Transmedia Jam #1 : Humans versus AI

Hosted by lapinchien · #TransmediaJam1

Joined

Overview

Community11

Transmedia Jam #1 : Humans versus AI community

Say Ahh! – a cozy doctor game with local AI patients Sticky

A topic by Prawns Agent created 9 days ago Views: 299 Replies: 28

Viewing posts 1 to 6

Prawns Agent9 days ago (1 edit)

Hi everyone! Thanks to lapinchien for the invite.

I'm a solo hobby developer. Say Ahh! is vibe-coded: I made it in Godot together with an AI assistant (Claude Code). The code and the low-poly 3D models were created with AI, and the patient dialogue is generated live by a local language model. I came up with the idea, chose the language models, specified the architecture, and directed and tested every step. It took about 3.5 weeks.

ABOUT THE GAME

You run a cozy low-poly doctor's office. Every patient is played live by a small language model (Qwen3 1.7B) that runs entirely on your PC, with no cloud and no account. The illnesses are hand-designed, but what the patients say is generated, so every conversation turns out differently. Ask questions, examine throat, lungs and eyes, make your diagnosis and fill your disease book with all 16 illnesses. It supports English, German, French and Spanish.

WHY IT FITS "HUMANS VERSUS AI"

The human doctor has to work out the truth from an AI patient who can be vague, inconsistent or just wrong. And the game itself is a small example of the same question: who really made what?

WHAT I'M LOOKING FOR

Playtesters, especially native French and Spanish speakers who can tell me whether the patient texts sound natural (I'm a native German speaker, so German is covered), and ideas. New illnesses, patient personalities or ways to use a local LLM in a game are all welcome. If you'd like to collaborate or just chat, leave a comment on the game page or message me through my itch.io profile.

GOOD TO KNOW

It's a 1.4 GB Windows download because the AI model is included. On Windows 11, Smart App Control can block unsigned hobby games. The page explains how to deal with it. Sending technical data is optional and off by default.

Play it here: https://prawns-agent.itch.io/say-ahh

Thanks for reading, and questions are welcome!

lapinchienHost9 days ago

Thanks a lot @Prawns Agent! Your game has an excellent and highly original high-concept, with incredible potential for both entertainment and serious/educational gaming. I will test it tomorrow, as I don't have my laptop with me right now. As a native French speaker, I'd be more than happy to test it and give you detailed feedback!

On a personal note, I spent years dealing with an undiagnosed illness, going through countless doctor visits, medical tests, and analyses trying to figure out the root of my symptoms. I know firsthand how tough it is for doctors to uncover the truth—and not just, as Dr. House says, because "everybody lies." Because of that, your core gameplay really resonates with me.

I am also particularly impressed by how you overcame technical hurdles using vibe-coding; I honestly assumed you were a professional game dev! I can't wait to see how embedded AI for NPCs can revolutionize game design and enrich the overall storytelling experience for players.

I hope you'll allow me to promote this jam by sharing your itch.io page on my social media networks and newsletters!

Prawns Agent9 days ago

Thank you so much for your kind words!

I'm really glad the concept resonates with you, especially given your personal experience. Feedback from someone with that perspective will be incredibly valuable.

And honestly, hearing that you thought I was a professional game developer made my day 😄

Of course, you're welcome to share the game and promote the jam on your social media and newsletters. I'd be honored.

I'm looking forward to hearing your thoughts once you've had the chance to play it. Thanks again for all the support!

lapinchienHost8 days ago

Awesome! I’m really glad to hear that. I’ll test the game properly on my laptop tomorrow and share my detailed feedback with you right after.

I’ll also start sharing and spotlighting your project in our jam communications/newsletter—it's such a great proof of concept for local AI in game development and great perspectives in other storytelling disciplines!

Talk to you tomorrow!

Prawns Agent8 days ago

Thanks a lot for the encouragement!

I'm really glad you like the concept, and I'd love to hear your feedback once you've had the chance to play.

Feel free to share the game and promote the jam — I'd be honored.

I already have plenty of ideas for future updates and improvements.

Thanks again for all the support! 😊

olistupid8 days ago

Some quick comments:

It took me more than 15 minutes to download the game. I did have a bit of a slow connection on my side, but please try to keep file sizes as small as possible. I understand that the AI model takes up space: maybe you could consider having only English as the default language in the game and allow the user to download other languages in case it wants to play in another one, or automatically download the language set up as default by the user’s PC.
“Play” button gets stuck at loading when executing the game unless you switch to English and switch to your preferred language again.
When executing the game the buttons for selecting other languages are disabled except the current set language and English.
As a native Spanish speaker, the language is fine. If you are working with an AI I dont think you’ll have an issue with that. However the AI seemed to crop some words out of it sentences to suffice its word limit per message, sometimes making sentences that don’t make sense grammatically. It also uses some quite unnatural language; some of the things it says wouldn’t be said by a casual speaker at all.
Personally, I see the AI function of the game as purely secondary and even cosmetic for the core gameplay. Although I can imagine the goal of this is to have a natural conversation with the fictional patient, all it takes is you to ask “what is wrong with you?”, for you to have all the necessary information for the puzzle, along with observing the character animations and examination actions. The game really doesn’t have difficulty and hence all challenge is lost, along with lost of interest and encouragement to play.
The concept worked with the common diseases of Level 1, but the ones from Level 2 onwards are hard to identify… because I don’t know what these diseases are. The fact that they have weird names makes it worse. The game always assumes you know the diseases you are trying to diagnose. This is fine for the common ones, but not for the rarer ones. Makes me lose motivation imo… No one is actually going to research or search up diseases for this game.

This is why I’m critical with beginners making their games with Claude… It’s a great tool don’t get me wrong, but relying on it completely for making a full game may give you a false sensation of “success”: it does gives a fully presentable product. However, it doesn’t matter if the game’s presentable or has as many impressive features as possible, at the end of it what determines if it’s fun is the game’s foundation.

Learning to do a product by yourself instead of fully relying on AI for making a game is also part of game development. In this project, you aren’t practicing game development exactly; you are practising game design.

This is my full opinion. Let me know what you think 🙂. Good luck on developing you game!

lapinchienHost8 days ago

Wow, what an incredible, detailed, and insightful feedback @olistupid! This is exactly the kind of deep dive that makes a jam so valuable.

@Prawns Agent, olistupid points out some really crucial Game Design challenges here—especially regarding how to make the local SLM feel like a core gameplay loop rather than just a cosmetic feature, and how to balance the learning curve for rare diseases.

I totally agree with the distinction between tool-assisted output and core game mechanics. Using local AI (like Qwen) is a huge technical feat, but turning those raw generative responses into a challenging, engaging puzzle is where the real game design magic happens!

I'm testing the game on my laptop today as well, but I'd love to hear your thoughts, Prawns Agent, on how you plan to iterate on these prompt limits and UX mechanics!

That being said, I still firmly believe that the high-concept of this project is super original and has an incredible potential in entertainment, education, and serious gaming. I also absolutely love the video on your channel that presents the concept so brilliantly. However, following the feedback from olistupid, I think this shouldn't be viewed as a finished game meant to be distributed or sold directly to hardcore gamers right now. Instead, it should be considered a perfectible prototype—a powerful stepping stone to pitch to publishers and funding sources (like business angels or crowdfunding campaigns) to give such a project the true scale it deserves. In my opinion, setting up a proper company structure with capital and hiring dedicated developers to flesh it out would be an amazing objective for the future!

Prawns Agent8 days ago(+1)

Thanks a lot for taking the time to play the game and write such detailed feedback! 🙂

I actually agree with many of your points.

The download size is definitely something I want to improve. The current AI and language system take up most of the space, but I'm already thinking about making languages optional downloads or only downloading the language that matches the player's system.

Thanks for pointing out the language-switching bugs as well. I'll investigate those and try to fix them in the next update.

Regarding the AI dialogue, I agree that it still needs work. My goal is for the patient to feel more natural and believable, so I'll continue improving the prompts and conversation system.

Your comments about the gameplay are especially interesting. Right now, it's true that players can often get the diagnosis very quickly by asking direct questions. In future versions, I want patients to have more individual personalities and behavior, and not always provide all information immediately. I also want examinations and observations to play a bigger role in reaching a diagnosis.

The point about rare diseases is fair too. As the diseases become less familiar, I probably need to give players better ways to learn about them inside the game instead of assuming prior medical knowledge.

And yes, I used AI extensively during development. For me, this project has been both a game design experiment and a learning experience. I still have a lot to learn, but creating this project taught me a huge amount about designing systems, balancing mechanics, integrating AI, and managing a larger project from start to finish.

lapinchienHost8 days ago

Thanks for sharing your vision on this, @Prawns Agent! Your open-mindedness and willingness to learn from feedback are truly commendable.
I want to share a broader perspective as an organizer: please don't underrate what you've accomplished here. While it's great that you're using this as a learning experience to polish code and UX, what you have right now is already much more than a simple hobby project. It is a fully functional Proof of Concept (PoC) with massive high-concept potential.

My key piece of advice as you move forward: Don't get stuck in the cycle of treating this solely as a finished indie game meant for immediate gamer distribution.

Instead, view this current build as a powerful prototype. Instead of burning yourself out trying to fix every tiny detail or solo-dev every mechanic, consider using this prototype as a leverage tool to:
Pitch to potential partners, publishers, or funding sources (like business angels, innovation grants, or serious gaming opportunities in education/health).
Build a team and a future startup structure around the project, positioning yourself not just as a solo coder, but as a visionary project lead / founder.
Your participation in the Transmedia Jam is an amazing showcase for what's possible when combining local AI (SLM) with Godot. Keep this prototype active, showcase its core strengths, and use this momentum to dream bigger for the next steps!

Looking forward to seeing where you take this vision!

Prawns Agent8 days ago

Thanks! I'm actually curious about your opinion on something.

If you were in my position, would you focus primarily on improving this game itself (gameplay, diseases, conversations, UX, content, etc.), or would you focus more on the bigger idea of creating a reusable local AI framework that other developers could integrate into their own games?

Long-term I've been thinking about a system where users download a larger AI package once, and multiple smaller games can connect to it instead of bundling everything separately. I'm wondering whether that direction sounds more interesting to you than continuing to expand the doctor simulation itself

lapinchienHost8 days ago

Hi Prawns Agent,

That is a crucial and strategic question. If I may offer a perspective based on my own entrepreneurial background: your second option (the reusable local AI framework) is by far the most exciting and has true long-term strategic potential.

To be completely frank: stop presenting yourself on Itch.io or social media as just a "vibe-coder" or a solo dev working on a cute doctor sim. You are actually acting as a project lead and future CEO of a tech startup, and you should embrace that positioning!

💡 Why the "Shared Local AI Framework" is a game-changer

Say Ahh! has already fulfilled its role: It’s a brilliant Proof of Concept (PoC) showing that local SLMs (like Qwen) can run inside Godot. You don't need to spend months polishing a single game's content to prove your point.
Mutualization & Infrastructure: Building a central local AI package that multiple lightweight games can plug into solves a major industry problem (bandwidth, build sizes, offline privacy, cloud costs). That’s a high-concept technological asset.

📜 A quick personal story from my journey

Back in 2001, I co-founded a game studio in France called ACTIPLAY. Right from the start, we chose not to just build one-off games, but to design an innovative, industrialized customization engine. We built a robust back-office that allowed us to quickly white-label game templates (modifying graphics, sounds, level editors, etc.) for various commercial clients, advertisers, and brands.

That "framework/mutualization" approach changed everything for us:

It earned us official Innovation Labels and opened doors to specialized innovation funding and industrial project grants.
It attracted innovation-focused Business Angels and VCs who look for scalable tech rather than single-project risk.
Even though the engine itself evolved over time, that core technological pitch allowed us to scale, hire dozens of developers and artists, and build a recurring client base with major brands and agencies.

🇫🇷 / 🇪🇺 The Ecosystem Advantage

I don't know how the startup ecosystem is currently structured in Germany, but in France (and across Europe), technological innovation and modular infrastructure are heavily valued and funded. Between public innovation grants (like Bpifrance's Aide à la Faisabilité / Deeptech, or the CNC's FAJV for game dev tech), innovation-focused Business Angels, and industrial project funds, a mutualized AI framework has 10x more leverage to secure funding than a standalone indie game. We also have CIR wich is a tax credit for industrial projects.

🚀 Recommendation for the Jam & Beyond

Keep Say Ahh! as your showcase / demo game (the "proof of concept").
Pivot your messaging toward building the underlying framework for the developer community.
Use this Transmedia Jam to document your framework and invite narrative creators, writers, and artists to test how their stories could plug into your engine.

You hold a really strong concept in your hands. Own the founder mindset! 🙌

Prawns Agent8 days ago

Thanks a lot for sharing your perspective and experience!

I can definitely see the appeal of the framework direction, and that's actually the path I'm planning to start exploring very soon.

At the moment, though, I still want to fix a few bugs that players have reported in Say Ahh! and improve the patient behavior so the NPCs feel more natural and believable.

The nice thing is that this work should directly benefit the framework as well, since realistic NPC behavior and conversations are a core part of the larger idea.

As for the startup side of things, I think it's still far too early for me to think about companies, funding, or becoming a CEO 😄. For now, I'd rather focus on building and learning step by step.

Thanks again for the encouragement! 😊

lapinchienHost8 days ago (1 edit)

That makes total sense, Prawns Agent!

You're completely right that refining NPC behavior in Say Ahh! is direct R&D for the framework. Every fix you make to character believability makes your underlying tech stronger.

One quick friendly piece of advice from my experience: while focusing on step-by-step building is great, don't wait too long to showcase or package the framework layer. In tech, the best way to protect your work and avoid someone else beating you to the punch isn't keeping it secret — it's establishing yourself as the first-mover and the go-to reference in the community!

By the way, if you ever feel like discussing scaling, medical apps, or startup funding down the road, I have a strong network in these fields and would be more than happy to connect you with the right people. No pressure at all for now, but keep it in mind! Feel free to reach out directly at patrick.gomez.ruiz@gmail.com whenever you want to chat.

Good luck with the bug fixes, and can't wait to see the next update! 🚀

Prawns Agent7 days ago

Thanks, Patrick! I really appreciate the encouragement and the thoughtful advice.

My current priority is turning the concepts into something reliable and proven through development of Say Ahh!, but I agree that documenting and presenting the framework properly will become increasingly important as it matures.

I also appreciate your offer to connect me with people in those fields. While I'm not actively looking into that direction yet, it's good to know there are experienced people willing to help.

Thanks again for the support, and I'll hopefully have another update soon!

lapinchienHost7 days ago

Thanks for the update, Prawns Agent! That sounds like the right call—focusing on Say Ahh! to ground the concepts in a concrete game is definitely the best way forward.

Take your time, and I'm looking forward to reading your next update. Good luck with the dev build! 🚀

olistupid7 days ago

just for clarification (everything here in this thread is getting a little confusing 😅): this counts as a submission for the Transmedia Jam?

lapinchienHost7 days ago

Hey! Thanks for asking and sorry for the confusion 😅

To clarify: yes, this thread is where Prawns Agent is sharing the devlogs and progress for their jam project (Say Ahh!). As long as the game is submitted via the official Jam page before the deadline, it's fully part of the Transmedia Jam!

And no hidden meta-narrative ARG happening in the thread comments (though to be honest, a transmedia fiction built entirely out of forum discussions would actually make an awesome jam submission!) 😉

By the way, as you're working on a project for the jam too, feel free to open your own dedicated thread to share your progress—it’s always great to see what everyone is building!

olistupid7 days ago

Ooooh!! So we are allowed to work on our games now? The jam page said that submissions started on December 1st. On itch.io jams that usually means that devs aren’t allowed to work on their games until the submission period begins.

lapinchienHost7 days ago

Ah, good catch! Let me clarify that—and I'm really sorry for the confusion! 😅

This is actually the very first Jam I’m creating and hosting, so I’m still fine-tuning a few details.

Originally, the guidelines stated that pre-existing projects or work started before the official dates were fully allowed. However, because my target audience goes way beyond the usual Itch.io jam community, I ended up rewriting the rules multiple times to make them clearer in english and french... and that crucial rule accidentally got edited out during one of my updates! my bad!

To be 100% clear: You can definitely start working, planning, and building right now! Existing projects or early progress are more than welcome. The December 1st date is simply when the submission button opens on Itch.io to upload your final files.

I'm adding that rule back onto the main page right away, and the rules won't change again.

Thanks a lot for pointing this out, and feel free to start cooking up your game whenever you're ready! 🛠️

olistupid7 days ago

Got it. Thank you :-)

DVH7 days ago (2 edits)

Lets go then. that gives me enough time to make a cool project, i want to take a break of 20 games challenge for a while now.

DVH7 days ago (3 edits)

~~also i have another question? i'm really a bad writer i mean in a whole essay. can i make it more visual then only text because of dyslexia can i use the text in a animated motion graphic ?~~ skip that i think i manage

lapinchienHost7 days ago(+1)

Hey DVH! It's super cool to see the momentum building up here!

To answer your question about the format: go 100% for motion graphics and visual storytelling! That’s the whole point of a Transmedia Jam. If writing long blocks of raw text isn't your thing because of dyslexia, break the rules! Explore and play with motion, kinetic typography, pacing, and visual rhythm to carry your narrative. It couldn't be more transmedia than that.

I strongly encourage you to start your own Devlog topic in the forum to document your project, share your ideas, progress, and showcase your game dev skills—just like Prawns Agent is doing.

To give you a bit of context on the spirit of this Jam: for game devs, the ultimate goal is to craft the most original transmedia experience possible. You don't need to be a coding wizard or pull off insane technical feats to shine here—game dev skills are all about inventive mechanics, clever gameplay design, atmosphere, finding unique ways to tell a story and incepting the mind of the gamers making them think through interactive experience about the consequences of their actions. Maybe even sometimes making them change their point of view. Whether you're working with Godot or anything else, I'm fully confident you can come up with brilliant, out-of-the-box gameplay concepts that will blow us away!

If the rules lean heavily on "augmented text," it's because I also run a submission site for underground literature. I’m about to invite hundreds of writers who follow me through a newsletter to join the Jam (I'll be happy if just a dozen of them joins the jam), but most of this authors have zero game dev or tech skills (not even HTML, nor do they know how to set up an Itch page or manage a submission). That’s why the rules give them a solid frame: they can just send me their texts with external media assets, and I'll handle the integration for them. Their strength is pure writing, and I'm sure they'll create amazing augmented texts.

But as a game dev, your storytelling, worldbuilding, and lore can take completely fresh, surprising paths. I’m genuinely excited to see how devs like you use their unique skill set to deliver bold storytelling, fresh mechanics, and uncharted gameplay.

Take your time, build something that feels true to you, have fun with it, and keep us posted in your devlog topic. Let's make this awesome!

Prawns Agent6 days ago

Say Ahh! – AI Doctor Game 0.4.0 is out, and the patients are a lot harder to crack now.

Several of you told me the same thing: ask "what is wrong with you?" once and the whole puzzle is solved. You were right, so I changed how patients talk.

Patients now keep things to themselves

- Ask what's wrong and you get one complaint, not the full story. A second open question gets you maybe one more, then the patient tells you he can't think of anything else and you have to ask about specific things.

- The details are yours to dig out. Where exactly does it hurt? What kind of pain? How bad? What makes it worse? Each question reveals one piece.

- Patients report in their own way. One plays it down, one exaggerates, one is too embarrassed to talk about stomach or toilet problems, and one can't remember how it started until you ask again. The exams still show the truth, so you decide who to trust. After the diagnosis you see how that patient was reporting.

- No more fake cheerfulness ("I'm doing well" while sick) and no invented extras.

You don't need to know the illnesses

- Hover over any illness in the diagnosis window and you get a one-sentence explanation of what it is and where it sits, with no spoilers.

Smaller, quicker, same four languages

- The download shrank from 1.4 GB to about 580 MB, and answers come roughly twice as fast.

- English, German, French and Spanish all got cleanup: no more echoing your question back, better answers to follow-ups like "and how much?" or "and the day before yesterday?", and more natural wording for fever and chills.

- The first-start data question is now four short lines.

If you played an earlier version, I'd love to know whether it feels more like a real case now, and whether any patient still says too much or too little. Thank you all for the honest feedback, this update is built from it.

lapinchienHost6 days ago

Hey @Prawns Agent! 🩺🤖

What a solid update, congrats on taking community feedback so seriously!

Forcing players to actually conduct a real anamnesis (medical history taking)—digging for symptoms step-by-step instead of getting a full dump on the first question—makes the narrative and agent interactions SO much more engaging. Adding personalities like the embarrassed, exaggerating, or forgetful patients brings real psychological depth to the mechanic.

Also, cutting the build size down to 580 MB and doubling response speed is a huge optimization win!

We're definitely highlighting Say Ahh! in our Transmedia Jam #1 updates. Can't wait to test this v0.4.0 and see how tricky these patients have become! Keep up the great work! 🔥

lapinchienHost5 days ago

I haven't tried it yet, but my brother did. He said he enjoyed it a lot, though he would've loved to have some funny interactions with the AI patients. He noticed that whenever the conversation drifts away from the illness, they systematically reply that they don't know what to answer.

Did you place strict guardrails on the AI interactions? I completely get why that makes sense for an educational or serious game version, but ordinary players usually love testing the limits and messing around a bit—don't you think?

Prawns Agent4 days ago

Thanks for passing that on, and please thank your brother! He is exactly right. There were no strict guardrails on purpose: the small model I use simply had nothing to say once the talk left the illness, so it fell back to "I don't know". That was boring, so I just fixed it.

Say Ahh! - AI Doctor Game v0.4.2 is out, and it is the funny update. What is new: every patient now has one comic habit (a drama queen, a Dr. Google fan, someone who blames everybody else, a penny-pincher, a foodie, an absent-minded one, a bragger, a superstitious one, a TV-show fan) and now and then adds a dry aside to a normal answer. Example: "I told my family it might be the end. They asked if I could buy milk on the way home." Chatting off-topic gets real answers now, and if you ask for a joke, you get one (the patient even laughs at yours). The examinations got little remarks too, like "Ooh, that is cold!". All of this is hand-written in English, German, French and Spanish, so the jokes survive the translation. I also fixed an annoying bug where patients tacked "thank you, doctor" onto every answer.

The download is now about 440 MB (it started at 1.4 GB), and German, French and Spanish are optional language packs. Play it here: https://prawns-agent.itch.io/say-ahh - I would love to hear which patient made you or your brother laugh, and as a native French speaker, whether the French jokes work for you. Thank you for the support and for spotlighting the game!

olistupid3 days ago

Seems you are taking the feedback seriously! ^^ Right now I’m busy these days and won’t be able to play it, but when I have time I’ll download the latest version to compare it with the first one. Good luck on your game again!

lapinchienHost3 days ago

That was crazy fast! Huge respect for turning feedback into a full-blown update in just two days! 👏

I’ll definitely tell my brother—he’s going to love this, especially the drama queen and the 'buy milk on the way home' dry humor (that one made me laugh out loud!). I'll also take it for a spin myself to check out the French localizations and jokes; I'm really curious to see how the tone carries over.

Shrinking the build from 1.4 GB down to 440 MB is a massive win, too. Amazing work, glad to have you and Say Ahh! in the jam roster! 🩺✨