This is my attempt at putting together a coherent history of the development process of I, Robot. When possible, I draw from Atari corporation documents.
In the Beginning there was Ice...
So what is the story behind the development of this game? Well, for starters, even though the game was released in 1984 in-game screens show a copyright of 1983. This suggests that by the end of '83 the game was mostly complete and in field test.
Atari collectors have unearthed a few prototype I, Robot game PCBs bearing the name Ice World. At the time games were often developed under a name that is different from production, and corporate records confirm that Ice World was the project name for I, Robot. Several sources online also claim that at one point the game was called Ice Castles, although there doesn't seem to be any existing Atari documentation that uses this name.
Poking around the internet, I found a few documents referring to the development of Ice World. The earliest document I could find was dated June 8, 1981. This is Dave Sherman's original hand written game project description.
Dave Sherman notes June 8, 1981
What's interesting is that the document came from Dave Sherman, and *not* Dave Theurer as I had assumed. In fact, Theurer's name doesn't appear anywhere on the proposal. The game (hardware at least) appears to be Dave Sherman's idea. As one of Atari coin-op's best hardware designers (maybe *the* best), my guess is he was dying to take a stab at designing real-time 3D polygon rasterizing hardware, and felt he could do it at a cost suitable for production, and the timing was right to make the attempt. Gameplay was likely an afterthought at this point. It's safe to assume that Dave Theurer (the game programmer) got involved later once actual hardware existed at a prototype level.
From this initial proposal we can gather that gameplay was intended to be some sort of first person view flying game, with the player flying through a canyon and attacking enemies. At the end of the canyon the player would battle a fortress that could emit "death rays". The "ice" theme seems to be reflected in the blue and grey colors that were proposed for the polygonal canyon. My guess is that Sherman though that the polygons reminded him of facets on a crystal, and so a crystal or ice themed planet would be a great stylistic fit for the game,. The initial gameplay proposal is not important however, and it would morph during development several times. The name "Ice World" would stick around until the final name was chosen.
Here are a few more Atari documents I found that refer to the development of Ice World.
November 1981 reference to Ice World
August 1982 reference to Ice World
The first document shows us that by November 1981 Dave Theurer has been added to the Ice World project team as lead programmer.
Also note the August 1982 document mentions a game called Heart of Ice... more on this later.
Well, what else do we know about the development of the game?
Crazy Mondo Hardware
Dave Sherman was responsible for the hardware, which included a custom designed matrix math and 3-space transformation coprocessor, as well as bespoke triangle rasterizer circuitry utilizing a custom designed ASIC. This was no small feat! In a 2023 interview Dave mentions that Atari attempted to leverage this technology into commercial CAD workstation business.
The hardware is composed of three main subsystems:
Main CPU subsystem
A Motorola 6809 CPU is responsible for running game logic and coordinating the work of the other processors (including the Quad-POKEY chip that generates audio). This is where the "game" programming resides. Dave Theurer wrote the software that runs on this portion of the board.
"Mondo Bondo" Mathbox
The mathbox is a custom 16-bit co-processor that can perform 3x3 matrix math, light vector shading, coordinate rotation, and camera projection. It also is responsible for generating the raw list of 2D dots/vectors/polygons to be drawn by the ICY video processor.
Sherman was not only responsible for the hardware, but he also did all of the custom microcode programming for the mathbox.
Owen Ruben says this was called the "Mondo Condo", due to the large number chips wire wrapped and stacked vertically on the board, suggesting high-rise condominiums.
"Mondo Boffo" ICY Video Processor
This chip (and supporting circuit) does all of the actual drawing, rasterizing objects a pixel at a time. It's capable of drawing dots, vectors, and polygons. Interestingly, the hardware was capable of rasterizing multi-sided polygons, not just 3 sided triangles. A technique that seems to have fallen out of fashion, all modern pipelines focus on triangles.
First designed with completely discrete components, Dave later moved some key functionality into a custom designed ASIC. This was to stop pirates and competitors from ripping off their design. A few prototype boards exist with the discrete circuit - hopefully this will get documented someday.
Ice World is ... A 3-D Driving Simulation?
Although the intial game proposal described a flying game, Ice World gameplay went through several different concepts throughout it's development.
Here's an excerpt from a message Dave Theurer wrote in 1989 talking about the development of I, Robot.
Wait... What? I, Robot started out as a driving game?
So it would seem the concept of flying through a canyon didn't last long, and was dead by the time Theurer joined the team. Atari had been wanting to do a real 3D driving game for a long time, since at least the late 70's. The most likely scenario here is that Atari management saw the project and said something along the lines of "our first 3D game should be a driving simulation" and that was that.
Here you can see a memo dated 1/7 talking about the different gameplay concepts that were being planned for the game. It's not clear what year this particular document is from, but from the overall timing I've pieced together in this article - it seems safe to assume this document is from January 1982 -- early in development - before any robot themed ideas crept in.
Jan 7 1982(?) document describing different game play elements in development at the time
Dave Theurer stated that the driving concept involved stopping at buildings and playing the games that were inside them. My take on the Jan 7 document is that the gameplay items being described ("rollerball", "flying game") would be the minigames or "metagames" you would play inside the buildings.
The driving concept was attempted, but cancelled early on as it was clear the hardware wasn't up to the task. Owen Rubin, who happened to share a lab with Theurer at the time, talks about this.
At this point the game concept had to be retooled. But into what?
Heart Of Ice
Note that the August 1982 document above references a little known game called Heart Of Ice. Here's the original April 1982 proposal for that game.
Heart of Ice game proposal page 1
Heart of Ice game proposal page 2
Heart of Ice was to be a 3D game that was "basically a raster scan analog of X-Y". From the description the game sure sounds a lot like I, Robot, albeit with robots and pyramids replaced with a princess and icy castles. While the driving game Theurer described bears little resemblance to the game he delivered, the gameplay of Heart of Ice sounds remarkably similar to I, Robot. Check out the numerous similarities between the two games:
In my opinion it can't be coincidence that these two games are so similar to each other. At Atari, it wasn't unusual for developers to pull ideas from other devs, and Atari would regularly hold brainstorming sessions to gather ideas which might get used for a new game.
There's an interesting interview with Atari alumni Michael Albaugh over at Atari Compendium where he briefly talks about Heart of Ice.
This is pure speculation, but here's the timeline of events I suspect occurred. The Ice World hardware appeared to be coming along well, but the game play involving driving wasn't working and they needed to come up with a new gameplay concept. And around the same time Heart of Ice has the 3D gameplay elements figured out, but work on hardware was slow and Cerny was distracted by Marble Madness. Putting two and two together, it seems that somewhere along the way the Heart of Ice gameplay concept was adopted by Ice World.
We still have the problem of figuring out where the idea for a robot came from.
The primitive polygonal graphics weren't well suited for the prince/princess theme in Heart of Ice, so it's not hard to see why the game setup and characters needed to be changed. The developers made some smart stylistic choices that worked within the limited capabilities of the hardware. It's not difficult to see why the game eventually ended up with an abstract "cubist" look - that style works well with angular polygonal graphics. The games unique visual style is clearly the result of the developers finding creative solutions to the problems they were encountering.
This reminds me of similar issues faced during production of the groundbreaking 1982 film Tron. There the animators had to create a digital world whose entire design aesthetic was in essence defined by the the abstract shapes and harsh angular features of the primitive CGI systems they were using. By working with those limitations, the team ended up creating a cohesive look and style for the film that is both striking and memorable. Many times the most interesting art comes from artists who find creative was to work around limitations forced up on them.
Lost Gameplay Ideas
Some marketing documents have been unearthed that add a lot more detail on the interim stages of game development. They detail features from early prototypes that were dropped during development. The first batch of documents relates to a focus group that was ran on November 4, 1982.
These documents show us that by November of 1982, an early prototype version of the final "robot" game concept was complete enough to run by a focus group. The 3D driving concept has clearly been killed off by this stage of development.
What's most interesting here is that you can see a lot of ideas they were playing with that didn't make it into the final game. For example:
It seems that the biggest ideas from the early prototype that were scrapped are:
Robot energy level, tied to collecting gold
Depositing of gold into an ATM at the end of the level
"scrolling forward as a timing device" -- which I interpret as meaning the level would slowly scroll forward to force you to keep moving forward (or else fall off?)
What's in a name?
At some point during development, the name Ice World was formally changed to I, Robot. We know that robot themed gameplay was firmly in place as of late 1982. Here's an August 25, 1983 document talking about the name change.
Note that the document talks about having the new name in place in time for the 1983 AMOA show, which was held in New Orleans from Oct 28-30.
It's not clear whether or not I, Robot actually made it's debut at the 1983 AMOA. Most of the press I can find on the show focuses on Atari's big LaserDisc, ahem... LaserVideo game, Firefox.
The copyright of 1983 in the ROMs suggests that the game was first put on field test in late 1983. The earliest date that can be pinpointed by Atari corporate docs shows that the game was in field test no later than February 1984. The gameplay at this stage resembles that of the final production game.
As we can see, player feedback describes the a game which plays very much like the final version of I, Robot.
Based on the timing, we can assume that most of the final gameplay features were developed throughout most of 1983, based on the feedback from the 12/82 focus group.
The game game was actually produced in cabinets that were left over from Firefox's production run, suggesting that Atari overestimated the demand for Firefox. Because of this, the bottom of the cabinet is completely empty, as this is where Firefox's LaserDisc player would have been. Also note the empty metal plate in the front of the cabinet below the control panel -- this is where the Firefox headphone jack and volume control were mounted.
June 1984 Release
I, Robot was released into production by Atari in June of 1984 (possibly June 6). Atari records show 750 units were manufactured at an original sales price (likely distributor price) of $1,995 USD. It was considered a major flop upon release. The video gaming public was unprepared for a 3D gaming experience, and the machines were notoriously unreliable due manufacturing issues and the failure rate of it's custom IC.
I, Robot was likely the last coin-op game released by the original "Atari, Inc.". On July 2nd 1984, having been battered by the great video game crash of 1983, the original Atari Inc. was split into pieces and sold off. The Consumer Division was sold to Jack Tramiel, leaving the coin-op division to continue independently under the new name "Atari Games Corporation".
The sale of Atari is seen as the unofficial end of the great video game crash, and in many ways the release of I, Robot coincides with the end of an era. Once the fastest growing company in history, Atari never fully recovered from after the sale and had to sit by while Nintendo took over the marketplace. But looking back, I, Robot reminds us of just how formidable of a force Atari was in it's heyday. It serves as a fitting swan-song to the golden age of gaming, the end of a wild journey that began with the release of Pong. And it shows that right up until the end Atari was determined to go out pulling no punches.
Rough Development Schedule
I've attempted to put together a rough outline of the development schedule. All dates, including estimates, are taken from Atari documents.
Comparison to Simutrek's Cube Quest
I, Robot was so far ahead of the curve that it's amazing the game saw the light of day at all. Atari had it's work cut out for them trying to bring 2-D gameplay into 3 dimensions. Breaking so much new ground both technically and gameplay wise, Atari could have easily ended up with a misfire, a forgettable "least common denominator" experimental arcade game. Although the game failed at the time, it's a remarkably fun game to play, an Atari classic that is still appreciated today.
To better understand how well Atari rose to the challenge, we need to consider the little known Simutrek 1983 arcade title Cube Quest. Rushed out the door in December 1983 and barely having a production run before Simutrek folded, Cube Quest is arguably in a neck-and-neck race with I, Robot to hold the title of first arcade game with 3D raster graphics. Like I, Robot, Cube Quest was also planned to debut at the 1983 AMOA, so it's possible they debuted at the same time. Hopefully someone can shed some light whether either of these games made an actual appearance. Cube Quest may have released first, but only because Simutrek needed to ship something before they went bankrupt (which they did in January 1984).
Simutrek's Cube Quest: Pretty backgrounds, but why must the polygons be so far away?
The Cube Quest hardware was very impressive. In addition to a polygon engine it also used a LaserDisc to stream pre-rendered CGI video loops behind the gameplay. The gorgeous backgrounds were created by Robert Abel & Associates (one of the companies that performed CGI work in Tron) and represented the state of the art at the time.
Although the hardware is impressive, from a gaming standpoint Cube Quest it's just a mess. The low-res primary color foreground polygons stand in stark contrast against the gorgeous high-res high-color backgrounds. The two styles don't work together at all. Worse, most of the game-play is focused at the center of the screen and the enemies are are mostly far way. With such small polygons ontop of a moving background it's hard to tell exactly what's going on half of the time. This looks and plays like a game that was unfinished and rushed to market. Which it was, as Simutrek had run out of funding and had no choice but to ship an unfinished game to avoid bankruptcy. [For those interested in more info there's a great interview with Cube Quest programmer Paul Allen Newell here).
It's unfortunate the Simutrek team didn't have the time to showcase what their hardware was capable of. Atari definitely took full advantage of the hardware's capabilities. I can just picture Theurer sitting at his computer terminal and grinning from ear to ear terminal as he watched the huge head of Big Brother fly onscreen for the first time. And regardless of what you think about the, um... questionable inclusion of Doodle City mode, it's obvious the development team loved showing off to the competition. They were engineering superstars, and they knew it.
Honestly, what's wrong with a little bragging when you consistently deliver games this good?
Atari getting in your face
The difference in overall playability between these two games reminds us that at it's peak, Atari was simply a force of nature in the coin-op industry. For over a decade they released hit after hit, relentlessly taking risks and continually breaking new ground technically and gameplay wise. They made it all look so easy, when it was anything but.
I, Robot serves as a fitting bookend to closing the door on the classic Atari, Inc. coin-op era. Here we see Atari's at its zenith: bringing together its excellence in engineering design, mechanical design, smart risk taking, creative gameplay innovation, and overall execution of product. No one did it better.
Hopefully what's written here won't be the final final word on the matter. If anyone has anything to add to this story, let me know.