Well, as per subject. Seeing them coming one after the other, maybe we should create a subcategory that will only include AI releases?
But then we’ll have to define what an “AI release” is. Some of them are obvious, when the AI does all the work, but where do we draw the line?
For me the line for going into this separated category is if some parts of the end game is generated by AI and author doesn’t know how to modify it if needed…
Code that author cannot correct.
Graphism he cannot update because he can’t draw a pixel.
Sounds/music he put inside.
On my side, I’m a programmer, I cannot do sound / graphisms. So if I ever publish something, I may ask AI to help with some subroutines or hard debug, but I will at the end own the source code enough to be confident to say it’s not AI, it’s mine.
But if I let AI do graphism and sounds, I totally agree that my work should finish in this category.
I think that if a program or game contains even a single line of AI code, it belongs in the dedicated section.
That’s a very hard definition. Someone may use AI a ton to help them but write the program themselves. The second may do everything by hand except for a tedious, say, data section for which they may ask a LLM to encode. Is the second AI and the first not?
Maybe just create a global AI topic.
I do think that the developer/team should be up front. It is up to them about how much they say AI actually did, and ‘AI assistance was used’ or an AI tag is probably mostly ok.
As to what defines an AI release, I am thinking more the AI made a code or data contribution that made it into the final release. If AI made something and the coder later modified it (after understanding) and it changed significantly then although initially assisted is it an AI release anymore?
I think if AI was used for build tools, or to make a level editor, maybe that doesn’t count.
Poll added to OP
If it sounds, looks or feels like AI it should be in that AI group.
And if it doesn’t there is probably enough human intervention that AI was “just another tool” under control of the developer. I don’t mind if AI was being used for code, graphics or sound, if it’s been polished (via good prompts or by hand with actual programming, pixel gfx or with a sound editor) to a level that the game is actually “good”.
Maybe similar to a group leader for a project at work or university. They can be part of the solution - or just want to be praised for someone elses work. If a group leader is deeply involved in design, decisions and was crucial to the output, it’s a different story from a group leader that tells the group “here’s the idea, no just do it”.
Just proper ai tagging works for me, and not just for games. I don’t think it’s worth the hassle to have questionable AI categories right now, let see how things turn out in a year or two. But I’d really love the option to hide certain tags 
I think this is basically how I feel, it’s about the creator understanding what’s going on so they can tweak it by hand rather than just telling an AI what to make for them.
I agree with the “actual programming” part, but not the “via good prompts” part, unless the original code was human-created in the first place. It’s not just about the game being good, as some AI-created games are much better than others, but IMO it should still be tagged as AI.
Given that identifying it relies on the author’s good faith, the “AI” category must be included whenever any part of the “code” that isn’t purely aesthetic—meaning a functional part of the program, excluding the loading screen or decorative elements—was generated by AI.
The author can then specify in the game description whether they assisted the artificial intelligence by fixing errors or detail their level of involvement. Furthermore, authors should ideally include the term “AI” on their own initiative to avoid suspicion or misunderstandings.
That’s a good idea. I’ll try to see if I can set up a template for that info.
But as you say, this relies on the Dev’s good faith as well
Ok, looks like we can have a template for the OP of the subcategory. Something like:
type: radio
id: ai_involvement
attributes:
label: What was the use of AI in this project?
choices:
- Completely vibe-coded
- Some routines
- Some coding, with human review and editing
- Only for tools / build scripts
- Only for ideas / design
validations:
required: true
type: textarea
id: game_details
attributes:
label: Game Description & Human Involvement
description: Detail what the game is and where you intervened (fixing bugs, optimizing loops, etc.).
validations:
required: true
Could you people suggest some options to be picked in the form? The ones I put are only samples to test it in the back end.
Identifying it just adds extra information and context to these types of creations over time. AI-programmed games are also part of our beloved AMSTRAD catalogue. Heck, if we can feel affection for horrible “Speccy-crap” ports made with someone’s foot, at least these ports work pretty well.
That’s a valid point.
The way you’ve designed it looks good. Down the road, as this way of working evolves, it might need some addition or field—for instance, the type of AI. Different “social” classes of AI are already emerging: cheap, restricted ones, and others with much more freedom and power. But that’s for later. Unless Mr. SKYNET takes over the AMSTRAD scene and dictates otherwise 
Heh, I had actually added your suggestion already. The form currently has these options:
- "Full Vibe-Code (AI generated almost all game code)"
- "Hybrid (AI generated key routines; human assembled/edited)"
- "Port / Translation (AI converted code from another platform/language)"
- "Debugging / Optimization only (Human wrote, AI analyzed/fixed)"
- "No AI in game code (Human coded; AI used only for assets/tools)"
And the Assets and Pipeline Involvement has these:
- "Loading screen / Title art"
- "In-game sprites / Background tiles"
- "Music / Sound effects"
- "Build tools / Asset conversion scripts"
- "Story, dialogue, or manual text"
- "None (Code-only or pure human assets)"
Then there’s a text field for LLMs, tools etc and finally a field for the main post.
This is what it looks like:
I think I’ll be making it public soon unless there’s some strong disagreement.
You’re an ace
Ok, looks like we can have a template for the OP of the subcategory. Something like:
The form currently has a typo. Hint: s/deails/details/.
What is the result of filling the form? Can viewers filter posts based on it?
The principle of a declaration looks nice. That said, we might be reinventing things that are being done elsewhere.
I see at least two projects:
After only a few minutes reading, I find the first one clearer.
Such a declaration could be included in any program release and source code repository.
As others said it forces to draw a line. I’m afraid wherever we draw the line it will feel wrong somehow. It could even feel wrong different ways to different persons, because each person may have different wishes about what they want to see and not see.
Perhaps a nomenclature of tags would be better suited?
Any viewer can see tags (or even filter their view based on tags) which allows in one glance to know what to expect.
I hope this solves many issues like " 
I don’t know how to receive this prod until I’ve got a hint about how it was made.“.
It would replace " 
I’m hunting for signs of AI”…
with " 
This project honestly declares use of AI for graphics but mentions code is made by hand, that’s a valuable effort".
In pseudo-regex form, inspired by AI-DECLARATION.md | Machine-Readable Standard for AI Usage Transparency :
ai-(code|titlescreen|graphics|sound|language|story)-(none|hint|assist|pair|copilot|auto)
For example a topic about a prod might set tags:
ai-code-hint
ai-graphics-none
ai-music-none
But even of that I’m not convinced: I can have a first draft written by AI (so, fully vibe coded) and massage it to a near-complete rewrite. What should I declare then? Declare about the final result? That surviving LD A,(HL) was originally written by AI but since it’s now in a fully rewritten context it’s ai-code-none?
Or start with a fully AI-generated image and rework it extensively. What should I declare then? ai-graphics-hint? Something else?
IMHO “build tools and asset converters” are out of scope here? Or please explain.
Who care so far which tool you use to convert modern formats? If some graphics is generated by AI based on another graphical source, then it’s covered by the graphics tag.
And build tools… to push it only slightly further, would I e.g. have to declare that my tools run on Linux because Linux has some code written by AI? I don’t see the point.
After much hesitation, I’d say now let’s try tags and not a separate category like “this post has some kind of AI at any place inside, push it to the AI-tainted category”.
Not saying it’s the only true answer, only time will tell.
That said, a category to discuss AI stuff is a good idea!
Replying to myself I might have found an answer.
How about:
So, the category to create would be: “AI-related discussion”, “AI experiments” or something.
I guess this might resolve a lot of friction? What do you think?