Index » Titanic! » Development » Rules proposals on AI usage on Titanic

Hello everyone! With the prominence of AI generated tools and services, agents/agentic coding, &c. in daily life, and the public distaste (including the Titanic community's) for this usage of AI, I'm presenting a couple of new rules proposals.

1. AI usage disclosure
This proposal is for requiring all creation or development involving use of vibecoding/agents to be properly disclosed before being posted on the forum. This does not include using AI to research or other tasks unrelated to the creation or development process.
We could have something like [AI] in the title as the disclaimer and that would be enough.

Here are the options you can vote for this proposal, ordered from most strict to least strict ruling:

• Option 1 - Any AI usage, regardless of extent, requires an AI disclosure.
  
• Option 2 - Usage where AI is the primary method of creation or development (vibecoding, agentic development) requires an AI disclosure. (This means it can be used to a lesser degree.)
  
• Option 3 - There should not be a requirement for an AI disclosure.
  

2. Vibecoding/agentic development ban
This proposal is targeted more towards our passionate developer community here, and suggests a complete ban of vibecoded/agentic developed contributions to Titanic's codebase and software posted on Titanic's forums.
Some developers may actively use AI for simple tasks, automating boring busywork, or debugging. These cases are not included in this proposal and should be a discussion for another time.

Here are the options you can vote for this proposal, ordered like the previous:

• Option 1 - All vibecoding/agentic development is prohibited on Titanic forums and codebase.
  
• Option 2 - Vibecoding/agentic development is prohibited, with the exception that the human controlling the AI is able to explain in their own words what the code does and why.
  
• Option 3 - Vibecoding/agentic development should not be prohibited on Titanic.
  

I will personally refrain from commenting on what I think should be done and remain neutral on the proposals.
When replying, please include which option you think is best for the proposal, or if you believe there should be a new option. I'll add any new options if they're distinct enough!

this shouldn't be a rule but rather a contributing policy somewhere in git repo.

Clarification, how does this work in practice if someone has a development project where they add a ui that is entirely vibecoded, for instance? Is that a primary use since it is all of the ui, or a minor use since the rest of their project is written by them? 

if an application requires a UI to function and the UI is vibecoded then that can constitute a major portion of the software being vibecoded

If someone is contributing in a significant way to Titanic, of course I think it is important that they know what they're doing, whether they are using an agent or not. & In my personal life, I stay away from using AI in some areas, because of my own values. But I do worry a bit about the environment this could foster. There are some cool osu websites and projects for bancho that I suspect were created at least partially with AI. If someone were to have to disclose regardless of the use case or project, and this disclosure created a hostile environment, would some people just not attempt similar projects for Titanic, or attempt to mask or hide their use of the tool? If the environment can stay respectful so that such disclosure is actually viable, then I would vote option 2. on AI usage disclosure.

As far as a ban, I don't think an overall ban is appropriate. If someone is contributing to critical infrastructure, perhaps AI should not be used. But if someone is just making a silly little tool, do they really need to even fully understand every single bit of the code? Perhaps I am out of the loop here, as I am not a developer.

As a vibecoder myself, I think it's still important to understand the majority of what your code is doing. Even if an LLM gives you something that works, it's worth asking it what it does, why it does it, and how it works. That way if something goes wrong, you'd be able to understand the problem well enough to ask better questions, verify the solution, and catch when the AI is wrong.

Learn from my dumbass mistake of the b99 client port to Windows 98, it did something I didn't even consider or know existed until my ban and Levi told me what I did.