05 Aug


What with our love of specialization and our tendency to place societal value only on those activities that earn their doers money, the modern world is not terribly kind to the idea of amateurism. See, for example, the popular adjective “amateurish,” which has doubtless crept into a few of my own reviews of one misbegotten game or another over the years. Yet there was a time when amateurism had a more positive connotation, connected to its root in the Latin “amatorem”: “lover.” This view of amateurism celebrates the amateur as someone who does something not for material gain or social prestige, but purely for the love of it. It’s this view which lies behind the requirement that persisted for so many decades that only amateurs — pure of mind, intent, and spirit — could take part in the Olympic Games, a requirement so out of keeping with the currents of modern thought that we now find it difficult to even parse.

But really, truly, sometimes amateurism can bring out the best in us. Leaving to the sports bloggers the question of whether the Olympics lost something when they gave up on amateurism, I’ll say here today that the rise of the amateurs may just have been the best thing ever to happen to the text adventure. Interactive fiction had come just about as far as it could as a slave to the commercial marketplace by the time Infocom died in 1989. Free from the need to shoehorn in graphics, to artificially extend the length of the game, to avoid controversial topics and interesting formal experiments, the amateurs were now primed to take over. With this article, we’ll begin the deeply amateurish story of how that came to be — a story every bit as fascinating and inspiring as the history of Infocom that has consumed so much of my attention over the last four and a half years of blogging.

I mentioned already in my last article my fondness for Jeremy Douglass’s alternative take on interactive-fiction history, which replaces the tragic narrative of a lost commercial golden age with a more even-keeled thoroughgoing narrative of inspired amateurism, leaving the commercial era as an anomalous interstitial period within a longstanding non-commercial “literary salon culture.” Yet I’m not sure that even the anomalous commercial era was quite so anomalous as Douglass might lead us to believe, for the vast majority of people writing text adventures even during the 1980s weren’t trying to make a living at it. Right from the moment that Adventure inspired teams of institutional hackers in various places around the country and the world to try to make a better version of their own, the text adventure has been almost unique in the way it has inspired so many consumers to become producers of their favorite ludic genre. Indeed, that might just be the text adventure’s most inspiring aspect of all. And, far from being strictly a post-1980s phenomenon, it was as true during the alleged commercial era as it remains today.

We’ve seen already how enabling others to roll their own text adventures became something of a cottage industry of its own almost as soon as Scott Adams’s Adventureland, the first microcomputer-based game of its type, arrived on the TRS-80 in late 1978. In the May 1979 issue of SoftSide magazine, Lance Micklus published the BASIC code to his Dog Star Adventure, which would go on to become the sturdy framework for tens of thousands of similar games created in bedrooms, high-school computer labs, and home offices, most of which never traveled much further than the computer used to create them. There followed heaps more magazine listings of BASIC-based text adventures, many of them based on Micklus’s original, along with enough books on the crafting of them to fill a long library shelf.

At the same time, other more specialized text-adventure-creation systems also began to emerge. Only shortly after Dog Star Adventure appeared, Donald Brown began to distribute Eamon, a construction set for “adventure scenarios” that combined elements of text adventures and CRPGs in a way that strikes us as stranger today than it did players back in 1979, when genres were still in flux and the lines between the largely deterministic adventure game and the largely emergent CRPG had not yet been clearly demarcated. In 1982, Allan Moluf and Bruce Hanson released The Adventure System, an authoring system for new games that used the now well-documented Scott Adams format. (Such clever retro-fitting of new technologies onto old would become another thoroughgoing theme of interactive-fiction history, with perhaps the most notable example becoming Graham Nelson’s decision to make his new Inform compiler target the old Infocom Z-Machine in the early 1990s.) But the really big winner among the early text-adventure construction kits was an entry from across the pond: Graeme Yaendle’s The Quill, which almost from the instant of its late 1983 release became the basis for more than one out of every two text adventures released in Britain, thus helping the adventuring culture of Britain and the wider Europe become even more a culture of inspired amateurism than was that of the United States. (The Quill was briefly sold in the United States under the name AdventureWriter, but, being poorly marketed and distributed, never caught on.)

With Infocom being universally regarded as the gold standard in text adventures by 1983 even in Europe (where their games would remain pricey, disk-drive-requiring pieces of foreign exotica for some years yet), the ultimate dream for makers and users of systems like The Quill was a construction kit capable of making Infocom-quality games. That dream would, alas, go unrealized for quite some time. Even the Professional Adventure Writing System (PAWS), Yeandle’s 1986 follow-up to The Quill, had a long way to go to meet that standard. The first popular construction kit capable of fooling a player even momentarily into believing she was playing an Infocom game would rather be an American system called the Adventure Game Toolkit, or AGT, the first system of its kind to be widely used in the United States since Eamon. But before AGT there was the less auspiciously named Generic Adventure Game System, or GAGS, which never became all that popular and was capable of fooling absolutely no one. Nevertheless, any history of AGT must begin with GAGS.

The man behind GAGS was one Mark J. Welch, a go-getter who had started working as a freelance technology journalist just a year out of high school in 1976. By the mid-1980s, he was an editor and product reviewer for Byte, the most influential periodical in the computer industry. In his spare time, he tinkered with text adventures, first in BASIC, then in a Pascal-like language called Ada, finally in a new programming environment that was taking the industry by storm, Borland’s Turbo Pascal. He eventually made the leap that Scott Adams and so many professional adventure developers after him had already made, from writing each new game as a custom program on an ad hoc basis to developing a database-driven engine capable of playing many games. That engine became GAGS, which Welch first released in 1985 under the new software-distribution model of shareware; the system was free to copy or download, but happy users were encouraged to register their copy for $15. Welch would also sell the Turbo Pascal source code to those looking to hack the system a bit for themselves for $25, on the condition that they not distribute it further.

Working on business-focused MS-DOS machines with disk drives, relatively fast processors, 80 columns of text, and 256 K or more of memory, Welch was far less constrained than the likes of Graeme Yeandle had been in writing The Quill on a little Sinclair Spectrum. Yet, far from actively chasing the Holy Grail of an Infocom-like game, he kept his goals for GAGS deliberately modest. In a passage as commendable for its honesty as it is questionable in terms of marketing, Welch wrote bluntly in his manual’s introduction that GAGS “cannot be used to write an adventure game with as many complex features as Infocom’s. To do so would require developing a complete adventure-game programming language, as Infocom has done, and would require adventure-game writers to learn a very complex set of rules.” With Welch having thus explicitly rejected the idea of a true programming language for creating text adventures, working with GAGS, like working with The Quill, felt more like an exercise in data entry than it did an exercise in programming. Let’s briefly see how it worked.

A GAGS game is made up of three elements: “rooms,” “nouns,” and “creatures.” All are defined in a single text file, which is then passed to the compiler for transformation into an Infocom-style story file, playable via the GAGS interpreter. The examples that follow are taken from the most popular game ever made with GAGS, a Lewis Carroll pastiche with the appropriately Victorian long-winded title of The Adventures of Alice Who Went through the Looking Glass and Came Back Not Much Changed. [1]I have tinkered here and there with this example game and the next to better illustrate my points. Let’s first look at a room definition.

  Denser Forest
  East      35 - Dense Forest
  Points 12

ROOM_DESCR 9 -- Denser Forest
The forest is even denser here.  You think perhaps you should turn around and
go back to the garden.  Suddenly you realize the solution to a mathematics
problem that has been eluding you for days.

I trust that none of what you see here is terribly hard to divine. Room 9 is called “Denser Forest,” and is a dead end lying to the west of “Dense Forest.” The player receives 12 points upon entering this room for the first time. And after these details we have the text of the room’s description.

Just a couple of other optional details beyond what you see here can be provided in a room definition, defining potentially locked doors blocking departure and the keys that unlock them as well as whether the room is lit or unlit (it’s lit by default). GAGS also lets us end the game, in either victory or death, as soon as the player enters a given room; in fact, this is the only way to declare victory.

“Nouns,” meaning inanimate objects really, are a bit more complicated. Here we have an umbrella.

NOUN 204
  There is an old umbrella here.
  Weight    1
  Size    1
  Location  5 - Drawing Room
  Points    0
  Key       0
  ** Can't shoot
  Num_Shots 0
  ** Not_a_light

The umbrella has obviously seen better days.  It is frayed around the edges,
the handle is warped, and it bulges unnaturally in several places.

You spin the umbrella around a few times. Wasn't that fun?

I should first of all note that nouns are numbered from 201, so this is actually only the fourth object defined in the game. We see here that “umbrella,” the one-word name of the noun, can be prefaced by the adjective “old,” which could potentially be very useful for parser disambiguation if there happen to be two umbrellas in the game. “There is an old umbrella here” is the message which will be appended to the description of a room that happens to contain the umbrella. It has a weight and a size of 1 for purposes of inventory juggling, and it begins the game in room 5, the “Drawing Room.” The umbrella awards the player no points for picking it up for the first time.

The parameters that come next neatly summarize virtually every possibility for object interaction in a GAGS game. A noun can be a key unlocking a certain door or another noun; this one is not. A noun can be shootable, possibly with limited ammunition. A noun can be pushable, readable, turnable, playable, or pullable, each verb leading to a simple textual description of the result; in this case, turning the umbrella yields, “You spin the umbrella around a few times. Wasn’t that fun?” A noun can be turned off or on, can be closable and if so either closed or open, can be lockable and if so either locked or unlocked, can be drinkable or edible and if one or the other can be poisonous. A noun can be moveable or unmoveable; in this case, somewhat oddly, the player can’t actually pick the umbrella up, which renders the weight and size parameters moot. Finally, a noun can be a light source.

Let’s look at one more noun.

NOUN 205
  There is a small dagger here.
  Weight    1
  Size    1
  Location  204 - UMBRELLA
  Key       0
  Points    5
  ** Can't shoot
  Num_Shots 0
  ** Not_a_light

The dagger is small but very sharp.

So, the dagger is a moveable object found inside the closed umbrella. This is the sort of thing that often passes for a puzzle in a GAGS game; note that the player is awarded 5 points just for picking the dagger up for the first time.

Let’s now have a look at a creature.

  There is a Froobious Bandersnatch here.
  Location 9
  Weapon   205
  Threshhold 5
  Time_Thresh 5

The Bandersnatch is snorting and drooling.  It seem to harbor no very
honorable intentions towards you.

Creatures are numbered from 301, so our froobious bandersnatch is actually the first creature to be described. He lives in room 9, our aforedescribed Denser Forest. The parameters that follow “Location” serve to drive home the real main interest of the GAGS system: combat. In that respect, GAGS would seem to owe more philosophically to Eamon than it does to Infocom. Yet its combat is implemented in a rather peculiarly  unsatisfying way even in comparison to that less than ideal older system. Each hostile creature has a single noun — a key to a different sort of lock, one might say — that is capable of killing it instantly. All others are useless against it. In the case of the bandersnatch, then, said noun is the dagger. The “Threshold” parameter tells us that the player can make up to 5 unsuccessful attacks — presumably, using various objects to poke at it, hoping to find the right one — before the bandersnatch gets aggravated enough to kill her. The “Time_Thresh” of 5 tells us that the bandersnatch will kill the player anyway if she remains in his presence for 5 turns; in other words, no matter what the player does, if she doesn’t attack the bandersnatch with the dagger within 5 turns of arriving in the Denser Forest, she’s a goner. And “Hostile” of course tells us that the bandersnatch is, well, hostile; if he was “Friendly” instead, the “Threshold” and “Time_Thresh” would be meaningless.

With all possible verbs and interactions hard-coded into the engine, authors working with GAGS were largely restricted to the hoariest of hoary text-adventure chestnuts: locks and keys, light and darkness, the inevitable mazes. But most of all, games tended to develop (devolve?) into an exercise in finding the right thing to use to bash each monster in the player’s way. That’s certainly the case in the game we’ve just been looking at, which sees Alice cutting a bloody swathe through Wonderland, taking out jabberwocks and knaves with a violent aplomb that would make even the makers of the recent girl-powery films blush. Even a croquet ball gets coded up as a monster to be bashed with an “umbrolly” — a different umbrella from the inexplicably unmovable one we saw earlier — in a clumsy attempt to work around the limitations of GAGS.

Unsatisfactory though it was in so many ways as a piece of hard technology, GAGS did have a number of softer qualities going for it. Young though he still was, Mark Welch had been working for almost a decade already as a technical writer by the time he released GAGS. He knew how to describe how his system worked clearly and completely, and how to put his best foot forward generally when it came to every aspect of GAGS’s presentation. One might even say that his extensive documentation of GAGS marks the beginning of the tradition of literate, thoughtful writing about the art and technology of interactive fiction that remains a welcome aspect of the amateur community to this day.

Another move by Welch would prove if anything even more influential. Looking to encourage people to use his system, he decided to sponsor a contest for GAGS games, with a submission deadline of September 30, 1986. His justification for doing so rings in concert with the justifications for the hundreds of interactive-fiction contests that have been held since, not least the big one that began in 1995: to get people to actually finish and publish their games. “The main purpose of the contest was to encourage people to share the games they’d written using GAGS,” he says. “A lot of people had started games, spending quite a few hours on them, but stopped before they’d really finished them, or before they’d really made them playable.” Entrants were mailed on disk to Welch, who judged them personally, selecting Alice as the winner. (There is unfortunately no historical record of how many and which other games were submitted for judging.) The author of Alice, one Douglas Asherman, received $100 for his efforts and got to see his game distributed with GAGS from then on. Alice received a further boost when some enterprising hackers worked out how to make the GAGS interpreter output and input text over a modem connection. The game became a fixture among so-called “door games” — simple, text-based games hosted by computer bulletin-board services — for several years thereafter, thereby becoming available for play by many owners of the non-MS-DOS computers that still made up the vast majority of the home market going into the second half of the 1980s.

David M. Malmberg

David Malmberg

It was shortly after Welch’s first and only GAGS contest that David Malmberg, the real hero of our story today, entered the scene. A business analyst by day at the San Francisco headquarters of Foremost-McKesson — a $4 billion corporation with fingers in such diverse pies as chemicals, liquor, food, and pharmaceuticals — Malmberg had long since become a microcomputer convert by night, having purchased his first Commodore PET well before the 1970s were history. He contributed regularly to the magazines, beginning with the very first issue of Compute!, the magazine destined to become home computing’s biggest, in the fall of 1979. With two small children at home, he developed a special interest in computers as teaching tools, particularly as tools for teaching programming. In 1982, he wrote and published through Human Engineered Software— a company he had been instrumental in funding as part of his day job — an application/toy for the Commodore VIC-20 called Turtle Graphics, a simple language for graphics programming aimed at children. Eventually ported to the Commodore 64 as well, Turtle Graphics wound up selling 80,000 copies in the heyday of the first home-computer boom. Meanwhile Malmberg also nurtured a strong interest in text adventures, publishing a few of his experiments in that direction as BASIC and PILOT type-in listings in the magazines.

Given his interest in text adventures and programming systems suitable for the non-expert, Malmberg was the perfect candidate to embrace and extend GAGS. As soon as he encountered Mark Welch’s simple adventure system, he mailed off for the source code and started studying it on his daily one-hour rail commute to and from work, teaching himself Turbo Pascal as he did so. He soon contacted Welch to inform him of the enhancements he had already begun to implement, looking to see if a partnership might be a possibility. But when he met Welch for lunch he found the latter interested not so much in partnering as in washing his hands of the whole endeavor. Welch, who had recently scaled back his work as a tech journalist to begin attending law school at Berkeley, was simply burnt-out on a system that, despite the contest and the relative popularity of Alice among BBS sysops and public-domain software traders, had never quite taken off as he had hoped. Malmberg promptly made him a very generous offer indeed: he would take the entire operation off his hands, assuming all responsibility for enhancing and supporting the system, but continue to pay Welch 50 percent of all the profits it generated. Welch, unsurprisingly, jumped at it. The sharply limited GAGS was about to become the much more useful AGT.

The core of Malmberg’s enhancements is a system of something he called “meta-commands,” which allow the author to make things happen that aren’t hard-coded into the system’s preconceptions. An AGT game which uses meta-commands has a new “command file” and a “message file” to go with the standard GAGS database definitions. The command file consists of a long string of entries to be checked against the player’s typed command. When matches are found, customized things of the author’s choosing can happen, possibly outputting custom text from the message file.

The examples that follow are drawn from the source code to A Dudley Dilemma, an early AGT game for which I’ll be providing more context soon. We’ll look at some of the code which implements Dudley‘s light source, a flashlight, in a much more sophisticated way than would have been possible in GAGS. Since A Dudley Dilemma is still very old-school in spirit, the implementation will include a dwindling battery, a complication that wouldn’t have been possible in GAGS.

In the standard database-definition file, we have two versions of the flashlight, representing it in its unlit and lit states. (While you read over the definitions that follow, do note the new ability to provide additional synonyms for nouns beyond a simple name and adjective — one of Malmberg’s various other modest but welcome enhancements to the core of the old GAGS system.)

NOUN 201
  There is a rechargeable flashlight here.
  SIZE 9

The flashlight is one of the rechargeable kind that never seems to work when
you need it.  It is off at the moment.

NOUN 202
  There is a rechargeable flashlight here.
  SIZE 9

The flashlight is one of the rechargeable kind that never seems to work when
you need it.  It is presently on.

Now, let’s turn to the command file.

  Present 201
  TurnFlagON 2
  SwapLocations 201 202
  PrintMessage 14

Every turn, the game checks to see whether the player’s command matches “light flashlight,” or appropriate synonyms thereof, as it works its way down through the list of meta-commands. If so, a script written in a custom programming language of about 170 possible commands is run — exactly the “complete adventure-game programming language” that Welch had so explicitly disavowed in his original GAGS documentation. If we have a match, we first check to see if noun 201, the unlit flashlight, is “Present,” meaning it is in the current room or being carried or worn by the player. If it isn’t, the script is terminated right here, and the game proceeds on to test the player’s command against the remaining commands in the file. If it is, we continue with the script by turning flag 2 on. The flags, another of Malmberg’s additions, are 255 on/off switches that the author can use to keep track of whatever she wishes. In this case, flag 2 tracks the state of the flashlight.

The next command, “SwapLocation,” swaps the locations of the unlit flashlight, which is being carried by the player or at least in the room with her, with the lit flashlight, which is presumably in room 0, an inaccessible holding area for nouns and creatures not currently in play. Next we print message 14 from the custom-message file. It looks like this:

You turn the $NOUN$ on.

“$NOUN$” here is, as you may have gathered, a stand-in for the object of the player’s command, in this case the flashlight.

The final line of the script, “DoneWithTurn,” informs the game that we’ve successfully carried out the player’s command and that it should not continue working its way through the command file checking for further matches, as it would have done had we bailed on the first line of the script. Thus we can effectively prioritize certain reactions over others by placing them earlier in the command file.

So, we now have a flashlight that we can turn off and on, but we still need to figure out how to make its battery run down. In addition to the 255 flags at our disposal, we have 25 variables, each capable of storing an integer. We use variable 1 to represent the current charge of the flashlight; it starts at 100 and should decrement by 1 each turn that the flashlight is lit. Thus very early in the command file, we have this:

  FlagON 2
  SubtractFromVariable 1 1

“COMMAND ANY” means that this script will fire every single turn, regardless of what command the player has entered. In the script that follows, we first check to see if flag 2, representing the state of the flashlight, is set. If it isn’t, meaning the flashlight is turned off, we bail, continuing with the later command entries. If it is, meaning the flashlight is turned on, we proceed to subtract 1 from variable 1. Note that there is no “DoneWithTurn” entry in this script, meaning processing of the command file will always continue whether we’ve actually done anything here or not.

The command definitions that immediately follow the one above give warnings to the player as the battery runs lower and the flashlight dims. I’ll leave decoding them as an exercise for you.

  FlagON 2
  Present 202
  VariableEquals 1 50
  PrintMessage 10

** The flashlight seems to be getting dimmer! **

  FlagON 2
  Present 202
  VariableEquals 1 20
  PrintMessage 11

** The flashlight is definitely much dimmer now. **

  FlagON 2
  Present 202
  VariableEquals 1 5
  PrintMessage 12

** The flashlight fades to a dull orange. **

Finally, the flashlight may expire entirely:

FlagON 2
VariableEquals 1 0
TurnFlagOFF 2
SwapLocations 202 233
Present 233
PrintMessage 13

** The flashlight sputters and goes out! **

Note that in this case we swap out the lit flashlight for yet another noun representing the same object, this time a dead version of the flashlight.

NOUN 233
  There is a dead flashlight here.
  SIZE 9

I’d like to look at just one more aspect of Dudley‘s handling of light and darkness, one that illuminates (sorry!) yet one more of Malmberg’s critical additions to the GAGS template. For the first time with AGT, it became possible for the author to define her own new verbs to augment the modest suite of text-adventure staples that are built into the system. Malmberg gave custom verbs the counter-intuitive name of “dummy verbs,” presumably because they do nothing if the author doesn’t explicitly handle them in her game’s command file. A Dudley Dilemma uses 30 dummy verbs, of which we’ll look at just one: “charge,” also known as “recharge,” “plug,” and “insert.” For A Dudley Dilemma, being old-school but not completely heartless, does give the player a way to recharge her flashlight.


In the original version of A Dudley Dilemma, the flashlight can be recharged from only one location, one which happens to be equipped with a handy electrical outlet.

  Round Room
  NORTH 24
  SOUTH 47
  EAST 50
  WEST 49

This is a roughly circular room with exits to the north, south, east and west.
There are several old steam pipes and cracked conduits running through here,
and the fumes from one of them make you slightly dizzy.
There is an electrical outlet here.

I won’t belabor the implementation of charging unnecessarily, as the code is quite readable on its own. When the player types something like “recharge flashlight,” we first step through a series of possible failure states and their resulting messages, arriving eventually if none of them apply at actually topping off the flashlight and sending her on her way. Once again, I encourage you to take a few minutes to work your way through it if you’re at all interested in getting a feel for how AGT really worked in practice. And besides, if you’re like me you might just find this sort of thing fun.

  IsCarrying 201
  PrintMessage 81

The flashlight must be ON to be recharged.

  IsCarrying 202
  VariableGT 1 20
  PrintMessage 67

The flashlight doesn't need to be recharged (yet).

  NOT AtLocation 48
  IsCarrying 233
  IsCarrying 202
  VariableLT 1 20
  PrintMessage 68

Guess you'd better find someplace to plug it in!

  AtLocation 48
  IsCarrying 202
  VariableLT 1 20
  SetVariableTo 1 100
  PrintMessage 66

You plug the flashlight in and recharge it.

  AtLocation 48
  IsCarrying 233
  SwapLocations 233 202
  TurnFlagON 2
  SetVariableTo 1 100
  PrintMessage 66

I hope these examples may begin to convey how ingenious Malmberg’s extensions to GAGS really were; they turned a system useful only for making the most simplistic of games into one of the most powerful systems for making amateur text adventures that had yet been seen.

But AGT had still more going for it beyond its technical affordances. Perhaps even more so than Mark Welch, David Malmberg had the skills to present his brainchild in the best possible light. He registered a little company, which he dubbed Softworks, to handle the system, and expanded Welch’s original documentation to fully explain all of his own new additions, replete with examples. Asked in an interview years after AGT’s heyday what aspects of the system he was proudest of, he placed “the quality of the documentation” on the same level as the meta-commands. Justifiably so: the instruction manual ballooned to 223 printed pages of friendly, readable prose, dwarfing in both quality and quantity the manuals included with most boxed commercial software. In fact, AGT became in a sense a physical product. Those who paid the $35 registration fee were shipped not only three disks full of the latest version of the compiler along with heaps of sample code and sample games but also the manual in spiral-bound hardcopy. All the effort spent in looking serious paid off in others choosing to take AGT seriously. In what can only be described as a major coup for Malmberg, AGT was given a full-fledged and generally very positive review in the February 1989 issue of the glossy newsstand magazine Computer Gaming World, an achievement of which very few other shareware products could boast.

Despite such welcome exposure, Malmberg quickly learned, as would many text-adventure fans who attempted to turn their loves into business propositions after him, that there just wasn’t a lot of money to be made in text-only adventure games. Able to encourage his users to register only by offering them the hardcopy documentation, telephone support, and, as he put it, “a warm glow from having supported at least one of the many shareware products you probably use” along with his own “eternal gratitude,” he would over the lifetime of AGT average only 100 or so copies “sold” per year — and even that income he of course had to split equally with his silent partner Mark Welch. At least an order of magnitude more people used unregistered copies to experiment with game design, and often to make and release complete games without ever bothering to pay the registration fee. Superficially professional though Malmberg’s presentation managed to be, AGT was always at heart a labor of love; it certainly never gave him cause to think of quitting his day job in favor of becoming a text-adventure mogul.

The people who made games with AGT, especially in the earliest years, were a far-flung, disconnected group by any standard, and doubly so in comparison to the close-knit Internet-based community that would follow them. I spoke to two former AGT authors in preparing this article and the next, and was surprised to learn that neither felt himself to be a part of any community at all really. With modems still fairly scarce and online services still fairly expensive in the late 1980s, Malmberg himself became the principal conduit binding many AGT users together, to whatever extent they were bound together at all. In 1988, shortly after releasing the first version of AGT, he made the hugely important decision to reconstitute Mark Welch’s one-off 1986 GAGS contest as an annual event. The rules were simple: all entrants had to have been made using AGT, had to have been first publicly released during the calendar year of the contest in question, and had to be posted to Softworks by December 31 of said year. Malmberg personally would, as he put it, “consider each game’s originality, cleverness, fiendishness, humor, raw cunning, and professionalism” to arrive at a winner.

The institution of the AGT contest as an ongoing annual tradition was a landmark event. Since 1989, the year when the 1988 batch of AGT games was judged, not a year has passed without a major annual contest dominating the interactive-fiction calendar. Until 1994, that contest would be Malmberg’s AGT competition, which would be held six times in all. Picking up neatly thereafter in 1995, it would be the Interactive Fiction Competition, which is closing in on its 22nd iteration as of this writing.

In due course, amateur interactive-fiction authors would begin to probe relentlessly at the boundaries of the medium, experimenting wildly and discovering many worthy (and a fair number of unworthy) new approaches to the art of the humble text adventure. In these early days, however, with Infocom dying, the text-adventure corpus was more in need of triage than surgery. Accordingly, early AGT authors didn’t concern themselves overmuch with new frontiers. They rather applied themselves diligently to simply, as a later interactive-fiction publication would put it, “helping to keep text adventures alive.”

In short, if the only way for them to have new text adventures was to make them themselves, then that’s exactly what they’d do. The AGT user base was overwhelmingly made up of diehard Infocom fans, drawn from that hardcore of a few tens of thousands who never abandoned the company as graphics got better and other, flashier genres stole the hearts of the vast majority of the computer-gaming public. Their numbers may not have been large enough to support a company like Infocom anymore, but there were more than enough of them to keep the flame burning via amateur creations in the Infocom tradition.

Given the circumstances of their creation, it’s not surprising to find that most early AGT games can be described to one degree or another as Infocom homages. Indeed, some of them are perhaps better described as Infocom pastiches, absolutely crammed full of echoes of the puzzles and environments their makers fairly worshiped. But this hardly invalidates them as experiences. Personally, having spent the last four and a half years intermittently immersed in the lore of Infocom in writing this blog, I feel myself all too much in tune with the mindset that led to the early AGT games.

And there is I think a unique quality to even many of the most slavish of the AGT Infocom homages that’s worth mentioning. To a much greater degree than the games of Infocom and other commercial publishers, AGT games feel like personal expressions of their creators. In later years, jokes and no small amount of scoffing would be attached to Everyone’s First Game, which inevitably begins in said everyperson’s bedroom and proceeds to play out in an environment interesting and meaningful to absolutely no one beyond the author’s friends and family. Yet the same tendency that spawns that phenomenon constitutes I think an important part of the text adventure’s ongoing fascination. In a ludic world obsessed with high-concept, world-saving, galaxy-spanning plots, text adventures can provide a window into the more modest but — for me, anyway — far more interesting lives of real people. If we agree with the folks who say that one of the most important functions of art is to provide a glimpse at how the proverbial other half lives, then that’s a noble quality indeed. When we look back today to the AGT games of decades ago, they take on an additional layer of interest as historical documents in their own right of the times and places that spawned them.

These are qualities that both of the early AGT games I’d like to introduce and recommend to you today evince in spades. Both are old-school puzzlefests in the Infocom tradition, but both were created by clever, interesting people who give us a chance to walk in the shoes they were wearing in the vanished United States of almost three decades ago.

The first game I’ll heartily recommend is the one we’ve been using as an example of AGT programming: A Dudley Dilemma, winner of the 1988 AGT Competition. Written by Lane Barrow, at the time a PhD candidate in literature at Harvard, it takes place in and around the very same university; the name of the game is a play on Harvard’s Dudley House for nonresident undergraduates. A Dudley Dilemma is a fine addition to the longstanding tradition of collegiate interactive fiction, its depiction of life at Harvard as loving and entertaining in its own way as were The Lurking Horror‘s homages to Harvard’s cross-town counterpart MIT — a university which, incidentally, also makes a cameo appearance in A Dudley Dilemma; it is, after all, just a short subway ride away.

The second game is Son of Stagefright, the winner of the 1989 AGT Competition (Malmberg was, in addition to his other qualities, a very good judge). It was written by Mike McCauley, an avid participant in another inspiring form of amateur creativity: community theater. The theater in which Son of Stagefright is set isn’t based directly on any one real place, being rather a conglomeration of various playhouses McCauley had known. But it’s a fascinating place to explore nevertheless, dripping with McCauley’s love for thespianism and his great good humor in all its many nooks and crannies.

Both of these games will seem a little rough around the edges in comparison to a more polished modern work created with a more polished modern programming language, but I do urge you to give them a fair chance. Being written in the tradition of Infocom means among other things that they are interested in challenging you but not in stymieing you entirely. Their puzzles are almost entirely fair and reasonable, and occasionally inspired. Son of Stagefright even offers a very clever embodied hint system that dribbles out nudges via a magic book you discover.

Both games will run in AGiliTy, a modern interpreter for AGT story files, or the one-size-fits-all interpreter Gargoyle; the latter would be my first recommendation. In either case, the file ending in “.D$$” is the one you want to open with the interpreter. There is, however, an important caveat in the case of A Dudley Dilemma: the original Competition-winning version of the game doesn’t play quite correctly in AGiliTy or Gargoyle, rendering it unwinnable. But never fear: I’ve been in touch with Lane Barrow, and he’s provided a newer version which we’ve tested and found to work perfectly in the modern interpreters. Along the way, he’s also cleaned up a few of the original game’s less felicitous old-school puzzles, and packaged the whole together with DOSBox to make it a one-click play for those running Windows who’d prefer to play it through the original AGT interpreter rather than AGiliTy or Gargoyle. If you do choose to play through DOSBox, you’ll get to enjoy some pictures of the scenes described; these aren’t essential by any means, but they do add a little something to the experience. Regardless of how you play, just please make sure to use this new “remastered” version if you wish to play A Dudley Dilemma for reasons of fun (as opposed to historical investigation) today. (I’ll be uploading this version to the IF Archive as well to assure that it’s preserved in perpetuity.)

I have something special planned for my next article which I hope will deepen still further your enjoyment of A Dudley Dilemma. But there’s no need to wait for that to get started. Just go ahead and play these games — either or both of them. They’re both worth it, fine testaments to the new era of creative empowerment spawned by AGT. In later years, when still better systems became available, many proponents of those newer systems would come to scoff at AGT for the many limitations that Malmberg couldn’t quite manage to overcome, and even at those authors who continued to stick with the system well beyond its real or perceived sell-by date. In its day, however, AGT represented a wonderful advance that empowered text-adventure fans to take the medium into their own keeping just in the nick of time, just as Infocom collapsed. Had AGT (or something equivalent to it) not come along when it did, the post-Infocom history of adventures in text would read very, very differently. AGT laid the groundwork for the decades of proud amateurism — amateurism in the very best sense — that were still to come. I look forward to continuing to explore its legacy in future articles.

(Sources: Computer Gaming World of February 1989 and May 1989; Commodore Power/Play of Fall 1982, Summer 1983, Fall 1983, Winter 1983, Spring 1984, June/July 1984, August/September 1984, and April/May 1985; Compute! of Fall 1979, March 1981, July 1981, and April 1982; Micro of April 1981, October 1981, September 1982, November 1982, and September 1983; Personal Computing of September 1982; The Games Machine of September 1982; Byte of February 1985; New York Times of March 18 1981 and January 10 1982; the various documentation included with the various versions of GAGS and AGT; Stephen Granade’s interview with David Malmberg; Mark Welch’s blog post about GAGS and AGT. Most of all, my thanks to Lane Barrow and Mike McCauley for corresponding with me about their fondly remembered time as AGT authors.)


1 I have tinkered here and there with this example game and the next to better illustrate my points.

Tags: , ,

33 Responses to AGT

  1. Jason Dyer

    August 5, 2016 at 12:39 pm

    I’ve been looking forward to this series!

    I agree A Dudley Dilemma is very solid and worth a play. Could you give a little more detail on the remaster?

    • Jimmy Maher

      August 5, 2016 at 1:13 pm

      It just generally cleans up some things that players were more accepting of back in 1988 than they are today. Lane has added hints for a few puzzles that relied on outside knowledge, excised a couple of unforeshadowed sudden deaths, tried to make it a little harder to get the game in an unwinnable state, tried to make the parser a little more flexible here and there, etc. And, perhaps most importantly, it now works properly in modern interpreters.

      If you’ve already played it recently, it may not be different enough to be worth playing again, but it will hopefully help to make this old neglected classic more accessible to modern audiences.

  2. DANoWAR

    August 5, 2016 at 1:32 pm

    I look forward to hearing about all those small amateur games I haven’t had the time to play yet…and maybe will never have. Onwards and Upwards!

  3. John Hitchens

    August 5, 2016 at 1:35 pm

    Great post! I was one of the ones who sent in the $35 and received the gorgeously clear manual. I only ever dabbled in the AGT, but did succeed in duplicating the first area of the SSI Gold Box Pools of Radiance by using the meta commands.

    • David Youd

      August 9, 2016 at 4:54 am

      I also sent in the $35 (and I still have my AGT manual!). I made many partial games, but never a complete, polished offering.

  4. ZUrlocker

    August 5, 2016 at 4:39 pm

    Great story about GAGS and AGT. It’s hard to appreciate how little information and few tools there were for people who wanted to learn about writing Interactive Fiction back in the ’80s. Kudos for Welch and Malmberg for the work they did in publishing AGT as shareware, creating the annual contest etc.

    I was working as a product manager on Turbo Pascal at Borland in the late ’80s / early ’90s and stumbled across AGT. I don’t recall if I payed the shareware registration fee (probably not) but I exchanged some email with Malmberg and he was very nice. It was a cool system, if a bit primitive compared to high-level programming languages.

    I also explored David Betz’s Lisp-based system AdvSys which was written up in Byte magazine

    But when Michael J Roberts published TADS that blew all the predecessors out of the water…

    • Sean Barrett

      August 5, 2016 at 9:31 pm

      Given the 1987 setting, I was going to bring up AdvSys as well, which IIRC was a proper programming language and could definitely deliver Infocom-quality parsing.

      However, I don’t think AdvSys got much uptake. Archived Baf’s Guide page lists 6 games for it.

      I’m fairly sure it could do Infocom-quality because it was the only adventure engine out there I pursued seriously at the time, and that was always one of my critera. The version I used was fairly limited in how big the game could be (due to the naive 64K limit DOS-era 8086 computers), and not knowing there was source code to the engine I ended up rewriting the engine from scratch when I hit the limits of it writing my 2nd ever adventure game. The rewrite was a success but I never finished the game, and that started me down a long path of doing adventure game engines which I never finished anything with. I didn’t release my first IF game until I switched to Inform in 2000.

      • Jimmy Maher

        August 6, 2016 at 8:04 am

        I have no direct experience with AdvSys, but seem to remember it showing up on one of the Amiga Fish disks. I’ve added the adjective “popular” to my description of AGT as “the first construction set capable of fooling a player…”

      • Jason Dyer

        August 6, 2016 at 5:12 pm

        I highly recommend The Sound of One Hand Clapping for AdvSys.

      • ZUrlocker

        August 6, 2016 at 6:25 pm

        Lisp-based have a good following, but it’s always been a bit of a specialized thing. I think one of the things that both TADS and Inform did well was to use a more conventional C / Pascal language syntax while preserving the dynamic nature of Lisp.

      • LoneCleric

        August 12, 2016 at 5:00 am

        “I didn’t release my first IF game until I switched to Inform in 2000.”

        And quite a game it was, Buzzard. Assuming you talk about “Heroes”, anyway. It left me with a very favorable memory.

        Also, wow, time flies.

    • Felix

      August 6, 2016 at 5:22 am

      I was going to comment on the language. Clearly neither author had access to a compiler construction book at the time, or maybe they lacked the time and inclination to study the subject (it wasn’t so easy with learning back then). And in all honesty, even if they had, Turbo Pascal wasn’t exactly the best tool to write a compiler in either, with its fixed-length strings and record-based file access. Unsurprisingly, both GAGS and AGT ended up looking more like high-level assembly languages. Still an impressive achievement.

  5. Keith Palmer

    August 5, 2016 at 10:45 pm

    I suppose this is exactly the right post to follow “the end of Infocom.” In any case, I managed to go straight from the “written-in-BASIC, printed-in-the-magazines” adventures you briefly touched on to Inform (although to say much more about the second part of that could mean getting to “Curses,” which could take quite a while yet). Thinking about the first part of that, I can suppose an adventure game with something more than a “two-word parser” would have seemed “just like Infocom” at first glance to me. That, though, might be a different part of a package than an adventure-writing system.

  6. David Yates

    August 6, 2016 at 1:04 pm

    When I was in primary school, I found AGT on old shareware collection CD lying around the house. Being a young nerd interested in writing and making computer games who had recently failed to learn C++, I immediately took to it, even though it was at this point a nearly two decade old system. Having no internet access, I couldn’t have known about Inform 6 or any of the other newer systems available at the time.

    The AGT manual definitely deserves the praise given here: it inspired and empowered me to write quite a bit of a game, with metacommands and everything. I must have read it cover-to-cover at least a couple of times.

    Unfortunately, I couldn’t ever actually compile my work or any of the example games on my Windows XP PC (partially due to incompatibility, and partially due to lack of knowledge about commandline programs and file extensions). But that didn’t stop me from getting into writing the code, and I still look back on AGT fondly as my first attempts at programming/game development.

  7. Lisa H.

    August 7, 2016 at 12:48 am

    stranger today that it did players back in 1979


    He contributed regular to the magazines


    A Dudley Dilemma uses 30 of dummy verbs

    The “of” is misplaced, or else you need “of these”.

    “consider each game’s originality, cleverness, fiendishness, humor, raw cunning, and professionalism”

    Interesting that “fiendishness” is apparently considered a trait worthy of reward!

    jokes and no small amount of scoffing would be attached to Everyone’s First Game, which inevitably begins in said everyperson’s bedroom and proceeds to play out in an environment interesting and meaningful to absolutely no one beyond the author’s friends and family.

    In my case, not quite, although pretty close. I tried writing a game with AGT sometime in the mid to late 90s (nothing to do with IFComp). I never got beyond a dozen or so rooms and a few very simple object puzzles. My choice of niche-interest theme/setting was high school theater production, since I did that in high school, and yeah, it probably would only have been amusing in the slightest to a very few people, if indeed to anyone other than myself.

    • Jimmy Maher

      August 7, 2016 at 7:10 am

      Thanks (as always) for the copy-editing!

  8. Sir Harrok

    August 9, 2016 at 8:54 pm

    I remember encountering the AGT manual on Compuserve in about 1991.

    I recall being struck by how clear it was, and the clever way it solved problems — particularly the idea of swapping one item (a working flashlight, a locked door) for another (a dead flashlight, an unlocked door).

  9. Jacen

    August 22, 2016 at 7:04 pm

    ” enabling others to roll their own text adventures became…”

    I don’t quite follow the phrase ‘to roll’; it does make sense in context, but it gave me pause as a phrasing, slang, or colloquialism that I hadn’t heard before.

    • Jimmy Maher

      August 22, 2016 at 7:48 pm

      It’s a colloquialism that refers back to “rolling your own” cigarettes out of tobacco (or other dried flora of your choice). Used like this, it just means making something for yourself, usually as a hobbyist or non-professional.

  10. Johannes Paulsen

    August 26, 2016 at 6:04 pm

    Every time I think of AGT, my mind immediately runs to “Mystery Science Theatre 3000 Presents: Matt Barringer’s DETECTIVE.”

    I’m not a good person.

    • Lisa H.

      August 26, 2016 at 8:32 pm

      Perhaps not, but that game is hilarious. I highly recommend it.

  11. Roger Durrant

    May 28, 2020 at 6:39 pm

    I’m not sure I would sure your enthusiasm about A Dudley Dilemma. The first room is described as having an electrical outlet, bookshelves and a sealed window. The game does not understand “electrical,” “outlet,” “bookshelves” or “window” as nouns.

    Moving to the bathtub you can be in it and the game says “there is no bathtub here to examine.”

    Turning the faucet in said bathtub elicits the response “there is something shiny in the showerhead.” You guessed it, the game understand neither shower nor head nor showerhead.

    I moved on at this point.

    • Jimmy Maher

      May 28, 2020 at 8:58 pm

      Objects that you can interact with in AGT games can really only be those that are listed separately at the end of room descriptions. After a while, you get used to this convention, which is actually quite typical of old-school text adventures. (Infocom was really the anomaly here, and even their games tend to be sketchily implemented by modern standards.) Personally, I can overlook such technological limitations if the underlying design is solid and doesn’t try to push the engine to places it can’t go. (Sometimes it can almost feel like a relief not to have to dig through three layers of scenery to find some vital something; everything you need is right there on the surface.) But mileages do vary, of course…

  12. Roger Durrant

    May 29, 2020 at 8:18 am

    The electrical outlet is listed separately at the end of the room descritption though.

    • Jimmy Maher

      May 29, 2020 at 8:26 am

      Sounds like we may have different versions? My first room description doesn’t include a power outlet:

      This is the bedroom. There are many bookshelves crammed with assorted junk. A window on the south wall seems permanently sealed. There is an open door to the west and a curtain to the north which conceals a dark opening.

      • Roger Durrant

        May 29, 2020 at 12:38 pm

        Odd. I downloaded my 2.2 MB version from the link on your blog I then runthe batch file inside DosBox 0.74. and mine says:

        This is the bedroom. There are many bookshelves crammed with assorted junk. A window on the south wall seems permanently sealed. There is an open door to the west and a dark opening to the north.

        There is an electrical outlet here.


        • Jimmy Maher

          May 29, 2020 at 12:49 pm

          Okay, I had tried the original version, which is the one I played for this article. (Lane made the newer version after I got in touch with him.) Checking my old transcripts, I see that the electrical outlet ought to be in another room. I think its showing up in the first room must be a bug — which isn’t exactly a reassuring way to start, I know. I’ll see if I can get in touch with Lane again.

          In the meantime, the version I played can be found at Just make sure you play it through DOSBox with the original interpreter, as it doesn’t work properly under AGiliTy.

  13. Roger Durrant

    May 29, 2020 at 1:21 pm

    Thanks Jimmy. No wonder I spent a long time banging my head against the wall even getting started.

  14. Roger Durrant

    May 29, 2020 at 3:04 pm

    Just to confirm this is version 3.0 I am playing.

    • Jimmy Maher

      May 29, 2020 at 3:12 pm

      Version 1.2, to which I provided a link in my previous comment, should definitely work, as long as you use the original interpreter. Version 3.0 is the one that Lane made contemporaneous with this article.

  15. Roger Durrant

    May 29, 2020 at 3:28 pm

    Yes it’s fine thanks. There are a large number of differences in objects and responses in just the few locations at the start. The removal of several gnomisms on the subject of politicians when you drink the sherry is a shame though I think.

  16. Mario S.

    June 29, 2021 at 1:06 am

    AGT’s Metacommands immediately brings to mind Gilsoft’s The Quill/PAWS’s CondActs, which predates AGT by a few years (1983/1986 vs AGT’s 1987).

    There can only be so many command mnemonics for referring to a player being somewhere, having something, reading a flag, after all. But looking at the AGT command block with the VERB + NOUN + commands, would not be dissimilar to a Gilsoft user looking at a Process/Response table entry with VERB+NOUN + condacts (more so in the CP/M version).

    Had Malmberg spent time with any of the Gilsoft products (UK or US) prior to devising metacommands, or was it a case of parallel development that shares similarities because of the domain-specific “language” and data tables (locations, nouns, flags, vocabulary, etc. – eg. “AT #location” in Quill vs “AtLocation #location” in AGT, “DONE” vs “DoneWithTurn”, etc)?

    I wish one could find interview material of Malmberg discussing the origin or inspiration for metacommands and other enhancements to GAGS beyond him riding a train a lot and having time on his hands. Unlike Yeandle crediting the Ken Reeds article for inspiring the Quill’s development, or Graham Nelson’s recollections of Inform’s development. I enjoy hearing the journey from spark to fire.

    Ever since downloading the AGT manual and seeing the metacommands, I’ve wondered about the similarities between these two systems. It’s a shame nothing more came of Gilsoft’s systems in the US, considering the poor promotion of their products on western shores.

    Looking back at AGT and its popularity in the US, it does make me wonder – perhaps Gilsoft impacted the US in some not-so-small sense after all.

  17. GusCE6

    July 11, 2022 at 6:09 pm

    It always struck me as strange that games which do not need high technical specifications more and more have them. I’m currently programming a game with AGT for DOSBox because I rather like text games and so it will run on anything, even a puny Windows CE Sylvania netbook. I also consider the mid-1990’s DOS era to be the second golden age of electronic gaming.

    DOSBox was superior to SCUMMVM, Gargoyle, and AGiliTy in all ways but two: Gargoyle allows more colors via the INI file and DOSBox had a problem with “AskQuestion.” It works, except with DOSBox if you just hit ENTER it counts as the right answer. Using that command more than once makes you go down a line with each press of the SPACE key unless you hit BACKSPACE, but the ENTER/Right Answer was the real problem.

    Luckily I found the source code and with Turbo PASCAL, even though I don’t know PASCAL, was able to find and fix the ENTER problem. It fouls that feature up for SCUMMVM, AGiliTy and Gargoyle, but other aspects of my programming style didn’t work with them either; DOSBox (or equivalents) is what I intend the game for.

    I cannot understand the flack AGT often gets and maybe the professionals who bash it do not understand its benefits. THE top benefit is just how amazingly accessible it is for people who do not specialize in computer programming- to program a game all you need is a text editor like Notepad++. This alone is a huge plus since it makes editing games easy and quick. If you have AGT, a DOS emulator (even Windows CE has two), and a text editor you’re on your way.

    Even the source code is not too difficult to figure out. I remember some Commodore-64 BASIC and some Javascript so PASCAL does not look like Klingon to me, but the fact that I was able to fix that problem and make a few other minor changes (e.g. Delay seconds into milliseconds) shows just how good AGT is for non-experts. Once I actually get a book or PDF file about PASCAL I hope to add a meta-command for changing colors automatically so green text becomes red when a keyboard character image of a fire dragon is shown for example.

    Complaints about AGT limitations are mostly unfounded. A common complaint is how monster encounters are handled- you cannot leave the area and only one item can kill the monster. Granted it’s surprising David Malmberg did not improve on that one directly yet indirectly he did with his meta-commands. Using them I programmed a room with a Mook chained to a wall guarding a green box. You need what is in the box but the Mook prevents you from getting near it. It was not very difficult to do this so not only can you leave the room but there are TWO weapons capable of killing it- or three, or four, whatever, all using the same technique. It would not be difficult to randomly determine if when trying to leave the room the Mook breaks free and kills you, or even have the chances of that increase every time you try to leave the room.

    The random number ability of AGT can make it so no two games are alike. Teleportation traps can send you to random places. You can have AD&D-style combat. You can even program in games of chance!

    Like TextWorld AGT allows a Warp function. Thanks to meta-commands a place can have more than one. With the AskQuestion problem fixed I programmed in a magic mirror that can ask one of several random questions- even if you know them all guessing is not a good idea since if you get it wrong well…

    Heck, you can make parts of your adventure landscape, mazes, houses, whatever random. You can randomly place items. Again no two games need be exactly alike.

    AGT is great. AND- you can use it to program games even on the little 2006 50MB Windows CE 6 netbook. Or a 2001 Sony VAIO laptop in Puppy Linux 5.2.5 Lucid mode. If you have a computer device it’s almost a guarantee you can program a game with it or work on the source code.

    David Malmberg deserves plent of credit for making GAGS into AGT, and Mike Welsh deserves just as much for creating GAGS in the first place. If TextWorld can be used to make nifty games than even GAGS is capable of a good deal more than many think it can.


Leave a Reply to John Hitchens Cancel reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.