Mike Russo's ParserComp 2024 Reviews

We all know the drill, so let’s get to it!


Iyashikei – The Fountain, by Adam Sommerfield

In his writings on so-called “entheogens” – hallucinogenic drugs used for religious purposes – scholar Huston Smith proposed a three-part model for analyzing the experience of those using them: set, setting, and drug. “Set” is more or less shorthand for mindset, the expectations and beliefs a particular person brings with them, which obviously enough shape how things play out, while the specific characteristics of the precise hallucinogen on offer similarly has a clear impact on what the experience will be. “Setting” here signifies the ephemeral details of the particular context in which the drug is taken: is it night or morning? What nearby objects might attract the user’s attention? Who is the friend or friends there to keep an eye on things? For whatever reason, this last element always struck me as the most elusive – while the first and last factor are reducible to psychology and chemistry, the middle one partakes of alchemy: the same exact person could try the same exact drug, but have a radically different experience from one time to the next based on something as small as the color of the drapes.

I’m not necessarily saying that playing IF is like taking psychedelics, but the model comes to mind because I suspect my response to The Fountain would have been entirely different had its cover art been different. The blurb, which is surely a central part of the setting, nicely conveys what the game offers: a low-key fantastical environment through which the player can wander while soaking up the peaceful atmosphere. But the art conveys how that’s going to be done, presenting an aesthetic that’s Thomas Kinkade by way of Midjourney – for the former, see the garish, over-saturated colors, for the latter, see the dinghy that’s tied up to the underwater part of a piling or the chaotic pattern of ripples on the lake. Without that visual prompt, I suspect I would have enjoyed this well-meaning game a lot more; with it, though, I found myself getting undeservingly irritated by its sometimes-schmaltzy prose and thin implementation.

The writing issue is the biggest one because the game is more or less a walking simulator: over the ten minute or so run-time, by far the thing you spend the most time doing is looking at scenery. There are a few actions required of the player – you need to cross a lake on a boat, there’s some limited interactivity allowing you to bottle some water from the eponymous fountain, and at one point progression is blocked until you realize one location has an unmentioned exit, though I wasn’t sure whether this was a puzzle or an oversight. And beyond looking around, you can better appreciate the atmosphere via LISTEN, BREATHE, and MEDITATE. But there’s not much to the gameplay, and as far as I can tell the responses to these latter verbs are identical no matter where you go.

So looking at stuff is where the game is at, which is fine by me: I’ve played plenty of similarly-structured games, and it’s an approach well suited to the parser format. But this is a structure that lives or dies by the quality of the writing; absent deep lore or a characterized protagonist with a backstory to peel back, the only reward the game has to offer is descriptive prose, and sadly I found it just wasn’t up to snuff. Here’s X ME, for example:

You see yourself as a tranquil traveler, immersed in the serene beauty around you. Your presence here feels harmonious, a perfect blend with nature’s calm and gentle rhythm.

Here’s X SKY:

The sky stretches wide, a vast canvas of soft azure blue. Wisps of white clouds drift lazily, their edges kissed by the golden sun. Birds soar gracefully, their calls echoing in the serene expanse. The air is fresh and crisp, carrying the faint scent of pine and wildflowers. Sunlight bathes the world in a warm glow, casting a gentle radiance that touches everything below. As you gaze upward, the endless sky fills you with a sense of peace and boundless possibility, inviting you to lose yourself in its tranquil beauty.

And one final excerpt, from when you make landfall on the island:

You arrive at the island shore, it welcomes you with a blend of soft, golden sand and cool, green grass. Tall, shady trees line the edge, their leaves whispering in the gentle breeze. The water, clear and inviting, laps softly against the shore, creating a soothing rhythm. Sunlight filters through the branches, casting dappled patterns on the ground. Colorful wildflowers dot the landscape, their delicate fragrance mingling with the fresh scent of the lake. The shore invites quiet reflection, its beauty a tranquil retreat. Here, surrounded by nature’s serenity, you feel a deep sense of peace and connection to the world around you.

I can see what each of these excerpts is trying to do, but unfortunately I don’t think any of them work. Adjectivitis is the first problem, with the overuse of descriptive words undercutting the power of the prose and reducing the power of any individual image. It doesn’t help that the palette here is an extraordinarily limited one, too – “serene”, “tranquil”, “peace”, “harmony” are words that recur again and again, flattened by repetition, and even particular details, like sun dappling across a surface, are overused. The descriptions also commit the cardinal sin of commandeering the player to tell you exactly what you feel and think, which is risky enough with a characterized protagonist; with a main character who’s an empty vessel, this feels like a lack of respect for the player combined with a lack of confidence that the prose is accomplishing what it should. Taken together, these flaws make the writing aggressively kitschy, which doesn’t convey the restful vibe the game’s going for – and its wordy blandness kept me wondering whether the prose was also a product of an LLM tool.

Some implementation stumbles also took me out of the world. Beyond the unmarked exit, I ran into some trouble with the bottle (once I dropped it, trying to pick it back up triggered two messages saying I didn’t want to get it again), and in the second half of the game, I noticed a fair number of mentioned scenery items that weren’t actually implemented. It’s nothing too awful, but in a small game that’s aiming to create a meditative mood, the impact of snarls like these is magnified.

I’m aware I’m probably being too hard on an inoffensive game, and it’s important to acknowledge that this puzzleless, plotless structure is a high-wire act that makes small flaws more visible. And god knows we could all use more peace and a place of refuge these days. So if the cover art hadn’t pushed me to be on alert for the prose getting purple or robot-y, possibly I would have judged The Fountain to be anodyne enough – and I suppose there’s someone out there who might have had the opposite reaction (Thomas Kinkade sold a lot of paintings). Using the IF medium to present short, meditative experiences seems like a promising approach to me, so I’d definitely be up for more efforts in this vein in the future – I just hope I like the drapes better next time.


Alphabet City Parser Edition, by Jgesq

I’ve talked before about the culture shock that ParserComp can sometimes engender, sitting awkwardly as it does between the IF community’s norms of polished games made with an expectation of substantive feedback, and a more laissez-faire itch.io jam vibe where hacking something together in a couple of days is a praiseworthy act of solidarity and an opportunity to develop some new technical and design skills. There’s nothing intrinsically better or worse about either of these approaches, of course, but it can make the job of a reviewer somewhat awkward, especially since I’m very much of the write-a-couple-hundred-words-minimum school. Alphabet City’s blurb makes no bones about the fact that it represents its author’s very first steps with Inform, with some of its features included more to provide programming practice than for design reasons. In its favor, it boasts a gritty, underexplored setting (the early-80s NYC demimonde) and an endearing ambition, but it’s also got a long, long way to go before it could even be considered an alpha.

I assume the author is aware that there are a myriad of issues that would need to be addressed before the game could be considered ready for release – if indeed that’s the goal, and this isn’t meant to just be a learning exercise (nothing wrong with that!) So I won’t belabor the negatives; some are flagged in the attached transcript, but in brief, the combat and scoring systems both feel superfluous and arbitrary; scenery objects are often under-implemented, completely missing, and/or incorrectly marked as takeable (my inventory by the end included “the air in the Mudd Club”, “a motorcycle throttle”, “a pile of puke”, “a folding fixed in place three legged stool”, and “a Squeegee kid” as well as “a panhandler”); many conversations and other interactions are triggered bottom-lined when you simply examine a person or item; the game’s senses of place (a George Washington Bridge overpass is immediate adjacent to your Alphabet City apartment, which is in the Lower East Side; going south from the Fort Lee area somehow gets you to the Triboro Bridge) and time (despite being set in 1982, there’s a “thanks, Giuliani!” joke, and ) are often loosey-goosey; there are omnipresent disambiguation issues; and the final (and only) puzzle is of the read-the-author’s-mind variety (you have to LOVE JAYNE; more concrete attempts to HUG or KISS her, much less talking to her first to establish consent, go unacknowledged). And the fact that this is notionally a story about addiction, dependence, and relationship trauma makes the jank even more farcical, because Alphabet City in its current form is nowhere near up to the task of engaging with such weighty themes.

But! Judged as a couple days of work by someone brand new to Inform, what’s here is by no means bad. Lots of these issues are things that bedevil experienced authors, or would be smoothed out with a modicum of testing, and there’s even a certain charm to a few places where the game’s reach exceeds its grasp (there’s a subway ride that progresses by moving from one location to the next, rather than by waiting for time to pass, which is clearly just the fruit of not knowing how the rules for that work but made me smile regardless). And I’m all for more games with grounded milieus; okay, sometimes the grunginess here is a little much, but give me an incompletely-renamed Max’s Kansas City over a generic spaceship any day. The writing, even in its rough form, is also sometimes cleverer than it appears, like X ME telling the player that you’re “young, dumb, and full of romantic aspirations.” So as I’ve said before about similar games: as a competition entry, this is a disaster, as a thrown-together jam entry, it’s potentially promising, and while I can’t recommend playing it in its current form, I’m definitely curious to see where the author might be going next.

AC mr.txt (87.3 KB)


PARANOIA, by Charm Cochran

One of the things that I really look forward to in ParserComp is seeing games that try to come up with different gameplay mechanics for the hoary old parser interface, because even almost 50 years on from Adventure, there’s still plenty of fertile ground to be plowed. PARANOIA’s twist is so clever yet so well-suited to its format that it feels like someone must have tried it before, but as far as I know this is a real innovation: taking the meticulous investigation of a fractally-detailed environment and making it into the core gameplay, rather than just a means to the end of solving medium-dry-good puzzles, by challenging the player to notice small (and not-so-small) discrepancies – it’s an interactive version of those puzzles where you’re supposed to find three differences in a pair of seemingly-identical images, livened up with impeccably-timed comedy bits. If there’s not much plot to speak of and the instructions could use some sharpening, those are minor blips indeed compared to what it gets right.

Might as well start with the plot, so we can get that out of the way: it’s your basic Portal setup, as you’re participating in an experimental scenario whose contours are at first unclear. After you’re given a chance to poke and prod at your sparse surroundings – a vase of flowers, some wall art, a couple pieces of simple furniture – you’re instructed to push a particular button, and then the fun begins. The lights go out, the scientist’s flunkies scatter around changing some key detail about the room – or perhaps they don’t. And you need to use your five senses and your memory (there’s no undo or transcript feature available, and the scrollback window clears for each round) to suss out what, if anything, is now different.

Sometimes it’s very easy, obvious just from seeing the updated room description print out, but sometimes it takes close, careful investigation to identify the change, and the game does a great job of milking the disjunction between those two modes for comedy: a couple of times, I got a couple of rounds in a row where nothing changed, which of course occasions the most thorough poking and prodding, and self-doubt before you hit the all-clear button, only to be greeted with a ridiculously over-the-top shift that had me burst out laughing. I won’t spoil any of them, but there are some great gags here that go beyond just changing the physical layout of the room and mess with the player’s expectations in really entertaining ways. Being funny is rewarding on its own, of course, but these eruptions of hilarity also help with the pacing, since they usually provide an easy win – you need to get 14 guesses in a row right to achieve victory – or at least switch up the steps required to solve the round, and help motivate the player to press on to see what might happen next.

My only real quibble here is that it took me a little while to get into a groove with the game, which I think could be streamlined. In particular, I found the opening instructions ambiguous about whether I was meant to be comparing each round with the original state of the room, or to how the room looked in the round that came before. It’s the former, which for good or ill keeps the madness from escalating too far, but I wasn’t sure at first, and combined with the counterintuitive way the buttons are labeled – the green button means there is a change and the red one means there isn’t, which makes sense from a yes/no perspective, but my brain interpreted green as “everything’s fine” and red as “watch out, something’s changed.” After a couple of restarts it all became second nature, but slightly clearer framing might have helped me get to the good stuff quicker. Oh, and the winning menu asks if you want to UNDO, like always, but of course UNDO is disabled. But those are my only bits of feedback – this is a unique, engaging piece of IF unlike anything you’ve played before, and well worth the fifteen minutes or half hour it takes to win. So long as ParserComp keeps turning up these kinds of gems, long may it continue.


Mystery Isles, by Oz Realms

Quick, close your eyes. Now imagine the most prototypical adventure game puzzle you can think of – not any specific iconic one from the classics of yesteryear, nor the dreary ones you’ve done a million times like the get-the-key-out-of-the-keyhole bit, just the Platonic ideal of a classic text adventure puzzle. Once you’ve got it, you can open those eyes (how have you been reading this?)

This exercise doesn’t admit of wrong answers, of course, but I’d submit that there’s a single most-right one: there’s a monkey, and you need to give him a banana so he’ll give you his wrench. I don’t think I’ve encountered this specific scenario presented quite so baldly before, but when I ran across it in Mystery Isles, I recognized its totemic power. And in fact the whole game is like this, in its stripped-down, old-school-yet-friendly glory: you could call it Text Adventure: the Text Adventure and wouldn’t be far off. You’re marooned on a desert island, you see, and to escape you’ll need to construct a makeshift torch, unearth pirate treasure, climb a tree, and offer up multiple food items as bribes (it’s not just the monkey); it’s all presented in breezy, unadorned prose and will either take you forever – because there are a couple of puzzles that are real head-scratchers – or ten minutes, and fortunately there’s a hint function included so you can choose your own adventure on that front.

Much as I enjoy ParserComp as a space for experimentation, it’s also clearly a place to play the hits. Even given its limited ambitions, though, Mystery Isles could have stood for several additional rounds of polish, because the implementation is fairly rough. Beyond the aforementioned underclued puzzles – there’s a bit where hitting a big rock with a little rock turns the little one into a makeshift axe, which is not how flint-knapping works, and the business of how exactly you’re meant to get the banana out of its tree doesn’t give much for the player to go on, not even confirming the existence of actual bananas in said tree – there are plenty of niggles and small bugs. Items don’t always disappear from the inventory once used, once you solve a puzzle to obtain an object you might need to resolve it to pick them up again should you drop them, there’s a spurious north exit mentioned in the jungle description, and the hint function is welcome but gets a bit confused towards the end (it kept wanting me to relight the torch after I’d already obtained the map, which I believe at that point was both useless and impossible).

This is a short game, so even game-breaking bugs are quick to recover from, at least, but the lack of any credited testers really shows: there is no parser game so simple that it can be credibly released without independent beta testing, in my experience. There’s a lighthearted simplicity to Mystery Isles, and a certain ramshackleness can be part of the charm of such things – only as I’m writing this am I wondering about the plural in the title, since there’s just the one as far as I could see – but classic premises and design don’t need to be matched by creaky implementation.


The English Restaurant, by Eric Zinda

I am, generally speaking, an optimist. Some of that’s just the fruit of being born with a lot of privilege and a brain that knows what to do with serotonin, I suppose, but it’s also by choice: many years ago I came across a bit of Karl Popper arguing that nobody knows what the future will bring, or what will move it one direction rather than another, so we have an obligation to hope for a better world and act as though the little things each of us can do might bring that hope a bit closer to reality. That was persuasive to me and so I try to live into it, but I’ll confess that some days it’s harder than others, like for example the end of Supreme Court terms and when I play a Perplexity game.

I’ve been reviewing games using this engine since 2021 – this is the fourth, by my count – and while the pitch for a parser system that allows the player to use natural language input remains compelling, the reality is still so stubbornly far from the promise that reader, I begin to despair. Like you’re told your goal here is to order lunch at a diner for you and your vegetarian son, but when you say to the maître d’ “I would like to get some lunch,” the game butts in to say “I don’t know the words: lunch.” That’s small beans compared to this exchange with the waiter, though (the question marks are the prompts for player input):

?:my son is vegetarian

my son is not veggie

?:my son is veggie

I don’t know the words: veggie

?:my son does not eat meat

I don’t know the words: eat

Trying to couch your input as regular English sentences simply does not work – even as simple a phrase as ORDER TOMATO SOUP makes the parser throw up its hands in despair. What does work is single-word input: typing TABLE, MENU, SOUP to indicate what you want, which of course any existing parser engine under the sun can manage. It’s hard to hold this against the system, truly – natural language processing is quite difficult, from what I understand! But still, pushing the player to try to use complete sentences sets expectations the game can’t come close to living up to, while the blurb’s promise that it’s a good way for English learners to practice their language skills feels frankly irresponsible. Judged as a game, meanwhile, there’s basically nothing here – the only thing resembling interest is that you have a terrible waiter who needs too be reminded to hand you a menu and then prompted to tell you the specials – with no details to speak of and the world’s most basic prose.

In my previous reviews of Perplexity games, I’ve generally wrapped up saying some variation of “hopefully the system’s authors will keep fine-tuning things so it works the way it’s advertised to do,” but after three years, it’s hard for me to see any improvement on this front (at least the lag that I remember afflicting the earlier games appears to be a thing of the past). Perhaps it might be time to bring this experiment to an end? That’s maybe an unfair sentiment – and one certainly biased by the fact that the game doesn’t appear to end, so I spent a final ten minutes frustratedly typing BYE and LEAVE and I’M GOING and EXIT to the maître d’ who kept asking how he could help me today over and over like a robot – and I’d love to be proved wrong! But I’m not optimistic.

Oh, and the cover image is an AI-generated picture with myriad issues – beyond the standard-issue nightmare fingers, there are light fixtures hanging off of others, a double-handled coffee mug, an olive oil bottle standing in for wine, and a robot with only one eyebrow – and no attribution. Can we please stop doing this?


Thanks for the review. I had two testers but we didn’t find all the bugs, it seems.

1 Like

Race Against Time, by Finn Rosenløv

Dear reader, I must make a confession: I am not at all good at the hard-as-nails old-school throwback adventures that seem to be the predominant offerings of the current ADRIFT scene, and while I respect that different folks enjoy different strokes the vibe is generally too masochistic for me to enjoy. But then every once in a while, I’ll be playing a game like Race Against Time and think to LOOK UNDER a piece of furniture only to find that there is a hidden keycard down there and I’ll feel a little reward-jolt, and I kind of get it: a view of the pleasures that could be mine if I were king of the pixelbitchers.

I’m not, of course, and it didn’t take me too long to get out of my depth, but the game is agreeable enough about offering hints and keeping things zippy. The plot is standard text-adventure stuff: there’s a deadly plague loose on a space station, and only you are a bad enough dude to plumb its charnel halls and set off the self-destruct mechanism. There aren’t any twists or living characters on offer, so it’s strictly a medium-dry-goods affair, with a classic set of numeric keypads, powered-down elevators, locked chests, and broken mechanisms standing between you and victory. The map is relatively contained and straightforward to navigate, and the threatening atmosphere is established through efficient prose and a minimum of unnecessary detail, which helps keep things focused on the puzzles – because you’ll have to EXAMINE, SEARCH, and as mentioned, LOOK UNDER every implemented object to make sure you’ve got what you need to progress (the game politely informs you of this fact, at least).

Most of the game’s obstacles are pleasingly organic – by which I mean they seem like natural consequences of the situation, like a mechanism being bent by a scientist’s death throes, though there is a door blocked by a crush of bodies, now that I think about it. A few do feel excessively gamey, though, most notably the box that doesn’t open when you try to put the captain’s key in the keyhole, but does when you lay the keychain down on a groove that I thought was just the box’s opening. But this is part of the draw for people who like hard puzzles, I think – thinking “what would make sense in this world” will get you started on most of them, but the target audience probably thinks it’s an advantage to have a few challenges where you need to think creatively about your inventory without being too fussed about narrative plausibility. All told I got through about the first half of the game while using only a few hints, but had regular recourse to the walkthrough after that, which feels reasonable for the genre.

The implementation is likewise a mixed bag; I didn’t run into any bugs, but there are some nice conveniences, like a type-in-the-date puzzle that allows you to use either US or European date/month conventions. But while it’s nice that there’s a simple UNLOCK command, I only realized that would work after spending five minutes trying and failing to type stuff like SWIPE GREEN FOB KEY ON EASTERN FOB READER in a way the parser would understand, stuck in disambiguation hell.

I’m ending this review with a conventional copout that I really hate: if you like this sort of thing, this is the sort of thing that you’d like. Race Against Time is too challenging, requires too much unmotivated trial and error, and offers too little to players other than pure puzzle-solving gameplay to recommend to people outside its target audience. But it is sufficiently well put together to give a lay player a sense of the appeal of this style of game, I think, which is not always guaranteed.

Race Against Time MR.txt (177.0 KB)


I’d be curious on what you’d think about The Fire Tower because it’s pretty much that.


Thank you for including the transcript for Alphabet City: The Parser Edition. I’ve been playing it and got into an unwinnable situation in the night club because Jayne is not there and I can’t get out of the night club, possibly due to a bug in the map. The following is a bit spoilery, so don’t peak unless you’ve played the game.

By looking at your transcript, I was able to see that Jayne is meant to be in The Belly of the Beast, but she wasn’t. By tracing back, I was able to see that she first appeared in the bathroom of your studio, but I hadn’t been there. I’m presuming that I had to visit the bathroom before she is transported to The Belly of the Beast. So, I’ll have to start again.

1 Like

First, I appreciate the fact that you keep trying out the Perplexity games. Since the last competition, the engine has been completely rewritten and open sourced. That is part of the reason for the performance improvement. There are many new things that have been done in the engine that won’t be completely obvious unless you’ve tried building a natural language game, but they honestly don’t matter. The important part is the experience of the player.

To that end, there are two things I always struggle with in building these games.

First is just classic vocabulary holes. Given that it allows for natural language, there are more ways words can be used and more types of words that need to be implemented. You won’t believe how many ways people have answered no to “Anything else?”, for example. Certainly more beta testing helps find these and I’ve also tried ChatGPT. Vocabulary holes are still a struggle, and every one is painful to experience (and for me to see). I’m exploring ways to flush these out better.

Second is getting players, especially players who are used to classic Interactive Fiction, to “get into/stay in character”. For this game, that doesn’t mean trying the wildest constructs, like, “Well, kind sir, might I have bit o’ the slop you deign to serve here?” nor is it the classic IF commands like “ask waiter about food”. The sweet spot is: what would you actually say to a waiter? That is how the game is tuned, and that is where it allows you to practice your English. Take a look at the script from your session and for each command ask: “would I literally say this to a waiter in a restaurant?” When you say no, that is certainly my failure for not getting you to do it. But it also can help guide you on what works best in the game.

Even if you do that, yes, there will be terms not implemented, and yes, long phrases will get “that’s too complicated for me”. I’m working on it! But: If you want to see the best of this game you need to play it differently than classic IF. I take full responsibility for not yet finding the right way to get people into the right headspace. I’m still figuring that out. My hope with this particular game was that moving from a “moving around the world” type of game (that really is easier with two word commands) and implementing a dialog-based game would help along those lines. And it has! It has definitely helped users more naturally interact in a spoken english style.

And: I’ve learned that these two problems multiply because, when users hit gaps in the vocabulary, the illusion that you are “talking to a waiter” disappears and I’ve noticed that players then tend to fall back to non-English phrases which makes things worse.

So, back to some of your examples:

?:my son is vegetarian
my son is not veggie

Yes, the game should implement this (and the other variations of it that you tried), period. That is definitely a vocabulary hole. I didn’t implement the knowledge that you telling me about your son’s eating preferences was one way that people ask for the vegetarian options. You could say, “Do you have vegetarian options?” or “what vegetarian food is there?” or “Do you have anything vegetarian?”, all of which work and are something I think is also commonly said in this scenario.

This is a great example where the vocabulary hole broke the spell and then you tried a bunch of things that (I think) you would never say to a waiter. For example, “ORDER TOMATO SOUP” is not something most people are going to say to a waiter. The single word commands do work (as you found) because they need to for some cases (like answering “Q: what else would you like?”, “A: water, please!”) but I probably should have the waiter get annoyed if you use them in other cases (“Q: How can I help you?”, “A: table” would come across as pretty rude in real life). Again: If you use phrases that you would actually use with a waiter I think you’ll find it works much better.

In the meantime, I’ll keep adding vocabulary :slight_smile:

I’m sorry you didn’t like the AI generated art. I’m always very careful to give artists attribution for their work so they get proper credit and some free advertising. In the case of generative AI, I guess I didn’t feel it was necessary.

1 Like

I sympathize with the challenges you’re facing (and appreciate your even tone after my somewhat cranky review)! I think you’ve exactly diagnosed the issue, which is that it’s not too hard to get the player to start out typing full sentences/natural language, but once they hit something that fails and they find that single-word input does work, it’s very hard to get them to go back to the full-sentence approach. I’m afraid I don’t have any insights into how to avoid the dilemma, except more testing - honestly seeing that the game about ordering lunch didn’t know the word lunch significantly decreased my confidence that sticking with conversational input was going to work. I’m glad to take a look next time you’re looking for beta testers!


Yes! Very much an inspiration in fact. I’m tempted to try running an IF Art Show which is where The Fire Tower originated.


Where does this say he’s reducing ratings?

Sorry all, bad day at work and I’m afraid it’s put me in an especially cranky mood. I’m grateful for anyone taking the time to play and rate the game (like it or hate it).

I’m going to sign off for the day! Tomorrow is another day!


Thank you so much for the review! I’m honored!

This has been a major piece of feedback I’ve gotten from several people—it’s going on the list for release 2 expeditiously.


Apologies for not weighing in earlier – was off at a natural history museum with the family since my son’s day care is closed (key takeaway: ammonites were very cool but also horrifying). I know folks have deleted their posts, which I definitely respect and I don’t want to resurrect old disagreements – especially on AI/LLMs, which can obviously be a vexed topic around here – but I thought it might be helpful to just clearly state how I approach games that use that content.

First, I really hope we as a community can adopt a norm of acknowledging the use of AI/LLM tools in any piece of IF that relies on them. I think everybody understands that there are a host of potential concerns about the environmental and economic impact of these tools, as well as legal and ethical questions about their training processes – and different people come down in different places in assessing each of those concerns, and ultimately what their overall attitude towards the tools wind up being. I think it’s respectful of that diversity of views to disclose AI/LLM use to potential players. I think it also helps prevent people from approaching games with an air of suspicion, trying to guess whether a particular piece of art or passage of prose was made with an LLM, which I think would have a really negative impact on our community’s culture of thoughtful, engaged criticism.

I should probably say that I do find many of criticisms I mentioned above persuasive, but I will continue to play games that have AI/LLM-generated content if they’re entered into an event that I’m reviewing (pretty much all my IF-playing is in service to my reviewing!), and I don’t directly take that fact into account in my scoring. However, I do try to evaluate how effective games are at achieving their aesthetic aims, and if I think AI/LLM-generated stuff is impacting those negatively (which, again, in the spirit of full disclosure, I find it almost always does), given my approach to reviewing I feel like I have to discuss that, and inevitably would include that reaction in my overall score or assessment of a game. And since I do think transparency is an important norm, per the above, I do try to flag where I think there’s unacknowledged AI/LLM use in my reviews, both to try to strengthen that norm and also to provide that information to other potential players

Hopefully that all makes sense and is a reasonable way to approach what I know is a contentious issue! Moving forward, I might add a short version of this under a details tag in my initial review-thread posts so folks can refer to it if they’re curious.

Beyond that…

I’ve never actually played Fire Tower, but I have heard that that’s the vibe it goes for, which sounds really appealing to me! …it’s on the list so hopefully I’ll get to it one day.


Separate from the legal and ethical concerns—coming from academia, I appreciate people disclosing if they used AI, because it’s important to know where ideas originated. A human can cite their sources and an LLM fundamentally cannot.

This is, of course, a lot more important in research than in fiction writing.


Thanks Mike. I deleted my original post because it was just grumpy, having had a bad day at work, but some of my worries that underpinned what I said are still there.

The game, with hindsight, wasn’t ready, I acknowledge that, and by all means people can raise that in their reviews and score down as they see fit.

The issue I took yesterday was that from the outset it seems like the cover art vexed you and this cast a long shadow over your entire review. This immediate annoyance shows a bias, it’s evident you’re not a fan of AI, but to carry frustration of cover art into the game itself is a problem. It wasn’t just my game either I should also note.

It’s important to approach each piece with an open mind, free from preconceived biases. Otherwise, we risk stifling innovation and the rich diversity of perspectives that make the community vibrant (as I’ve noted several times before as having observed ruining other communities).

I won’t be signing up to anything that asks people to tag their creations as AI Assisted so that it can be “taken into account” (marked down as inferior in other words). That’s gate-keeping and elitist, and I don’t want to be a part of that. I don’t think anything good has ever come of tagging things so others can adjust their view of them.

EDIT: I don’t want this to become an argument so two final points- I’m not creating a bloody “strawman” here, these are legitimate concerns, and lastly I’m not going to reply to any questions or try to debate this because as mentioned previously it’s unwinnable by any side. Thank you.

I mean, isn’t the point of cover art that it gives an initial impression of what your work is going to be like? People (generally) don’t judge interactive books by their covers here, but the cover does set your initial impression. For example, from earlier in this thread:

I don’t think I’ve seen anyone taking points off because of AI-generated cover art, but it’s important to think about the impression you’re giving. Not all of us can or will commission professional artists for every game we make, but an AI-generated cover, a stock photo, some word art, or a pencil sketch all set very different tones for starting out.

I’m currently trying to figure out how to convey “lighthearted Agatha Christie pastiche” for my IFComp entry, and I’m currently thinking of commissioning the same friend (Heronmark) who did the Labyrinthine Library cover:

The lineless, cel-shaded style seems perfect. As opposed to Stormrider, where I took some public-domain photos and messed around with the hue-saturation sliders until it looked appropriately Dramatic™:

Okay, tangent over. I don’t mean to argue, just to point out that it’s not anything so simplistic as “AI bad”, and I don’t think Mike is the kind of judge to take off points for that sort of thing. (In fact I’d be shocked if he had any strict criterion that would always lose points.) But the cover art, the blurb, the introduction, the gameplay, are all parts of a larger whole, and it’s important that they fit together.