General policy in regards of the use of AI (LLM)

[luginf]

Is there a general policy towards using LLM for contributing to TIC-80? I'd really like to work on the inclusion of the FORTH language into TIC-80, but if AI is not tolerated it doesn't worth to try it :)

[imsys]

Using AI to code is inevitable and probably most everyone is using. I have done many contributions to TIC-80 using AI and I have disclosed it when I thought I should.

The main thing is that, at least currently, LLMs do not assume reponsability. So we do have to review, test and assume reponsability for the chuck of code we contribute.

When contributing to something in the core mechanics of TIC-80, we do need to understand a bit of how nesbox has coded it to not make things out of place. But as AI is improving over time, this is becoming less of a concern.

But a thing I'd like to mention about adding a new language is that it can make a bit more difficult to maintain, specially porting to many different devices. You can see the struggle we had at libretro#28

And here is a table with the number of TIC-80 games written in different languages. (March 2026)

Language	Quantity Games
Lua	1427
Js	128
Fennel	57
Wren	24
Python	22
Moon	13
Janet	9
Squirrel	6
Ruby	5
Wasm	5
Scheme	1

[hexabeeretro]

I use AI like Gemini, Qwen, etc., when writing all my games. I hope it's not banned; if it is, I'll feel really lost. I might even give up game development.

[Skeptim]

For your games you can do whatever you want, you're free, no worries.
The question is about using AI for contributions to the code of TIC-80, the software itself. Because AI can make mistakes, cause security issues and that can be harder to maintain a code written by AI.

[NNNN618]

Base on my experience, communities are not tolerant of this kind of ai-coding stuff , so I suggest you try something else.

[ddelemeny]

I'm going to chime in on this topic with an opposition to AI on principle. I'm not writing this to antagonize anyone, I only want to express why I would prefer that this project, to which I'm only a minor contributor but which I am fond of and is meaningful to me, steer clear of what is called AI nowadays.

For anyone who cares about checking me up, my line of contributions to the project is pretty sparse and narrow, mostly related to a nibble in VRAM I named the blit segment. I crafted my PRs with a lot of care because I wanted to match the quality of the codebase (I remember complimenting Nesbox on how much of a pleasure it was to work on these files) but I'll openly admit that haven't been the most reliable steward of the features I pushed onto the project. I'm glad that it hasn't been reverted and that some people embraced it.

When I started tinkering with TIC-80, I was engrossed by the whole idea of creativity under constraint. The idea of reading spritesheet data with lower bit depth wasn't new, but it wasn't really viable without proper support from the tic machine. The whole concept of the blit segment feature to me was to provide a feature tailor-made for tinkerers, a balanced helper that makes it possible to handle the bitmap budget a bit more tightly without the prohibitive CPU and character cost of doing it from within the language runtime, neatly interlacing and interacting with the design of existing features so that you can feel that the capacity of doing somewhat advanced things on that front is intentional.
It's been my love letter to the project, the concept of fantasy console and the community who gathered around it in 2020. It's infused with an ethos : learning IS fun, making IS fun, playing with constraints IS fun. I wasn't a game developer, I hadn't touched a line of C since university, but I cared enough about adding my little hand-carved stone to this edifice, to be found and manipulated by people of the same mindset. It's a piece of me added to the commons.

This is roleplay. Fantasy console have no reason to exist but to allow us to be LARPing as pioneers of the early computer era, modern language runtimes and extremely powerful host machines being the blunt weapons we use to pretend to be learning to do amazing things within the constraints of a limited environment. It's a performance. It's an art form.

Now I see people on the regular trying to push things they do not care enough to actually learn how to do by themselves, and I'm asking : what are we even doing here ? What is the purpose of this project if its users and contributors aren't in on the role-play anymore ?

If you are a contributor to this project since before ChatGPT rolled out, what did it mean to you back then ? did that change ?

To me, this is a project for artists and tinkerers, to which I find the use of generative "AI" to be the antithesis. I contributed as a gift to a community I thought shared the same values, and if these values do not mean anything nowadays... then I have no desire to contribute anymore.

Now if my interpretation of the purpose of the TIC-80 project doesn't sway you, please spend a moment thinking about the social and environmental cost of this technology. Please spend a moment thinking about the hazards of cognitive offloading and ever increasing reliance on a thinking crutch. Please spend a moment to ponder what it means to create without labor. Please spend a moment to reflect on how the community that created the immense amount of knowledge that's blended and incorporated into a language model is just erased when you interact with a chatbot. Please spend a moment to realize that this technology is currently making personal computing a luxury and its proponents openly push the commodification of thinking and creativity as a model.

Nothing about it is inevitable. Nothing about it is even remotely desirable to me, so here's my vote about the use of LLMs in the project : please don't.