Interview: Early Access, Stealth And Mods In Invisible, Inc.

Invisible, Inc. is coming to Steam Early Access on August 19th. It's developer Klei's return to stealth after 2012's Mark of the Ninja, but this time it's turn-based, tactical, and about steering a team of operatives through a Syndicate-inspired world of corporate espionage. It's also their return to procedural generation and permadeath after 2013's vastly successful Burtonesque survival game, Don't Starve.

I spoke to company founder and programmer Jamie Cheng about why they came back to stealth game design, the challenges of procedureal generation, the right way to do Early Access, and mods.

RPS: Klei have done stealth before with Mark of the Ninja. Why do you like the stealth genre?

Cheng: There's a lot of really interesting things about stealth that we think we didn't explore the first time. In general I think that stealth is fun before you feel powerful but at the same time you feel that you're really vulnerable. We're also just kind of sick of shooting people in the face or chainsawing. We did that a couple of times, and there's more to explore in this medium, let's try something else.

The hilarious thing is we thought, hey, we know how to do stealth now. We'll do this turn-based thing. As it turns out, there's three major factors that make this game a lot different than Mark of the Ninja, a lot of different problems.

The first being that it's multi-agent stealth, and then the second being that it's procedural, and the third being that it's turn-based. Those completely change the game in many different ways.

RPS: What are you exploring about stealth this time that you didn't before? How do those change the design?

Cheng: The most interesting I think is procedural again. I think it's super key here. We decided to do it because we like to play with that feeling that information is power. That was also key in Mark of the Ninja, but you could play through the game and then you're done. Here, every single time I play, I still have to be extremely careful. In fact, I often die on the first couple of levels. It's difficult and if I rush it then I can get myself into really bad positions, but I also know for sure that it's a completely fair system and that I'm able to be a good spy and put myself in solid positions.

The procedural part is simply a great way for the game to renew itself every time, and I think it's especially key in a stealth game rather than, for example, if I was building a tactical combat game. Procedural would be interesting, but it's not nearly as interesting as in stealth where information is actually a huge part of the game.

RPS: Can you take me through the step-by-step of the procedural geneation?

Jamie Cheng: A programmer literally just re-wrote it a few weeks ago. Basically we tried one way and the one in alpha is the old way. And it has some serious problems with it that we couldn't resolve. Like it was, it was one of those cases where it was 80% there, but you can't get 100% ever.

RPS: What were the problems you encountered with the original?

Cheng: The original method was that we would have a rectangle and then you start subdividing the rectangle into smaller rectangles. Now you've got a bunch of different rooms, so now you want to make some hallways so you start deleting some random walls. That starts making hallways and then you find paths between them.

But what then ends up happening is that you'll have funny shaped hallways that the game thinks are rooms. It was very hard for the system to know, is this a hallway or is this a room? Because after that, the next step is to add the furniture. What ended up happening is we'd have a desk in the middle of a hallway, and it just makes no sense to have that happen.

The other annoying thing about it that I really didn't like was that the level was exactly rectangular. It was always rectangular shaped, so you knew that [as a player] you just had to fill in this rectangular shape to see the whole level. Whereas the new system is more interesting; it can have a room jutting out and then nothing around it, things like that. One of the things about procedural generation that's so interesting is not knowing what path you need to take. So you're trying to optimise your agents to work in an efficient manner to search through your rooms.

RPS: How are you handling mission objectives? Does the system make decisions about what kind of mission to pair with a generated level?

Cheng: That's right. So there's two different kinds of places that we have. Let me talk about the overall structure of the game first.

The structure is that there is an end mission in 72 hours. Your job is to prepare for that final mission. There are all these different places that you can go to, and you choose to go there for the purpose of preparing. There's a lot of places that you go to that don't have an explicit objective; it's loot the place and get out, that's really what it is.

In those places we build interesting locations, and they have something special in them. So for example one might be a corporate server farm, and with that you can raid their program database and install a program onto your system. You could call that your objective, but at the same time it's, what are you comfortable with? Because if their security is too hot, you might want to just leave before you get the programs. That's OK, that's totally fine within the realm of you trying to prepare for your final mission.

On top of that, we have explicit missions. So for our early access, we're only shipping with one, because the other one's we're still polishing. And there will be a final mission, but it's a temporary final mission that we have right now. That's a hostage mission, so you've got to get the hostage and you've got to get them out.

For that one, it's still procedurally generated, it's just that during the proc-gen stage we have parameters and we say that the hostage room has to be a certain distance away from you, which has to be a certain distance away from your final exit. That's the constraints we have put in to the generation when we're building missions.

RPS: A lot of stealth games aren't really tactical, they're puzzle games. They're about watching a patrol route and finding the one correct way to play. That can be fun, but how do you provide gradients of success and failure and introduce tactical decision making to a stealth game?

Cheng: I would say that's one of the fundamental differences between Mark of the Ninja and this game. In Mark of the Ninja, there's many different ways to approach a problem, but we thought of about 60% of them. Maybe 80% of them. We designed it in a way where we expect people to approach it this way, but they might surprise us. If you die, then you can look at it again and plan your route from there.

In [Invisible, Inc], it comes back to that sense of not knowing, and I don't know either as a designer. I don't know what that situation is going to look like. There is no correct solution; or rather, there isn't a correct solution that anybody actually knows. If I reveal the map to you then yes, you could plan out perfectly your puzzle. But here, instead a lot of your time is gathering information, but if you want to push it - because there is an alarm that's going, so time is ticking as well and you can't take forever scouting out. You have to put yourself in good tactical positions in order to move forward.

RPS: Does the player select the types of missions they do in any way, or is it random?

Cheng: It's not quite random right now. There is a map. The map allows you to choose your locations, and there's a travel time to the different locations. You're trying to pick between locations that you want to go to versus how much time it's going to take you to travel there, because you've only got 72 hours.

Each location does tell you which corporation it is. We went through like four different ways of going about the whole system, the whole game.

RPS: What was the original pitch for the game; was it, 'let's make a turn-based tactical stealth game?'

Cheng: Nope! [laughs] The original pitch was-- I really loved Syndicate. We felt like we could explore something with special agents, a team of agents going around. And we weren't sure where we were going with that. We tried a whole bunch of different things. We built it out of a card game first, we did a little boardgame about it, it was very mechanical at first. It took lots of twists and turns before it became what it is.

RPS: What did you learn from doing Early Access on Don't Starve?

Cheng: We love it, that's the number one thing. We were actually early access before it was called Early Access for Don't Starve. I guess we're happy to say we were also first to graduate as well - become Early Access and the first to come out of Early Access.

What we love about it is that it genuinely makes the game better. At least, if that's the way that you want to use early access. That's the way I think it should be used, and that's also the reason why we did the alpha access on our own website. We did that for Don't Starve as well, where it was available on the Chrome Web Store. Almost exactly two years ago it was already available to play, and so a small number of people would play our game and we'd get a whole bunch of feedback before tons and tons of people bought it.

We did the same thing for Invisible, Inc where we just sold it on our own website, so only the people who really knew about us would be playing it. We could test out the base mechanics to see if that was any good at all before we moved on to the next step. Now we feel like, OK, we have a solid game that a much larger set of people can try out. And we know that a whole bunch of people are going to be like, 'hey, I'm going to wait until the game is done,' and that's totally cool.

RPS: Do you have a plan in Early Access, for how often you're going to release updates and update the community?

Cheng: We definitely do the planning thing, although it usually it goes right into the garbage as soon as it touches people. That's what we've found.

Yeah, we have plans. We have plans for new agents, we have plans for a new corporation, we have plans for new missions, a plan for daily challenges. We're not exactly sure how long it'll take us to build those things. Currently I think it's going to be around monthly. For Don't Starve it was every two weeks and that was a pretty intense pace. I also think that works better in Don't Starve because it's more agreeable to adding things that don't quite work yet. In a stealth game, it's really, really obvious that it's broken, and it totally doesn't work. At all. The whole game falls apart if something doesn't work. We've got to be a little more careful about that, and we definitely want the game to be solid the entire way through.

So I think it's going to be about once a month, but when it hits the public we'll see.

RPS: You mentioned that you were the first out of Early Access with Don't Starve, but why come out of Early Access at all? If you're going to keep adding things, why not keep the label?

Cheng: I think it literally is a game of semantics, but it's kind of weird to call a game 'early access' when-- how is it early access? I don't know. I mean, you can call it 'continually updated', but to me early access is literally it's an early access before any general public should be able to play it. So we had a point where we said, at this point, we think that anybody can come in and they'll enjoy the game and it will be solid for them. That's where we drew the line.

RPS: You mentioned when we spoke to you back in January that you didn't know yet whether it would have mod support. Do you know now?

Cheng: We don't have it there right now. It's not implemented right now. That's about all-- I don't know. I actually still don't know.

What happened in Don't Starve was that the game was working, so now let's look at what other things we might want to do and mod support is one of them. So we went ahead and implemented basic mod support and slowly built from there.

[Invisible, Inc] has taken everything that we got, basically. [laughs] Just to make the game really solid has taken everything we've got. Maybe after that, we can think about what else we might want to add, but right now there's no mod support.

Invisible, Inc. launches into Early Access on August 19th. We'll have hands-on impressions soon.