Three Witches

View Comments

• The machine-generated cover illustration should have been a red flag. If an adventure has actual art it means the publisher spent money on it or the adventure was itself made by an artist. If an adventure has no art you know at the very least that it's not a cynical cash-grab.

• The blatantly AI-generated cover image should have been the indicator of the AI generated text. My dude asked for suggestions to chatGPT and copy pasted those as an “adventure”.

  • Also, the fact that the product description says it "Contains AI-Generated Content" might have been a clue that it contained AI generated content.

    Also, the author has put out 37 titles since May of this year. I looked at 8 of them; the total pagecount for those 8 modules was 1542 pages. One of them was 459 pages by itself.

    So either they are all entirely AI generated, or Malrex really needs to step up his game and get modules out quicker.

  • I don't know - art is distinctly separate from text. Anyone can write text; few people can make passable art. My baseline assumption when I see AI art is "that guy didn't want to pay an artist, but also didn't want his adventure to look like a Word doc", which I can forgive for hobbyist/free work, but not for something that cost money.

    The better indicator of AI text is churn, as Beoric states. If this is indeed the author's 37th title in 2 months, that's a HUGE red flag. Even a cocaine-fueled Stephen King would struggle to reach that output.

    • Just recently under 'The Snake Temple Ruins' I said that that people are too quick to accuse adventures of being LLM text and tend to overestimate machine writing. But looking at the preview for this one it is absolutely written by a machine with minimal editing.

      • I used AI when the craze started. Privately, for fun. I kept reading articles about how it can write adventures on its own with prompts, and I thought I'd test it out of curiosity. That's exactly the way it replies. With maybe, perhaps, could be. This designer copy pasted the answers and sells them for 4 dollars.

        I am not against ai. It's a good tool and it's here to stay. There are ways (dnd or not) that it can be helpful. But copy pasting like that is the worse way to do it.

    • This content is 100% AI slop, and any editing sone to it is minimal. Rutibex is a blight churning out endless amounts of AI slop onto Drivethrurpg, and does not deny it when accused of doing so.

      Just take a look at their products. This time the accusations are real and apt.

      • Agreed. What a slimeball. I can forgive some use of AI art for free illustrations (and I’m an artist for chrissales…!). But when you are using AI *text* to churn out so much junk, how can you even pretend to yourself you are making something. Plus youre polluting the world literally and figuratively. Just pathetic.

• Short of altered statblocks, rebalanced treasure amounts, and the absence/presence of terms like "CR" and "DC", what exactly differentiates a 5e adventure from an OSR adventure? And please don't say "open world agency" and "problems solved off the character sheet", because that stuff is purely authorial choice, and not at all system inherent.

  Someone really should publish some kind of conversion manual, considering the number of times it gets brought up.

  • OSR is a style, not just a set of systems, and plenty of people have already published style guides on it. But even when doing a 1:1 conversion from one system to another, you have to understand the game you're converting to and play-balance it so all the gameplay loops work as intended and nothing breaks, either as the writer or the DM. If you give someone a conversion manual but they're never played the game they're converting to, they'll still mess it up.

    • Yeah I get what you're saying, but you're not quite picking up what I'm putting down.

      When Bryce says "this says it's an OSR adventure but it's clearly a 5e conversion", what does that mean exactly, beyond adjusting some numbers? That is to say, what are these so-called "gameplay loops" that are apparently only capable of existing in OSR and not in 5e? Because I can't name a single thing you can *only* do in an OSR game that you *can't do* in 5e, and vice-versa.

      The only issue I can identify between the two is the imbalance of direct numerical conversion (monster hit points and issues with gold-for-XP, mostly). And if it is the case that numbers are the only difference between the two, than simple conversion metrics would fix that, no? Literally a conversion manual (a 5e creature with AC 15 and 100hp = a B/X creature with AC 11 and 50hp", or whatever). "Half all the hit points", "Multiply GP values by 1o", and so forth.

      Assuming numbers are converted, what other factor could possibly distinguish a 5e adventure from an OSR one, that isn't just an author's personal bad practice choice to railroad or what have you?

      • Older edition modules have a different aesthetic than newer ones, and make different assumptions as to how the DM and players are going to relate to the game and each other.

        IME, newer edition modules are far more structured in ways that limit the choices, not just of the players, but also of the DM. A DM *can* colour outside the lines, or allow players to do so, but newer modules make it much harder for the DM to do so.

        An example of this would be the way travel, or other transitions between encounters, are handled. An area map allows the players to choose routes or modes of travel that were not contemplated by the module designer; it is an aid to improvisation. Whereas if movement between encounters is directed or handwaived, and the players want to go another direction, the DM has to invent the new route from nothing.

        There are also differences in action resolution. An early edition module is likely to assume that many obstacles will be resolved narratively, with a discussion between DM and player respecting the approach a PC will take, with the DM determining the odds of success after the player has described an approach. This is supported where a module either describes an object that can be interacted with in a fashion that encourages players to make choices about how to interact with it; or the module give no significant description of the many objects stated or assumed to be present, so the DM and player end up figuring out together what things look like and how they may be approached. Later edition modules instead present a list of objects with no particular description, but a list of mechanics with DCs attached to them. By default, this encourages interactions that are related to relevant skill and related DC.

        Another issue is presumptions of difficulty. If you are converting from an edition that has save-or-die mechanics to an edition that does not, you either have to invent a mechanic or lose some flavour. Another example would be converting from a system that gives a thief a 20% chance of finding a trap, to an edition that gives a 50% or more chance of finding a (probably less lethal) trap.

        • What you describe are things that, while common to OSR modules, are nowhere near exclusive to them.

          Choices for example; there is nothing mechanical in modern editions that limits choice. Setting a DC for an action's difficulty does not exclude the possibility of a different solution altogether. "Structure" is a writing issue, not a system issue. Any Dragonlance module is proof enough of that. Ditto for travel - the rules are so obscure that there are just as many "how to do overland travel" videos for 5e as there are for any old edition, and they all offer the same advice. The way travel is done at the table is 100% based on how the DM does travel (I mean, obviously). Modules, be they modern or OSR, simply try to capture whatever way best communicates the writer's vision to the DM (which is also subjective based on the author).

          Ditto for action resolution - a lot of people fail to take into account that a DC (seen as the primary resolution mechanic for 5e) is an entirely optional thing, and that a decent DM is one who is willing to allow any resolution that satisfies the situation. I do it in my game all the time - "tell me how you plan to accomplish that"; "I do X, Y, and Z"; sounds good, it happens". Sure, DCs are used if there's an element of chance, but when there's no chance required, DCs are likewise not required. And a percentage is just a DC roll in another form - a target number to be rolled on dice, with modifiers applied depending on the character and/or situation.

          Presumption of difficulty is just as easily at the author's whim - there are literally no rules in modern D&D that say "you cannot use save-or-die situations". Yes, there are death saves, but again, entirely at the whim of the DM, and a module author is perfectly within their rights to say "they die, no death save" in their works. But they don't because players don't enjoy those kinds of games.

          I concede that modern authors adhere to certain trends of design (as you say), but there no codified guidance to that effect; they're all just copying one-another, which goes back to my main premise that all the difference between editions are more due to authorial decisions rather than actual mechanical differences.

        • It occurred to me after the fact that Bryce could very well be referring to the author tendencies rife in modern modules (basically what you've identified Beoric) being present in older "conversions-which-aren't-really-conversions" as being a sign that something was lazily converted, and I suppose in that sense I can see where he comes from. It just irks me that what he gripes on are writing problems inherent to shitty authors, and not actual issues in the conversion of one edition to another. But then that's just more proof that some kind of guidance documentation is required, because nobody seems to be getting it right.

      • @DP: I can't think of anything you can *only* do in an OSR game that you *can’t do* in, say, a GURPS game either, but different systems create different modes of play that call for different approaches, even on a more granular level than just the overall style of play that ruleset encourages (and which is often recognisable in a conversion).

        Lots of little things add up, like: is the 'long rest' system and reduced focus on resource management in 5E relevant when running a hex crawl? How about the PCs' unlimited access to cantrips, including good odds that at least one PC has a 30' range telekinesis power they can use any time they want from level 1, is that going to affect dungeon crawling? These things, because there are a lot of them, tend to express themselves a lot in adventure design and assumptions.

        I'm presenting it from the opposite perspective because it might make it clearer: If you run a particularly lethal, trap-heavy, secret door riddled dungeon in OD&D, the bulk of playtime will consist of the party side-eyeing the architecture, bickering with each other, asking careful questions and coming up with elaborate precautions before making any moves. None of that is mechanical, but it will be the meat of the game. If you convert that dungeon to 5E changing only the numbers, the players will roll insight and perception checks before moving on, possibly getting upset when a low-level spike trap insta-kills a PC. It would not be fun, it would not be suited to the rules or expectations of that game. The same is true in the other direction, when you convert adventure to OSR without taking into account what makes it suited to a particular system.

        • > I’m presenting it from the opposite perspective because it might make it clearer: If you run a particularly lethal, trap-heavy, secret door riddled dungeon in OD&D, the bulk of playtime will consist of the party side-eyeing the architecture, bickering with each other, asking careful questions and coming up with elaborate precautions before making any moves. None of that is mechanical, but it will be the meat of the game. If you convert that dungeon to 5E changing only the numbers, the players will roll insight and perception checks before moving on, possibly getting upset when a low-level spike trap insta-kills a PC. It would not be fun, it would not be suited to the rules or expectations of that game. The same is true in the other direction, when you convert adventure to OSR without taking into account what makes it suited to a particular system.

          I think that may be more subjective to the players/DM than you give it credit, but I will admit that it certainly is *likely* to happen, sure. I still maintain that a DM can run old-school style with new school rules, and therefore an author can just as easily write a new school module in the old school style, but it seems to be either a skill or desire that is lacking.

          > Lots of little things add up, like: is the ‘long rest’ system and reduced focus on resource management in 5E relevant when running a hex crawl? How about the PCs’ unlimited access to cantrips, including good odds that at least one PC has a 30? range telekinesis power they can use any time they want from level 1, is that going to affect dungeon crawling? These things, because there are a lot of them, tend to express themselves a lot in adventure design and assumptions.

          This part I agree with, and was exactly the sort of difference I was looking for. The long rest system and cantrips do indeed change long-form session play, you're right.

        • Speaking from experience, you can run an OSR-feeling campaign in 5E with some house ruling but you do have to either wrestle with the system or amputate it because it's not designed for that, and because it *is* designed to be internally congruent for what it is. It's simply not true that it all comes down to the writer's preferences.

          To use your example above, death saves are at the whim of the DM in that you can *change the rules*, but the rule is that a PC reduced to 0 HP falls unconscious and makes death saves. It has a real effect on play whether an orc will chop your head off or put you in time-out. The mechanics of the game are more than incidental to play, and the issues that come up are not whether you're "allowed" to do anything in particular as a DM (because you can of course do anything, including and play a different game) but the structure and assumptions of the game rules you're using and how they inform play.

• I suggest a new category “ AI Slop” and a blacklist of AI authors.

  If only for your own good - this is the third thing youve reviewed by this shitepeddler.

• Should a no-AI policy be instigated? I'm not against the judicious use of AI for spell checking, minor art generation, etc, but wholesale adventure writing that results in slop will surely result in Bryce expiring of apoplectic rage and dismay.

• Please don’t review AI generated content.

• > “A fairy ring. Entering it could transport the players to a fey realm or cause other
  > magical effects.” No? How about: “A cursed statue in a small pond, with a puzzle
  > that, if solved, breaks the curse and grants a reward.” These are typical.

  They are also obviously AI-generated. So is all that other tripe you quote. It's blatantly AI-generated chum.

  Bryce, you really need to screw around with AI. Just go to "chatgpt dot com" and play around for free. It's fun.

  If you do, you can pretend you have one friend in this world!

• This does illustrate pretty well why special technologies for "AI detection" aren't actually necessary: because everything made with them sucks ass, simply filtering for quality will also exclude it, too.