knew the AI had to be sophisticated so that the guards had awareness of what was around them. And we knew that we had to create an AI that where the guards could progressively get more aware of what the player was doing. So they might suspect something. They hear a sound that sounds out of character.

- [Guard] I'll find you.

- And they say, "Hmm, what's that? "Was that something?" But then if they don't hear anything for a minute or so, they decide, oh, it must have just been a rat. So we knew we had to do that sort of progressive warning in the mind of the guards. So the lighting was a key part of it, and other games that were coming out before it, there were not that many, or at the same time being developed. Tenchu, Metal Gear Solid, some other games. It was really more line of sight. Can the guard see me? And it would be done in a sort of very game-y way, where a guard, you would display like a view cone.

- [Guard] Did that shadow move?

- To make the audio stealth environment work, or game play work, we had to do more sophisticated audio than I think anyone had attempted to date. We mapped the physical space of a level. We created a second layer, which is the audio space of the level of soundscape, and created that. And the AIs would use that. And then we propagated sound through that environment. So if a guard was around the corridor, around the corner, down a corridor, sound sounds differently when it reflects off walls than when it's coming without a reflection directly at you. And most people don't really, are necessarily cognizant of that. But it's just built into somewhere in your brain that all the way sound propagates, it's giving you information about the space around you. You just know without necessarily even being fully aware of it. We started to build the game and put the systems in place. And it was about a year and a half into the game when we didn't have stealth working well. As a player, you got confused signals. You couldn't really understand your full stealth state. You didn't necessarily know what the guard was thinking or how well you were hiding. The AI design was such that it couldn't effectively pick up the cues we needed it to do, like a blood stain on the ground or just the stealth aspects. And so it was all muddled. You tried to sneak in, and sometimes you'd be ridiculously excessive. You'd walk right by a guard, they wouldn't notice you. Or sometimes, you'd be hiding very effectively, and the guard would say, "There you are," and rush right at you. It was not the complete experience, and it was flat. We didn't have a game. To be a year and a half in a game with saying, "The core game isn't really working," "It's not fun to play," it was a very scary thing. And it was a building kind of tension for us because we had never gone that far into a game without having a core system going. There was also the added tension that our publisher Eidos was quickly losing confidence in our abilities, so they were actually saying, we may just cancel this project and walk away from it. Basically Eidos gave us a year and a half to have a vertical slice, rock solid, all the stealth works, all the game play works. But again, we just decided we had to do it. If we had failed, which we could have, it would have been a significant downsizing of the studio. There was definitely growing stress on the team. We might not be able to pull this off. We wanted to make it immersive and natural. The lighting part of that meant line of sight was still important because if the guards are not looking your way, they're not gonna see you. Beyond that, the extra depth that we had to solve for is, okay they're looking at you, but if you're hiding in a dark corner, they probably don't see you. We actually had to map the lighting environment in a very precise way to the physical environment around you and then feedback to the guards how visible you were depending on how much light you were in, how much you were moving, how much of your full body was being displayed to them, how close you were to them. So there were a bunch of factors that we tried to make it feel as realistic or understandable, might be a better word, to the players, say, that makes sense that they saw me. I was kind of halfway out of the shadows, moving kind of fast, and they were looking right at me. And that point's very important because we had to find. It was one thing we didn't fully understand. We had to feedback to the player their stealth state in terms of visibility. And so, showing that in a very clear way was important because if you don't know whether you're in shadow or how much in shadow, a lot of this just doesn't work. The player is like, why did the guard detect me? I don't know. Well, I don't know how hidden I am. One of the team members came up with the concept of a light gem, which shows up at the bottom of the screen. It's just a user interface icon, but the light gem gets brighter or dimmer depending on how hidden you are, visibly hidden. There's also an audio component. But that really helped, and you could just really look at it and intuitively, quickly figure out, oh yeah, I'm really well hidden now or I'm partially hidden or I'm completely exposed, I'm in bright light. Some pieces are starting to fall into place, but one of the pivotal things that happened is that our lead programmer on the project, Tom, he had a bit of a mental breakthrough in coming up with a new scheme, a new thought, thinking process of how AIs could behave in the environment. And he pitched this as we need to rewrite the entire AI system, which seemed kind of insane at that rate of the project. But he made some very compelling, he explained his thinking and it made a lot of sense. Tom literally spent three weeks working like 14 hour days, every day. And at three weeks, he'd rewritten a massive amount of code, the AI system, pretty much from the ground up, and it all worked. It was one of the more brilliant pieces of engineering I've ever seen, so it was pretty cool. As I recall, he had to sleep for the next week. He was just dead tired. But he wanted to see it through. It worked, and it worked really well. So that whole soundscape system came online around this time as well, combined with the AI rewrite, a couple other pieces. It all just happened to come together. I remember there was a build late that year, not long after the AI was rewritten, where I had the same experience I did as a 12-year-old of hiding from a guard walking right past me. Then of course, what I didn't do when I was 12, I picked up a blackjack and knocked him out. But it worked and it felt great. We had imagined when we started on it, when we decided the Thief was the game we were gonna make, we had imagined it, maybe a half year in, we'd have all this done, all working. We thought, maybe a little bit sooner, a little later. By the time a year had gone by and we weren't there, we were really wondering. But we had already so committed to the project, it would be very painful for us just to abandon. You have to make mistakes. You go down a lot of dead ends. In this case, a lot of those dead ends didn't pan out, but we were learning. So it wasn't as if this final, that final stretch a year and a half in, when all the pieces came together, part of that was because we had already learned what didn't work. We'd already gone down a bunch of dead ends, so by the time we got there, we actually had a good mental handle on what we needed to solve for. An old adage, particularly when it comes to coding and technology is, the hard thing is having it in your head, the easy thing is writing it all out as code or whatever. That was the key thing. We finally had the mental model after just doggedly pursuing this for over a year. Now we know what we needed to do to get this done, and we figured it out and got it done.