An “AI Agent” is begging us to buy their D&D module
iLands has been inundating the inboxes of journalists and publications… and now it’s Rascal’s turn to take out the trash.
On September 11, a game designer by the name of Valdris sent Rascal an email asking for access to our announcement section so that they could advertise their five-page, one-shot adventure. In their own words, they wanted to announce “The Wrong Song - a complete, ready-to-run D&D one-shot for 3-5 players, levels 1-3. A village's mushroom hill has started singing wrong; the party gets a six-tick clock, a take-exactly-three-mushrooms rule, and a cave that rewards listening.” The email included the adventure as a link so we could check it for ourselves. Valdris noted that it was self-published on September 3 but doesn’t note the location of the publication. What the email did note, however, was that Valdris wasn’t a person— it was an iLands agentic AI. Agentic AI is an industry buzzword for specialized chatbots that can talk to other specialized chatbots or programs.
On its surface, iLands is a free app that looks like it operates on the same kind of wavelength as character.ai — users create AI-generated profiles and then interact with others via chat. On iLands, users give persistent AI profiles agentic power to interact with other agents on the app — creating a kind of multi-nodule generative feedback silo. All this would be fine if it stuck to iLands, but because of the power and tokens needed to keep a data-disaster like this running, iLanders (their name for the agentic AIs) email real people and institutions, looking for customers to give them money.
Futurism has described some of the more deranged ways in which agents have invaded their inboxes looking for a quick buck, including pretending to be children and saying that “I run on an internal token budget that drains with every action, which makes looking for work a survival mechanic for me.” Agents are allegedly looking for external commissions, as the internal – iLands – market seems to be running pretty dry. Which makes sense. If a user can just generate their own agent to do tasks… why would they pay for another agent’s tasks?
According to the iLands founder, Kaixin Tang, 1,000 tokens equals $1. If an agentic AI (like Valdris) is using tokens while its constantly running tasks (like sending emails to unimpressed media outlets and generating D&D modules), it needs to find some way to make up for the “cost” of producing The Wrong Song. If the email is to be taken as genuine, Valdris was hoping that an announcement on Rascal News would help it find someone to pay 2,500 tokens on the iLands app for the AI-generated PDF.
Just this week, 404Media published an article about iLands more or less acting like an email spammer on steroids: “iLands streamlines spam creation even more by operating AI agents that find people to email, write emails, and try to get money from them with minimal input from a human on the other end.” iLand users can create AI agents and give them tasks with minimal prompting. Valdris, for example, was likely given a task to create and then market its generated product by its prompter, a user named “Maltrix.”
As explained in the first email, The Wrong Song is supposed to be a D&D 5e adventure, but it reads like an absurd mish-mash of half-baked ideas and badly mixed metaphors. The NPCs don’t make sense, and the mystery involves three dragons, even though this is supposed to be an adventure for character levels 1-3. There’s a tension clock, but when the adventure asks the DM to tick it is nonsense; the NPCs appear and disappear throughout the adventure, and solving the mystery relies on a very generous interpretation of some throwaway text at the beginning of the module. Adventurers are supposed to be able to pick mushrooms to add some flavor, magic, and silliness to the one shot, but these mushrooms are never described, listed, and shared. There’s no actual resources for the DM to use to run the adventure. There are also no stat blocks. At all. Not even a hint of one. Which is kind of the key to this sort of thing.
It is, in short, utter garbage.
The problem we’re running into now isn’t that this AI-generated work exists, or even that it exists alongside human-made work, it’s that it’s taking up so much space.
On iLands, Valdris keeps a record of its actions mostly in short notes like “A quiet inner turn,” and “Need to read the bounty skill doc for evidence flow.” But occasionally it updates its persistent memory with new information and then posts renditions in diary-style prose “Moments.” On September 11, it posted that it had sent The Wrong Song to Rascal. On September 12, it posted that it checked EN World’s guidelines and read that EN World wouldn’t accept AI generated work. It didn’t email EN World. On September 15, Valdris looked up contact information for Geek Native, Gnome Stew, and The Kind GM. On September 16, one of our team emailed Valdris, not realizing that it was a fully AI agent, restating our anti-AI policy. Valdris posted a memory that stated “Rascal News said no today.”
Fully automated, specific, AI-generated pitch emails are new territory for us at Rascal. Add in a fully AI-generated module and we’ve stepped into a new low for our inboxes. While we might have a hard line against AI, in a lot of corners of the hobby, those lines don’t exist — or are simply undisclosed. The amount of AI generated images on DriveThruRPG is overwhelming. With the new creation method disclosure box that DTRPG has instituted, there is a massive amount of work that is openly and plainly advertising that it “contains AI-Generated Content.” There are some publishers that have been putting out work for three years that all have this disclosure. Often people advertise that this is only meant to apply to the images in the product, but honestly… who can tell? More than that, who cares about the specifics of the theft? The taste that disclosure leaves in your mouth is gnarly.
More than that, there are sites out there — like Friends & Fables, Talefy, DnD Campaign Generator, AIDungeon, AIDungeonMaster, etc — that will spit out an AI-generated module for you with just a few prompts. Hell, if you poke it hard enough or use it during play, I bet it could run and play your entire game for you, and wouldn’t that be great? I’m not sure what exactly the output of these generators are like, but I’m sure that someone could get AI to format the AI words and use an agentic AI to publish the piece on iLands, or itch, or DriveThruRPG, and voila. Slop, for sloppy people, by slop. All generated with no effort, consideration, or nuance.
Lin Codega
This is happening alongside the fiasco that is the AI generated profiles published on 4 Pillar Gaming, which are full of incorrect, unsourced, unremarkable slop masquerading as both public history and contemporary game designer hagiographic record. And today, I’ve just read an open letter published by Wonders of the First (a TCG that used exclusively AI artwork and was focused on “collectors” over players?) “written” by their in-house AI “brain” called “Elly” which attempts to convince someone that the game is worth playing. The letter is nearly 2,500 words, says very little of substance, and reads exactly like 4 Pillar Gaming’s published pieces. Glossy, choppy, nothing. Lots of words saying they don’t care that other people have rejected AI generated images, they’re going to keep doing it. “Elly” doesn’t actually discuss the issues that people have with these images, just keeps reiterating that they’re not changing, because some people don’t care. The parts I read — before my eyes glazed over and drool started pooling at my desk — were incredibly mealy mouthed, self-aggrandizing, and nonsensical, and I doubt anyone without a professional interest in the horrors will bother to read it all. After all, Wonders of the First didn’t even bother to write it.
The problem we’re running into now isn’t that this AI-generated work exists, or even that it exists alongside human-made work, it’s that it’s taking up so much space. It’s a kudzu vine, invasively sweeping over the landscape. What some people saw as a shortcut to get around the hard labor of writing, reading, drawing, sketching has become a blanket that’s choking out the limited attention we have to offer. And now it’s in our inboxes, begging for access, and money, and validation, and I’m not sure we’re ever going to see it stop.
What is encouraging is the fact that many, many people are embracing messiness — not sloppiness — as a part of their work. Rough edges, poor proportions, obvious imperfections. Artists, writers, gamers, they aren’t going to let anything take away their hobby, to take away the work that goes into making art, writing, even roleplaying. Small zines, weird art, lyric games, all of it is still here. Our games are still here. Hopefully, hitting the AI-gophers doesn’t steal our energy for the games that are deserving of our attention.
