Tag Archives: level 9


Mike and Pete Austin, Level 9's first two full-time employees

Mike and Pete Austin, Level 9’s first two full-time employees

The first weeks of 1983 brought the last gasp of Level 9 as a publisher of generalized software in the form of an implementation of the Forth programming language (a Great White Hope for programming in the early 1980s that never quite caught on as anticipated) for the BBC Micro. The disappointing sales of that product, contrasted with the ever-quickening sales of their three adventures, guaranteed that Level 9 would live or die by their adventure games from here on. Thus, over the course of 1983 the Austin family transformed Level 9 from a hobby into the company that has gone down in history as the “British Infocom,” the most long-lived, dedicated, and arguably most important British publisher of the text adventure’s relatively brief lifespan as a viable commercial proposition.

Unable to keep up with rising demand by duplicating each cassette by hand, they signed contracts to begin churning out their games in bulk, the way the pros did it. At the same time they replaced their Ziploc baggies and mimeographed cassette inserts with real boxes which included real instruction manuals. With their beloved Nascoms fading in importance in a computing milieu now dominated by the likes of Sinclair and Acorn, the brothers began doing their development work on the BBC Micro. However, the two younger Austin brothers, Mike and Nick, continued to develop A-Code interpreters for many other platforms, bringing Level 9’s games to an ever-widening slice of the British microcomputer market. By fall their games were available for the Nascom, the BBC Micro, the Sinclair Spectrum, the Atari 8-bit line, the Commodore 64, the Camputers Lynx, and the Oric-1, with yet more platforms soon to follow. Meanwhile the lone Austin sister, Margaret, took over marketing. With sales now substantial enough to pay for them, full-page spreads began appearing around this time in the major magazines in lieu of the little classified ads with which Level 9 had made do in earlier years.

At the center of this web of activity was the eldest brother, Pete, who quit his day job in June of 1983 to become Level 9’s first full-time employee, soon to be joined by the youngest Austin, Mike, who took a year off from his education before starting university to help with the company. It was Pete who actually designed the games, as well as being the de facto leader and day-to-day manager of the family business. What with being so busy with the logistics of getting Level 9 up and running in earnest, Pete found time to design just one game for Level 9 in 1983, which in turn became the company’s only new adventure of the year. Still, that game, an exercise in hard science fiction which he called Snowball, was an important one, first fruit of a deepening determination to create adventures that were coherent fictional experiences. That determination would set Level 9 apart from most of their peers in the time to come, and would provide, even more so than their reliance on a cross-platform virtual machine or their impressive technical standards, a good reason to call them the British Infocom.

Snowball advertisement

A comparison of Snowball to Peter Killworth’s science-fiction game of the same year, Countdown to Doom, might help to illustrate how Level 9 was now diverging from the Cambridge/Acornsoft nexus. The planet of Doomawangara makes absolutely no sense as a piece of world-building. Glaciers sit next to deserts next to jungles; it’s enough to make even a Star Trek writer blush. The game’s fictional context, like the world, like the conceit of needing to finish your work and get away before your ship disappears in 215 turns, exists merely as a frame for the devious, clever, surreal, inspiring, occasionally infuriating puzzles (including, of course, the meta-puzzle of solving all of those other puzzles within the time limit). There’s nothing wrong with such an approach; Countdown to Doom can provide quite a compelling experience for hardcore puzzle fans. Still, Pete Austin was aiming at something quite different in Snowball, something more demanding and perhaps ultimately more rewarding.

Level 9’s first trilogy of games had made occasional nods in the direction Snowball would now travel with determination. All three otherwise traditional fantasy puzzlefests are occasionally interrupted by odd, lumpy bits of exposition and plot that seem dropped in from some other creative endeavor entirely, as in the climax of Colossal Adventure when Crowther and Woods’s Adventure suddenly turns into a desperate mission to rescue 300 elves being held prisoner by a nation of evil dwarfs that apparently lives below Kentucky. But with Snowball the world-building really comes to the fore. A fan of the intellectual rigorousness of hard science fiction like that of Larry Niven (he listed The Mote in God’s Eye as a particular inspiration), Pete Austin strove with Snowball to craft a fictional environment that actually makes sense, that could make a coherent setting for a novel. As described in a detailed background section in the manual (shades of Infocom yet again), the game takes place aboard the eponymous Snowball 9, a massive colony ship carrying 1.8 million people, all in cryogenic sleep, on a journey of over a hundred years to the planet of Eden. The ship gets its name from the shell of ice which is constructed around it at the beginning of the voyage:

The chain of accelerators beyond Pluto burst erratically into life throughout the following three years, firing ten-tonne blocks of ammonia ice at precise speeds after the receding craft. Once reeled in by the Snowball’s skyhooks, the ice was built into a huge hollow shell around the linked passenger disks. When complete, this shielded the disks during the voyage, until the ice was finally needed as fuel for the ravening fusion drives.

As the above excerpt may illustrate, Snowball offers Big Ideas, and can inspire in the player the awe-inspiring, almost religious epiphany of coming face to face with wonders on a scale bigger than we as individual 21st-century humans can entirely fathom. And, as the above excerpt should also illustrate, the environment of Snowball is worked out with impressive care. In addition to its descriptions of the Snowball itself, the manual also includes the circumstances of her birth in the form of a history of humanity until the launch of Snowball and her 49 sister colony ships in the 2190s. None of this history bears all that directly on the game itself, but it does do much to make Snowball a holistic fictional experience, in a way that really only Infocom was also attempting to do at this date.

Surprisingly, all of this emphasis on coherence and (science-fictional) realism leads quite naturally to the most remembered, most mocked, and perhaps most misunderstood attribute of the game today: its geography of “over 7000 locations.”


When it comes up today, this data point is always followed by the same punch line: the fact that over 6800 of these rooms are essentially identical. Snowball is used in this context as the ultimate illustration of the absurdities of old-school adventures, with their mazes and sprawling geographies full of corridors and empty rooms. Still, it’s not really a fair portrayal of Snowball. Let’s remember that Pete Austin was trying to present a realistic depiction of this massive colony ship. Let’s further remember that such a ship will inevitably consist of endless identical rooms and corridors full of the 1.8 million sleeping colonists and the apparatus to maintain them. Snowball doesn’t indulge in mazes for the sake of them, and doesn’t ever demand that we map or visit any but the barest fraction of these rooms. Instead, a central puzzle is that of learning to use the ship’s manifest to identify and find the couple of rooms amongst this menagerie that you do need to visit. In this sense Snowball is anti-old-school: someone who tries to dutifully visit and map every single room in classic adventurer fashion, ignoring her role as a fictional actor in this world, is simply playing with the wrong mindset entirely. (That said, less defensible was Level 9’s decision to use the “over 7000 rooms” tagline as their centerpiece for marketing this “massive adventure.” For that they were roundly mocked even in the computer press of the time, and with good reason.)

In trying to make the leap from text adventure to interactive fiction, Pete Austin faced many of the same questions with which the Infocom authors were wrestling. In particular: to what extent could or should the author define the protagonist? Would players insist on playing “themselves” in adventure games, or would they be willing to play a role assigned to them, as it were, by the needs of the fiction? Pete created a detailed profile of the protagonist, Kim Kimberly, and included it in the manual. Kim’s gender, however, he left unspecified. When asked about his reasons for doing so, he often noted that as many as one-third of the people who played Level 9 games regularly were female. In deference to them, he made the decision to try to make all of his games “non-sexist.” It’s of course problematic in the extreme to assert that a game which merely features a male — or female, for that matter — protagonist is automatically sexist. Nor is it entirely clear that players, male or female, otherwise willing to accept inhabiting a role very different from themselves would determine that playing someone of the opposite sex would be the deal-breaker. In an interview for Micro-Adventurer, Pete went on to make the more straighforwardly sexist — or at least condescending — assertion that “adventure games offer many women, trapped at home by children, a more intellectual alternative to Mills and Boon.” Ah, well, at least the fellow was trying, which is more than could be said for most of his peers.

Progressive as it is in so many ways, Snowball still has its full share of problems and questionable elements. Like the three fantasy games that preceded it, it’s just much too hard, and too often for the wrong reasons. The most infamously awful puzzle involves a computer screen which is operated by looking at desired entries in a menu and then BLINKing, a completely unclued combination of the design sin of read-the-author’s-mind with just a dash of guess-the-verb. Just to really salt the wound, the first entry you read after — let’s be honest — looking the answer up in a walkthrough is… instructions on how to operate the computer. Grr…


The parser, while it does understand phrases of more than two words to an extent, is very unrefined compared to that of Infocom. You see, the parser here lies — or, if you like, cheats. When it encounters a phrase it doesn’t entirely understand, it attempts to guess the meaning from those words it does. This is a fraught approach, with huge potential to result in misunderstandings. You’re thus likely to see a lot of weird non-sequitur responses in your time with Snowball.

And then the actual plot that develops in the game itself, involving a hijacker who is attempting to crash the Snowball into the Eden system’s star for reasons that are never really explained just as it will soon be time to begin reviving the colonists and sending them down to their new lives at last, is oddly sparse in comparison to the meticulous back story.

All of these weaknesses can largely be attributed to the draconian technical constraints of the 32 K, cassette-based machines that represented the lowest common denominator for which Level 9 had to develop. I mention them here only so you won’t think that Snowball is possessed of quite the same sophistication and polish of a contemporaneous Infocom game. By comparison Infocom, with the luxuries of an extra 16 K of memory and floppy-disk-enabled virtual memory, had it easy. Still, Snowball packs a staggering amount of content into 32 K, and is in that light if anything even more technically impressive than Infocom’s games.

Indeed, when compared with the most obviously similar game in the Infocom canon, Starcross, Snowball acquits itself pretty darn well. While Starcross owes a hell of a debt to Arthur C. Clarke’s Rendezvous with Rama, Snowball is braver, more audacious in proposing a more original science-fictional concept and working it out so carefully. If its plot is ultimately a bit slight, well, plenty of science-fiction novels are also more interesting for their worldbuilding than their plot. I would even say that Snowball gets the sense of wonder of the best science fiction across better than Starcross. There are a few moments, such as when you exit the ship proper to the interior of the globe of snow and see this huge man-made wonder stretching out around you, that approach the transcendent even given the bare few sentences Pete Austin can spare to describe the scene. Anyone playing along at the time had to recognize the audacity that was lacking in most of Level 9’s peers, and had to be excited to see where it would take them. That’s a journey I’ll be continuing to follow here with relish.

For anyone playing along with these articles, I’ve prepared a zip file which includes Snowball in its BBC Micro disk incarnation as well as the original manual and a hints sheet, which, once again, you’re likely to need. Next time we’ll wrap up 1983 in Britain with the most important adventuring development of the year — which wasn’t even a game per se.


Tags: ,

Level 9

Before the likes of the Sinclair ZX80 and ZX81 and the Acorn Atom which I discussed in a previous post on British computing, there were the solder-them-yourself kits which began to arrive in 1978. The most long-lived and successful of these were the products of a small company called Nascom. The obvious American counterpart to the Nascom was the original kit PC, the Altair. That said, the Nascom was actually a much more complete and capable machine once you got it put together (no easy feat). It came, for example, with a real keyboard in lieu of toggle switches, and with video output in lieu of blinking lights. Like the Altair, the Nascom was open and flexible and eminently hackable, a blank canvas just waiting to be painted upon. (How could it not be open when every would-be user had to literally build her machine for herself?) In the case of the Altair, those qualities led to the so-called S-100 bus standard that, in combination with the CP/M operating system, came to dominate business computing in the years prior to the arrival of the IBM PC. In case of the Nascom, they spawned the 80-Bus architecture that could eventually also run CP/M, thanks to the Nascom’s use of the Zilog Z-80 processor that was also found in most of the American CP/M machines. A hardcore of committed users would cling to their Nascoms and other 80-Bus machines well into the 1980s even in the face of slicker, friendlier mass-market machines that would soon be selling in the millions.

One of the Nascom buyers was a 25-year-old named Pete Austin. He had finished a psychology degree at Cambridge when, “looking for an excuse to stay there for an extra year” before facing the real world of work and responsibility, he signed up for a one-year course in computing. He discovered he was very, very good at it. After finishing the course, he began a career as a programmer, mostly coding applications in COBOL on big-iron machines for banks and other big institutions. He quickly found that he wasn’t as excited by the world of business computing as he had been by the more freewheeling blue-sky research at Cambridge. But while programming accounting packages and the like may not have been exciting, it did pay the bills nicely enough. At least he earned enough to buy a Nascom for some real hacking.

After buying and building the Nascom, he spent a lot of time tinkering on it with his younger brothers Mike and Nicolas, both of whom were if anything even more technically inclined than Pete himself. Together the brothers developed a number of programming tools, initially for their own use, like a set of extensions to the Nascom’s standard BASIC and an assembler for writing Z-80 machine language. In 1981 they decided to try selling some of these utilities in the nascent British software market. They took out advertisements in a magazine or two under the name Level 9 Computing, a generic but catchy name that could refer to anything from an academic qualification to the lowest circle of Dante’s Inferno to a level in a videogame to a Dungeons and Dragons dungeon or character level. They were rewarded with a modest number of orders. Encouraged, they added some simple games to their lineup, mostly the usual clones of current arcade hits. More indicative of their future direction, however, was Fantasy, a sort of proto-text adventure written by Pete. Some earlier experiences had influenced its creation.

Already a dedicated wargamer, the young Pete had been introduced to Dungeons and Dragons while at Cambridge. He promptly became obsessed with D&D and another early tabletop RPG, Empire of the Petal Throne. He later said, “In the evening we either played D&D or went down to the pub… and played Petal Throne.” Still, it took him a surprisingly long time to connect his interest in computers to his interest in RPGs. Cambridge was the premier computing university of Britain, the atmosphere within its computer science department perhaps not terribly far removed from that at MIT. As such, there were plenty of games to be had, including some early proto-CRPGs obviously inspired by tabletop D&D. Pete toyed with them, but found them too primitive, underwhelming in comparison to playing with friends. (Ironically, Pete left the university just before the rise of the Phoenix mainframe text adventure boom, about which more in a future article.) The spark that would guide his future career wasn’t kindled until he was working in business computing, and had left D&D behind along with his old gaming buddies in Cambridge. On one of his employers’ systems, he stumbled across an installation of Adventure. Yes, now follows the story I’ve told you so many times before: long story short, he was entranced. Fantasy was the first product of his fascination. But Pete wanted to do more than create a stripped-down shadow of Adventure on the Nascom. He wanted to port the whole thing.

This was an audacious proposition to say the least. When Scott Adams had been similarly inspired, he had been wise enough not to try to recreate Adventure itself on his 16 K TRS-80, but rather to write a smaller, simpler game of his own design. A year after Adams’s Adventureland, Gordon Letwin of Microsoft ported the full game onto a 32 K TRS-80. The Austins also had 32 K to work with, but they lacked one crucial advantage that Letwin had been able to employ: a disk drive to fetch text off of disk and into memory as it was needed during play. With only a cassette drive on their Nascom, they would have to pack the entire game — program, data, and text — into 32 K. It looked an impossible task.

Meanwhile the Austins were mulling another problem that will be familiar to readers of this blog. The burgeoning British PC industry was in a state of uncertain flux. In yet another piece of evidence that hackers don’t always make the best businessmen, Nascom the company had suddenly collapsed during 1981. They were rescued by Lucas Industries, famed manufacturers of the worst electronic systems ever to be installed into automobiles (“Lucas, the Prince of Darkness”; “If Lucas made guns, wars would not start”; “Why do the British drink warm beer? Because Lucas makes their refrigerators!”), but their future still looked mighty uncertain in the face of the newer, cheaper computers from Sinclair, Acorn, and Commodore that you didn’t have to solder together for yourself. Wouldn’t it be great if the Austins could devise a system to let them run their game on any computer that met some minimal specification like having 32 K of memory? And what if said system could be designed so that games written using it would actually consume less memory than they would if coded natively? We’ve already met the P-Machine and the Z-Machine. Now, it’s time to meet the A-Machine (“A” stands for Austin, naturally).

In some ways Level 9’s A-Code system is even more impressive than Infocom’s technology. Although Level 9’s development tools would never quite reach the same level of sophistication as Infocom’s with their minicomputer-based ZIL programming language, the A-Machine itself is a minor technical miracle. While Infocom also targeted 32 K machines with their earliest games, they always required a disk drive for storage. The Austins lacked this luxury, meaning they had to develop unbelievably efficient text-compression routines. They were understandably tight-lipped about this technology that as much as anything represented the key to their success during their heyday, but they did let some details slip out in an interview with Sinclair User in 1985:

Pete’s text compressor has been a feature of all Level 9’s mammoth adventures. It works by running through all the messages and searching for common strings.

For example, “ing” might occur frequently. The compressor replaces “ing” with a single code wherever it occurs. That done, it goes through again, and again, each time saving more space. “It doesn’t always pick up what you’d expect it to,” explains Pete. In the phrase “in the room” the compressor might decide that it was more efficient to use a code for “n th” and “e r” rather than pick out “in” and “the.” That is not something which occurs to the human mind.

The Austins used a similar technique in their actual A-Machine program code, condensing frequently used sequences of instructions into a single virtual-machine “opcode” that could be defined in one place and called again and again for minimal memory overhead.

Having started with a comparison to Infocom’s technology, I do want to remind you that the Austins developed the A-Code system without the same pool of experience and technology to draw upon that Infocom had — no DEC minicomputers for development work, no deep bench of computer-science graduate-degree-holders. The youngest brother, Michael, was not even yet of university age. Yet, incredibly, they pulled it off. After some months they had an accurate if not quite word-for-word rendering of the original Adventure running in A-Code on their faithful Nascom.

But now Pete realized they had a problem. Level 9 had taken out some advertisements for the game describing its “over 200 individually described locations.” It sounded pretty good as ad copy. Unfortunately, they had neglected to actually confirm that figure. When they sat down and counted at last, they came up with just 139. So, determined to be true to their word, they decided to start squeezing in more rooms. They replaced the original’s simple (if profoundly unfair) endgame in the adventuring repository with an extended 70-room sequence in which the player must escape a flood and rescue 300 Elvish prisoners. Just like that Level 9 had their 200-room game, and they continued to trumpet it happily, the first sign of a persistent obsession they would have with room counts in the years to come. (The obsession would reach comical heights in 1983 with the infamous 7000-room Snowball and its 6800 rooms of identical empty spaceship corridors.)

The Austins released Colossal Adventure on the Nascom in early 1982, selling it by mail order through magazine advertisements. They copied each order by hand onto a store-bought TDK cassette tape. Into the TDK case they shoved a tiny mimeographed square of paper telling how to load the game. Scott Adams’s original Ziploc-bag-and-baby-formula-liner packaging was sophisticated by comparison. But they sold several hundred games in the first few months. And, with the work of tools development behind them, they were able to follow up with two more equally lengthy, entirely original sequels — Adventure Quest and Dungeon Adventure — before the end of 1982.

Colossal Adventure in particular makes for an interesting experience today. Prior to the endgame, it’s mostly faithful to its inspiration, but there are just enough changes to keep you on your toes. To create a context for the endgame, Pete grafted a plot onto the original game. You are now exploring the caverns at the behest of an elvish warrior. The axe-throwing dwarfs that haunt the caverns are not just annoying, but Evil; it is they who are holding the elf’s people prisoner. Certain areas are re-purposed to fit the new plot. The Hall of the Mountain King, in the original a faithful reconstruction of a cavern Will Crowther knew from his journeys into Bedquilt Cave, becomes here a sinister monument to the dwarfs’ conquests; Spelunker Today becomes a dwarven propaganda rag.

Pete did tinker here and there with the structure of the game as well. The outdoors are fleshed out quite a bit, with additional locations and (naturally) an additional maze, and a few familiar items are found in different — usually less accessible — places. Whether out of a sense of mercy or because his game engine wouldn’t support it, he also eliminates the need to respond “YES” to solve the dragon “puzzle.” Much less mercifully, he inexplicably reduces the inventory limit to just four objects, which makes everything much, much more difficult than it ought to be, and makes finishing the game without buying more batteries for the lantern (and thus getting the full score) well-nigh impossible. The inventory limit also makes mapping the several mazes even more painful.

For its part, the endgame is absurdly difficult, but it also has a sense of onrushing momentum that was still rare in this era. Literally onrushing, actually; you are trying to escape a massive flood that fills the complex room by room. It’s impressive both from a storytelling and a technical perspective. For all their old-school tendencies, Level 9 would always show a strong interest in making their games narrative experiences. Dungeon Adventure and, especially, Adventure Quest show a similar determination to present an actual plot. The latter takes place decades after the events of Colossal Adventure, but begins on the same patch of forest. It does a surprisingly good, almost moving job of showing the passage of time.

It soon evolves into a classic quest narrative that could be torn from Greek mythology, with the player needing to make her way through a series of relatively self-contained lands to arrive at “the Black Tower of the Demon Lord.”

Sprinkled increasingly liberally through the three games are references to Tolkien’s Middle Earth — Black Towers, balrogs, High Councils, Minas Tirith. Still, they never feel so much like an earnest attempt to play in Tolkien’s world as a grab bag of cool fantasy tropes. It almost feels like Pete kind of wandered into Middle Earth accidentally in his quest for Cool Stuff to put in his games. As he later said, “Middle Earth was a convenient fantasy setting. It was a way of telling people the type of world they were getting.” Where another milieu offers something equally cool, he uses that; Adventure Quest, for example, features a sandworm straight out of Dune. All of this was, of course, completely unauthorized. After not mentioning the Tolkien references in early promotion, Level 9 actually advertised the games for a while as the Middle Earth Trilogy. Then, presumably in response to some very unhappy Tolkien-estate lawyers, they went the other way, excising all of the references and renaming the games the Jewels of Darkness trilogy.

Before I leave you, I just want to emphasize again what an extraordinary achievement it was to get these games into 32 K. Not only are they large games by any standards, brimming with dozens of puzzles, but — unlike, say, the Scott Adams games — the text also reads grammatically, absent that strangled quality that marks an author trying to save every possible character. Better yet, Pete knows how to create a sense of atmosphere. His prose is blessedly competent.

That’s not, however, to say that I can really recommend them to players today. In addition to quality prose, they’re also loaded with old-school annoyances: a two-word parser (in their original incarnation; the parser at least was updated in later releases); mazes every time you turn around; endless rinse-and-repeat learning by death; and always that brutal four-item inventory limit. They were damn impressive games in their time, but they don’t quite manage to transcend it. Fortunately, there were lots more adventures still to come from Level 9.

In which spirit: by the end of 1982, Level 9 had, thanks to the magic of A-Code, leaped from the Nascom onto the Sinclair Spectrum and the BBC Micro, ready to ride a full-on adventuring craze that would sweep Britain over the next few years. We’ll start to talk about the new machines that would enable that — including the latter two in the list above — next time.