Doom Guy: Life in First Person

by John Romero

When Carmack and I met, it was an absolute meeting of like minds. He was a fresh-faced nineteen-year-old who looked like he had yet to shave. I was all of twenty-two. I eventually discovered that we came from different sides of the tracks—his family was far more affluent than mine—but as we talked that evening, it was almost as if we had inhabited parallel lives when it came to games and programming, having both spent thousands of hours mastering our three obsessions: games, computers, programming. When I say “thousands of hours,” I mean that literally. Over the course of my life, I have bonded with programmers, including some of the most legendary figures in games and computer design, from Steve Wozniak to Bill Budge to Nasir Gebelli and others. However, on that first meeting, Carmack and I instantly connected on multiple levels because we shared a deep knowledge base and passion that creates instant bonds; we understood each other and the significance of our respective work, and we admired each other’s abilities. We spoke a language that, while rooted in normal, everyday English, was unique. We saw a world of possibility in games, and we knew we had the knowledge and drive to master it. We saw that in each other. It’s hard to describe how it felt. Imagine caring deeply about something, so deeply that you feel compelled to master it. Imagine no one around you gets it—not your family, not your partner, not your friends. Not really. Imagine meeting someone else just like yo...

He took the job, and I was beyond excited. I was even more excited when I saw Carmack in action. He could and would sit for hours and hours studying and coding. Like me, he didn’t smoke, drink, or do drugs. Unlike me, he didn’t have a family or even a girlfriend. He lived alone, which meant he could spend every minute of every day poring over the PC. And that’s what he did. The only time he wasn’t programming was when he slept or walked to our office refrigerator to get another Coke. He was in boot camp, working 24/7, trying to learn as much as possible about PCs. It was incredible to watch someone else besides me run a programming ultra-marathon. The three of us set up our office for serious work: In addition to Intel 386 PCs provided by Big Al, we invested in a fridge, a microwave, and a Nintendo Entertainment System (NES). We liked metal, so we cranked our favorite tunes. Our eyes were locked on code, on books, on magazines, downloading it all into our heads. Carmack and I shared information with each other, teaching techniques picked up along our solo paths. We didn’t do this because we had to. We did this because we wanted to. If you’ve ever stayed up far too late because you were playing a game, bingeing a series, or reading a book, code had the exact same effect on us. Learning more was its own reward.

Fortunately, I knew how to harness my abilities to do that, and kept my mind busy trying to accommodate Big Al’s request. The plan—I dubbed it “the death schedule”—was to work from 10 a.m. to 2 a.m. We realized the quickest solution was for us to repurpose previous work. In 1988, I’d made an Apple II game that UpTime had published called Dangerous Dave in the Deserted Pirate’s Hideout! And I decided to rewrite it for the PC. Meanwhile, Carmack decided to rewrite his Apple II game Catacomb.

Carmack introduced us to his campaign, which he had been running for several years. Carmack was the dungeon master, and Tom, Jay, and I rolled new level-one characters. Carmack led us through his eerie, dark fantasy world full of demons, mysterious places, ultrapowered beings, political intrigue, and much more. He was an excellent dungeon master and loved crafting campaigns. I admired his abilities to design gameplay for us within the D&D system, which was, at the time, a shadow of what it would eventually become. This was the beginning of a pattern for us that would continue in the future as we grew closer and more united: Make games during the week and play games on the weekend.

A lot of id’s biggest decisions happened late at night. Our unofficial motto was “We are the wind,” meaning we could blow out of wherever we were or change direction at any time. We worked superhuman schedules. We developed games faster than anyone. To do that, we quickly absorbed information, analyzed it, and made decisions. We also were id-driven. We wanted whatever it was we wanted—to make the best, coolest, most fun games—and we wanted it now. We called our decisions bit flips—when a computer value changes from on to off, or vice versa.

Set up in new offices, making Wolfenstein was some of the most fun we’d ever had making a game. A lot of that had to do with our liberation from Madison and our new life in Texas. It was warm. It was sunny. We had apartments at the La Prada Club Apartments and a separate apartment that served as the id office with a swimming pool right behind it. You could open the door and take a dip and lie in the sun. We’d had fun in Shreveport, too, going kneeboarding on the lake, but we were consistently working seven days a week back then and had all been in survival mode. The truth is that we relished the enormous pressure. You know that moment in a game where you’re on your last life? Sometimes that’s what it felt like. We had to keep going, keep playing, or the dream would be over. Now, in Texas, we had a financial cushion from Keen’s success, and Wolfenstein was bound to make far more money. If we wanted to hang out and play games, we could. Street Fighter II and Fatal Fury were our two favorites at the time. There was pressure and a lot of work, but life felt a lot more balanced now that we were focused on delivering just one game.

The hardest thing about making Wolfenstein was finishing the levels. The level design was not visually sophisticated, but that is because we made a trade-off, sacrificing visual nuance—textures, colors, angles—in favor of rapid-fire speed, which was part of the thrill of our FPS killfest. Adding those features would have slowed the game down and required more work on the engine. We felt that the slimmed-down feature set was exactly the type of gameplay we wanted. That kept the core loop clean and easy to understand, which was important since it would be the first fast 3D game that anyone had seen.

Earlier in March, Tom and I noticed Wolfenstein was missing something, something fun that we were used to, but that this engine didn’t support: secret areas. They were a staple of exploration and of our games. We lobbied Carmack to add some way of pushing a wall to reveal a hidden area, but he didn’t want to violate the purity of his engine to hack in a secrets-revealing “pushwall.” The next couple of months, we pressed him on it until he finally agreed in April to hack it in. It was a trying situation for Tom because, as the creative director, his design was expected to be supported by tech, and now th...

We decided to call it Wolfenstein 3-D. It was the third in the series and 3D.

Everything was ready to go on May 5.

Wolfenstein: January to May

We were finally done. The shareware episode took us four months from start to ship, and the other five episodes of Wolfenstein 3-D were done in one-and-a-half months. In total, just shy of six months for a six-episode game.

It was November, and we were releasing in a month. I brought it up to Carmack. “So when are we going to make multiplayer mode?” The short answer was that Carmack was ready to take it on. Looking from the outside in, I suspect some might wonder if I wasn’t just more than a bit concerned since we were hoping to ship in 1993. After all, John had never programmed a multiplayer game before. The truth is that I never had a doubt, not for a second. Back in March, Carmack had already done some innovative network programming in DoomEd. He wanted to play around with the distributed objects system in NeXTSTEP, so he added the ability to allow multiple people who were running DoomEd to edit the same level. I could see him drawing lines and placing objects on my screen from his computer. Then, I’d add to his room by making a hallway, and so on. For multiplayer, Carmack’s plan was to explore peer-to-peer networking. It was the “quick and dirty” solution instead of a client-server model. Instead of one central computer controlling and monitoring all the action between two to four players, each computer would run the game and sync up with the others. Basically, the computers send each other updates at high speed over the local network. The speed of Carmack’s network programming progress was remarkable. He had some excellent books on networking, and fortunately, those books were clearly written and explained the process of using IPX* well. In a few hours, he was communicating between two c...

On December 10, 1993, DOOM was on the verge of release and everything in the world of games was about to change. It had been an intense eleven months, and we had achieved everything we set out to accomplish in our initial press release. We were completely exhausted, and Sandy was sleeping under his desk. We had been up and working for thirty straight hours to get DOOM to this point.

The multiplayer aspect of DOOM transformed games, opening a path that has led to competitive gaming as we know it today. That’s why id is often credited with spawning competitive gaming. However, competitive gaming had been around for a while. The 1980s were incredibly active with arcade tournaments all over the nation. Even I competed in a tournament in 1981 at a local arcade, Video Bob’s, and it was packed full of kids and excitement. That said, our games did drastically change gameplay and unquestionably influenced competitive gaming. DOOM was used for the first modern computer-based tournaments, giving rise to a fledgling eSports community.

Within two weeks of releasing DOOM, we put the data structure used to build the game’s levels on the internet. Sharing that data made good on our promise. The decision to open our game up to the public was something that John Carmack and I felt strongly about. It was part of our core as programmers; we believed in sharing technology so that players could build their own creations and, ideally, lead to new improvements. We wanted and expected people to build editors, create new levels, make new sound effects, add new wall textures, and design new monsters. In fact, we looked forward to it!

Everything we had discussed from 3D to QuakeC sounded great to me. We had a preliminary idea of what we wanted, and as we began the new year, we had to create all the tools and tech necessary to make an engine capable of building a full 3D world. Even nowadays, FPSs push engines and computers to their limit, and creating an engine, the tools to support it, and a design to show it off is a Herculean task. We’d done it three times before, but this engine was the most complex one yet. We guessed it to be about ten times more difficult than DOOM’s engine. This metric allowed us to estimate the time it might take, but in the end, it was only a guess. We were charting new territory.

American purchased a house next to Carmack, and Carmack, who was by now possessed by the problem in front of him and working constantly, took to visiting American at all hours to bounce things off him, all of which would have been well over American’s head as a non-programmer, but American was a good listener and provided a sounding board, at least. Adrian and I used to joke that there was a tunnel between their houses.

Based on our past experience, I think we all expected that we’d be able to dig into building rudimentary Quake gameplay in a month or two. With Wolfenstein 3-D and DOOM, Carmack had gotten us working engines within weeks, but this time, the months passed. This engine was a massive technology jump. Not only were we raising the bar for 3D, we also wanted to connect players to one another through the internet.

By now, the engine had been in development for ten months, one month shy of DOOM’s entire development cycle, and gameplay programming had still not begun. We did not have a game. What we had was a technology still in development. Quake still had no hammer, and the things we had discussed with players and the press were not yet happening. This wasn’t due to any fault of the programming team—quite the contrary. What they did and were doing was monumental and transformational. It was Carmack’s most difficult challenge. He was doing what he had to do—making something incredibly awesome—and as a programmer myself, I knew it was going to take the time that it took. As the engine evolved, I steadily updated QuakeEd to account for its new features and to improve the experience for us level designers.

A week later, after eleven months of development, John Carmack and Michael Abrash declared the Quake engine ready. It was a Herculean engineering success.

American drove the nail in the coffin. “We’re just burned out, John. We want to make this game and get it over with. We should just put DOOM weapons in the game and call it done.” The design team and art department had spun their wheels for the better part of a year. We had designed about fifty R&D levels during that time, and while the experience was critical toward building our understanding of crafting play in a 3D space, the levels themselves were essentially experiments that lacked the necessary gameplay code to make them useable in the version of Quake we envisioned. That work would need to be remade or abandoned since the engine had been under development and changing constantly. The code team was worse for wear, too. Michael Abrash had had to rewrite his highly optimized renderer eight times. John Cash had written a bunch of network code, some of which Carmack didn’t like, and so Carmack rewrote it over a weekend. The team was brittle. It wasn’t a team that was ready. It was a team that was done. At the same time, I wasn’t done, and I couldn’t believe what I was hearing. It was opposite of the innovation that had defined us. I’d discussed necessary explorations once the tech was ready: We had opportunities for characters to engage in melee combat, possibly in third-person perspective. I planned to explore using a first-person perspective for aiming and shooting a crossbow. I wanted to experiment with new ways of playing in a true 3D world. Perhaps American, Sandy, ...

Trying to get people to do something they didn’t want to do would not make a great game. Still, I imagined how Carmack would respond if his tech plans were shot down by Dave, Sandy, or American. At best, he would ignore them and walk away. After taking a whole year to get his programming vision together, Carmack was now entertaining talk about stopping any similar exploration of design using his engine. It was a painful moment for me. In that room, with everyone standing around, I realized that our shared vision was over. After five years of dedicating equal importance to his tech and my design—a formula that always pushed for equal innovation and absolutely worked—Carmack was listening to people who hadn’t seen us go through a new engine cycle, and who were too burned out to put in the effort it took to reach for the stars. Merely copying design ideas of the past was not what I was here for. And with the loss of design’s importance, id lost me.

In December 1995, Carmack wanted to move everyone into that area, call it the War Room, and enter crunch mode to finish Quake there: that’s seven days a week, twelve hours a day minimum. We never had mandatory crunch mode at id before this, but we wanted to finish the game, and we knew being together in the same room would allow us to finish the game faster, so we agreed. In a sense, the War Room was a reflection of the team it housed. It worked, it did what it needed to do—but it was frayed and, well, wasn’t what it used to be.

Most days, Carmack worked until after midnight and would come in around noon. The rest of us worked 10 a.m. to 10 p.m.

In April, Michael Abrash was having some trouble while testing his code. Occasionally, when he was moving around in the game, the screen would show a completely random view that was from a totally different location for a split second. He debugged all the way down to the exact instruction that was responsible, but determined it was impossible to get the calculation result that would produce that random viewpoint. Carmack looked over the code as well, and this error didn’t make any sense. They were confident that there was not an error in the code. So Abrash called a friend who worked at Intel to come check it out. His friend saw the error happen, closely inspected Abrash’s code, and noticed that a floating-point instruction was causing the trouble. He then asked Abrash what kind of processor he had in the PC. It was an overclocked 100Mhz Pentium. As it turned out, there was a hardware-level bug in the Pentium’s floating-point divide instruction (FDIV)—if he divided certain pairs of numbers, it would return incorrect results. This calculation was responsible for the player’s viewpoint angle. Abrash modified the calculation so it wouldn’t run into that problem again. This Intel Pentium hardware error is known as the FDIV bug. It remains one of our favorite debugging stories.

In the end, Quake’s levels were divided among the four designers: • American/Tim co-created: 1 level • Sandy Petersen: 7 levels • American McGee: 9 levels • Tim Willits: 10 levels • John Romero: 11 levels It’s a tremendous credit to the entire team that Quake was as great as it was, particularly given that it was redesigned from an action RPG to an FPS with just seven months to go.

On Saturday, June 22, 1996, I came in to the office and was alone. The previous day, I told the team that since the game was finished, I’d be in the office the following day to upload the shareware version. Quake was ready to be unleashed, and none of the others were there. That’s how burned out and disconnected we’d become. It was both surprising and sad. Every id release before Quake was a company-wide, sometimes industry-changing, event. It didn’t matter if it was four in the morning, everyone was there to watch our games launch. Newsgroups and bulletin boards provided us with a live, interactive feedback loop. We could see the reactions. We’d celebrate.

With all our previous releases, there had always been some sort of celebration, even if that was just us turning our chairs to face one another to talk about what we had done and what we would do next. The empty office was, in a sense, a reflection of the energy we had left. It was unfortunate that no one else was there to celebrate the release of the game that would redefine the modern shooter. Truly, there was a lot to celebrate, with features like internet multiplayer, a true 3D world with fast perspective-correct texture mapping, level design that took advantage of the engine’s capabilities, an in-game console for executing commands into the game, mouselook for viewing the true 3D world in a new way, and a programmable QuakeC language that allows modders to create anything they like in our game. I took two weeks of vacation, relaxed, and enjoyed spending time in our new house. I was so exhausted, I have little recollection of it. When I returned, I was ready to master Quake’s retail shareware CD.

There’s a figure I sometimes hear kicked around in the game industry—for every ten games started, one makes it to completion. I suspect its actually more games started than that, like for every one hundred games pitched in demo form, one makes it to completion. Even that estimate might be generous.