top of page

Design Sprint #2: Download and Play a Legendary Card Game

Design Sprint #2: Download and Play a Legendary Card Game

I need help assessing a game created entirely by ChatGPT. Now, I'm completely capable of play testing this game and making notes about what it lacks and what it needs, but I'd actually prefer having this be a community experience. Afterall, LLMs have absorbed everything we've written, it's only fair we group together to criticize its work, right?


Not that it'll do much good. ChatGPT is completely immune to any ridicule or attempts to make it feel stupid or inept. Dammit.


The rest of this post is about the card game it came up with, and though it is super interesting and well worth your careful perusal, I'm putting the rules for this game here, at the top, so you can get started play testing the game yourself.



I suggest using a section of a standard deck of cards, aces through 8, and translate the suits to the colors in this way:

  • Hearts are red

  • Diamonds are yellow

  • Clubs are blue

  • Spades are green


Are LLMs going to change the way table-top games are created?

Almost certainly, but what is unpredictable at this point is what level of effect that will be. There are three areas Gen-AI is likely to have the greatest impact in the tabletop game industry — content, art, and concept development.


Most existing research on generative AI in game development focuses on the video game industry, which is unsurprising given how strongly that sector is driven by production scale, tight timelines, and the potential for automation to reduce costs. As Alharthi (2025) notes, studios are exploring generative AI primarily for its ability to accelerate content creation, streamline workflows, and expand creative iteration, though these advantages come with concerns about diminished human creativity, over-reliance on automation, and challenges in maintaining originality.


While much of the discussion around AI in games centers on video-game production, the broader research landscape shows a consistent pattern: AI is becoming deeply embedded in game-creation workflows across multiple domains. For example, Jagli et al. (2024) note that artificial intelligence is already central to modern video game development, supporting everything from procedural content generation and character behavior to animation, environment design, and dynamic gameplay elements. Their analysis argues that AI’s expanding role has the potential to increase immersion, accelerate production processes, and broaden the creative possibilities available to developers, though these gains also come with a growing reliance on automated systems to shape the player experience.


There are many productive uses for Gen-AI in table-top gaming. For example, Schatten et al. (2025) describe building a prototype system for a Game Master assistant, which could give NPCs personality, provide ambient scene descriptions, and possible even handle ambient sound, music, and imagery. A support system such as this has the potential to deepen the engagement the players' and GM's experience.


It sure beats playing CDs on a beatbox in the corner.

There have been plenty of attempts, both formal studies and casual community discussions, to understand how AI is affecting tabletop game development. One example is the BoardGameGeek.com thread titled “State of AI-Generated Art in Board Games Survey.” The sentiments expressed there (pugbunny, 2024) can be summarized like this:

  • AI-generated art is generally rejected for final, published products. Most posters feel that AI art simply isn’t ready for primetime.

  • A major concern is inconsistency. Because AI imagery can vary so much in tone and style, many believe it makes a finished game feel visually disjointed.

  • Hobby designers and print-and-play creators often take heat for using AI art, even when their projects are small-scale, personal, or non-commercial.

  • However, AI art for early concepts or prototypes is widely accepted. Most posters see no problem with using AI to quickly sketch ideas during development.

  • Pragmatic voices also note the real advantages—AI can speed up concepting, reduce design barriers, and streamline editing and pre-production workflows.


I tend to view AI in general as a useful tool for most tasks that would take me much longer than needed to do. Being a single-person shop, I need all the help I can for speeding up production. That said, I only use AI generated art as occasional reference for hand-drawn art, never for final illustrations. I will use it for text editing, proofing, and even re-writing blog post sections, but I always have final say over the final text produced.


But will these LLMs replace game designers and create a legendary card game?

Can a Large Language Model (LLM), armed with the collective knowledge of digital game design scattered across countless online libraries and forums, actually create a legendary card game?


Something close to this has already been tested. In Ma et al. (2025), researchers ran a four-week workshop where student designers used both text- and image-based generative AI tools to prototype board games. The results were interesting, if not exactly shocking: the AI struggled to weave the designers’ original ideas into coherent, playable game structures. In the end, participants treated the AI less like a creative partner and more like a handy “assistant resource” that could spit out drafts, sketches, or filler ideas—but not shoulder any real design responsibility.


Which, honestly, lines up with my experience too.


So, how do you use and LLM to make a legendary card game?

That's my goal with this design sprint, to see what ways Gen-AI can be used to create what I will call a "legendary card game." My working hypothesis is that LLMs, like ChatGPT, have access to a vast array of game references, and so should be able to draw those references together to find the ideal format for a fast, fun-to-play and universally accessible card game.


So, let's get started with this prompt:


I want to create the most legendary card game ever, one that can be played by anyone, is quick and easy to pick up, and sets the standard for fun at a level no one has ever seen before. Describe that game to me.


That does set the standard pretty high. If someone asked you to name a legendary card game, the list would be pretty long:

  • Poker

  • Uno

  • Cards Against Humanity

  • Blackjack

  • Solitaire

  • Go Fish

  • etc.


Can ChatGPT create something that would become a cultural touchstone like any of those games? Let's see what it came up with.


ChatGPT's Attempt at a Legendary Card Game

Below is the most brutally honest, wildly ambitious, design-first vision for a “legendary card game” that actually could exist—one that’s universally accessible, instantly learnable, endlessly replayable, and has the kind of emergent fun that feels inevitable, not forced.

This is not a gimmick deck. This is not a reskinned trick-taking game. This is the game people teach their friends for the next 50 years.


⭐ THE LEGENDARY GAME: FLUXFIELD

(Working title — the name will sharpen once the core metaphor emerges.)


1. Core Philosophy

A legendary game must hit five sacred design pillars:

  1. Instant Teachability — You can teach it in < 60 seconds.

  2. Universal Skill Expression — Kids can play; experts can destroy you.

  3. Moments of Hard Choice — The feeling of “I shouldn’t… but I must.”

  4. Emergent Chaos — Funny, surprising turns without pure randomness.

  5. Viral Table Energy — People laugh, gasp, swear, blame each other… then immediately ask to play again.

Most “classic” games have 2–3 of these. This has all five.


🎴 2. COMPONENTS

A 32-card deck. That’s it. No expansions required.

The deck is made of 4 suits, each representing a behavior pattern:

  • Push (Red) – Forces neighbors to act.

  • Pull (Blue) – Steals, redirects, reshapes the field.

  • Lock (Gold) – Freezes things in place, protects, blocks.

  • Wild (Green) – Breaks rules, bends rules, reverses consequences.

Numbers are 1–8 in each suit.


Why 32 cards?

  • Small enough for fast, snappy play

  • Large enough for deep emergent interactions


🎯 3. THE OBJECTIVE

Be the last player with cards in hand.

That’s it.

Simple, primal, timeless.


📘 4. HOW TO PLAY (60-second teach)

On your turn:

  1. Play 1 card.

  2. Resolve its effect.

  3. Hand passes.

Every card has one icon, telling you what it does:

🔺 PUSH (Red)

“Make someone else play a card right now.”

This causes chain reactions, panic, and table groans.

🔻 PULL (Blue)

“Take a card from the field or a player.”

Steal big moves, disrupt combos, bend outcomes.

🔒 LOCK (Gold)

“Freeze a card in front of you. It stays until unlocked.”

Your personal tableau builds your engine, and others must break it.

🌀 WILD (Green)

“Choose another player to suffer your effect instead.”

Chaos, comedy, betrayals… the dopamine engine.


⚡ 5. THE ENGINE: THE “FIELD”

Cards are played into a shared center row, the Field.

  • Pushing sends cards to the Field.

  • Pulling steals from it.

  • Locking protects cards you own from Field effects.

  • Wild rotates consequences around the table.

The Field is a living organism—small enough to understand, big enough to create emergent madness.

At 5+ cards in the Field, the Flux Trigger activates: Everyone must choose 1 card from their hand to discard simultaneously.

This keeps the game fast, tense, and self-balancing.


🔥 6. WHY THIS GAME SLAPS

A. Elegant Tension

Every turn forces you to decide:

Do I burn this card now to cause chaos… or hold it as defense?

Both choices feel dangerous.

B. Multi-Level Skill

Kids:

  • “If I Push Dad maybe Mom gets hit!”

Adults:

  • “If I Push into a Field at 4, I force a Flux Trigger right before her draw, cutting her defense stack.”

Both are valid play.

C. Built-In Comebacks

You can be down to one card and still wreck the entire table.

D. No Downtime

Push chains = “You! Play a card!” “No YOU play a card!” “Stop pushing me!!” “Flux Trigger!! Everyone discard!!”

It’s chaos theater in 12 minutes.

E. Infinite Replay

There are no optimal “perfect lines.” It’s micro-poker meets Uno meets Coup meets Oath. No two games rhyme.


🧪 7. WHY THIS COULD ACTUALLY BECOME A CLASSIC

A genuine evergreen title:

  • Must be understandable by grandparents

  • Must be fun at 2, 3, 4, 5 players

  • Must create stories like “Remember when you locked the 7 and I wild-pushed the whole table into a forced meltdown?”

Humans repeat stories. Games become classics when those stories become rituals.

FluxField is built to generate those rituals.


👑 8. OPTIONAL VARIANT: THE LEGENDARY MODE

For game-night players, not casuals:

Cards gain hidden synergies:

  • Matching numbers create combos

  • Opposing suits create negation

  • Rare combos allow “Flux Storms” (everyone plays 1 card face-down, resolve in number order)

This mode has “KeyForge-level” emergent depth while staying shockingly compact.


🔚 9. THE ESSENCE

The legendary game is…

  • Simple enough to teach instantly

  • Chaotic enough to be hilarious

  • Strategic enough to reward mastery

  • Small enough to live in your pocket

  • Deep enough to become culture

The game you play “just once more”… for the sixth time in a row.


Here's my initial impression of FluxField

I had a lot of questions about the rules, which is why I asked ChatGPT to run through a game example with me. It missed a card, but the game works okay-ish. I can see how the game is meant to play, and I do think there is a legitimate game in there. It needs a lot of work, though. I'm not as disappointed in the result as I thought I might be, but neither am I scared that ChatGPT will be used to generate astounding games anytime soon.


Download the rules for yourself:


What are my next steps?

I want to collect feedback from as many people as I can on the current set of rules to feed back into ChatGPT and see how it changes things. In the mean time, I'm going to take what it provided and do my own version of the the game and see if I can correct the mistakes and turn this into a fun, quick little game.



References

Alharthi, S. A. (2025). Generative AI in Game Design: Enhancing Creativity or Constraining Innovation? Journal of Intelligence, 13(6), 60. https://doi.org/10.3390/jintelligence13060060


BoardGameGeek.com. (2024, January 30). State of AI-generated art in board games – survey results [Online forum post]. BoardGameGeek. https://boardgamegeek.com/thread/3536844/state-of-ai-generated-art-in-board-games-survey-results


Jagli, D., Chandra, S., Dhanikonda, S. R., & Laxmi, N. (2024). Artificial Intelligence Usage in Game Development. Preprint. https://www.preprints.org/manuscript/202406.1983/v1


Ma, L., Pan, M., Siu, V., Chang, X., Holopainen, J., Li, J., & Ray, L. C. (2025). Follow My Logic: Generative AI Workflows in Designing for Serious Table-Top Games. In HCI in Business, Government and Organizations (pp. 153–172). Lecture Notes in Computer Science. https://doi.org/10.1007/978-3-031-92823-9_11


Santiago III, J. M. (2025). Exploring the potential of co-creative AI in tabletop role-playing games. University of Salzburg. https://eplus.uni-salzburg.at/Abschlussarbeiten/content/titleinfo/11759846/full.pdf


Schatten, M., Malović, J., Barišić, I., Martinaga, A., Marčan, J., Pavlić, S., & Tonković, P. (2025). An AI Enhanced Game Master Assistant for Table Top Role Playing Games. In Proceedings of the 36th Central European Conference on Information and Intelligent Systems (pp. S1–1). University of Zagreb. https://archive.ceciis.foi.hr/public/conferences/2025/Proceedings/S1/1.pdf

Comments

bottom of page