There’s a few interesting points made at nesdev.org regarding how to handle the use of AI. And I quote.
AI rules
The purpose of these rules is to ensure this community remains focused on human creativity, encourage people to learn to develop for this platform themselves, and foster a welcoming environment for developers.
Vibe-coding and genAI art
Projects that are primarily AI-generated are banned unless they provide meaningful utility to the community.
If the human’s role in the project was primarily prompting AI rather than directly making a majority of the actual assets or code, do not post about it here.
If a major component, such as art, music, or code, was made by you, the project is allowed even if some other components are entirely AI-made. However, you must focus on the parts you made. We also strongly recommend collaborating with another developer on the parts you can’t do instead of using AI.
If you have something like an AI-written tool, script, or disassembly that is sufficiently different from anything that already exists and would make others more productive, that is allowed. Just please don’t come here to show off your vibe-coded version of something we already have and works well.
Disclosure
Projects that contain any AI-generated work must disclose what work was done by AI and what work was done by humans.
This refers to actual project content; using AI for external things such as ideas, code review, debugging, or tooling does not need to be disclosed.
For code specifically, disclosure is not required for small amounts of AI-generated code that can reasonably be argued as non-creative, boilerplate, or similar in scope to what you might pull from Stack Overflow. The intention is practicality, not purity.
This disclosure must be done here; an external page is not sufficient.
Disclosure isn’t required constantly, but is required at minimum when you first discuss your project in a thread.
You can get help here without having to disclose whether this ultimately benefits a project that uses AI, but do not use us to debug AI’s work.
If your project is entirely human-made, we encourage you to tell us explicitly; it will remove doubt and make people more interested. However, this is not required.
Responsibility
If you post AI-generated work, you are responsible for that work.
It is expected to be of good quality: you should thoroughly test it and ensure the AI’s decisions make sense.
You must respect the copyrights and licenses of others’ work.
Laundering others’ work through AI is not acceptable.
Writing
AI-written posts are banned.
Machine translation is allowed.
We prefer your writing over AI’s writing, even if you think your writing is bad.
We don’t recommend having AI edit your writing because that often makes it sound like it was written by AI.
No witch hunts
Do not start discussions to speculate or investigate whether a community member’s work is AI-generated.
If you’re unsure, you may ask once politely and move on, but please don’t make a habit of asking everyone.
If you have good reason to believe someone is lying about their use of AI, please reach out to staff with your concerns and evidence.
Don’t use others’ work as a soapbox for your opinions on AI. Be respectful to your fellow community members.
As I’m currently reading the developments and opinions posted elsewhere, I stumble into things that are interesting. So, I hope you wont mind me ‘flooding’ the discussion with those. They may be beneficial.
Anybody who wanted to already could make a game. I learned Z80 assembly through following a few tutorials online and it’s hardly an inaccessibly arcane skill to pick up.
What has changed is that people who previously couldn’t be bothered can now generate something with very little care or effort. And so we suddenly see a lot of careless work you’d expect from people who could never be bothered doing anything the hard way. This is not unique to AI tools, and the PC and console online stores were drowning in crap already, but until recently the retro community was somewhat insulated by there being a higher level of effort required to get started (and no money to be made).
I don’t doubt that people will do good things with AI tools but I don’t think much of it will be made by the people who “couldn’t” do something before.
And because we can no longer presume that something has had a degree of care and attention invested in it just by the fact of it existing, it will inevitably affect how all of our work is evaluated. Now if we see a screenshot or video somebody has posted, we can’t assume this is something that a person has been diligently working on for the last six months, it could just be the output of a prompt typed in yesterday.
Which is too bad if you really have been working diligently on something for the last six months.
Genuine question: Why would you start with the Atari 8-bit port? If you were going to do it, surely you just start with the arcade original? The idea of compromising the result by porting a port seems very weird.
I still think the bigger question is why at all? I can see the benefits of a manual port if you want to learn development or the GX hardware in particular. And I can see the benefits of an AI port of something where you are challenged and have to learn more about AI prompting, RAG, MCP etc to get a result (regardless of what it is). But if you can just single prompt a solution, where is the fun? Are people that desperate to play slightly inferior versions of classic games as long as they’re on their favourite system?
But I’m assuming that they’re not necessarily one shot prompts. Or if they are, there is a certain amount of effort going into crafting a prompt that actually delivers what they want. And some will be from the initial frenzy of “wow does this really work?”
Now I’m curious about why people are still doing it. What are they hoping to achieve? Or are they just late jumping on the bandwagon?
Why would people do it you ask? Well, after messages, then pics, then short videos… videogames is the new branch in social media content, I d say. The wiki isn’t social media though, nor its about fastfood consuming.
You know, ‘using AI’ literally means you (Dragon in this case) can pay a subscription to Claude and literally pass the prompt above and have the game baked for your platform. Why do you need someone else to ask Claude (and pay Claude) for the game you want?
But British 8-bit retro platforms are so astonishingly niche that there’s no real audience for such videos - at least not in the sense you can make money from it. And that’s before you get into how even more niche new games rather than nostalgia fuelled ones are. It’s like getting into currency forgery by making one pence pieces, it might work to some degree but it doesn’t make sense as a “get rich quick” scheme. You can see this with how many retro YouTubers pivot towards platforms like the NES (where there is at least a sizeable audience) or to retro TV shows which are better as nostalgia bait.
Heh. So, I guess the rhetoric of 2026 onwards will shift from ‘I wish a developer would make that type of game’ to ‘I wish someone with money can tell AI to make that type of game’.
What a great end to the CPC scene. I always imagined it would die with us. But I guess we are going to have more time for other things from now on.
@AndyCadley I don’t believe the goal is to make money. Not even in the trad social media. Vanity would be what I’d point my finger at. We’ve been through the ‘everybody can be an artist’ era of the Photoshops and the likes. I believe it’s not hard to guess how it will play out.
Je,seriously, i have a month of Claude. I have never have Claude early, i have copilot free,because Its free by the university. But It so idiot compared to Claude.
My experience from now.
To test It(Claude) i was try to port the hombrebrew levels from spectrum to amstrad of the laser squad. Something that its really a little nightmare to made alone manually. Because you need compare amstrad vs spectrum vs homebrew levels. To find the differences and to find the equivalente adress in ram. In the other hand you need manage the tr-dos and russian languajes from the 90 scorpion time.
You can made a level manually taken binary parts from amstrad and take binary parts from the russian levels, buy then you need to deal with change +100 sprites for each level and the internal IA if they modify It.
So as my goal is put in a cartridge all the homebrew levels can i found >35 that probably requires using the picogx with more that 512k.
I,ve given a Claude my amstrad dissasembly+ the spectrum dissasembly+ russian levels and i tell It taking my dissasembly as base and swap the parts that can be swap with the spectrum russian levels. And Port spectrum sprites.
Then i tell Claude made a cartridge as i usually do, putt It there as is the disc subrutine and we have space here.
Well the result was i ported the first level and it appears that a nuclear atomic bomb was released under the map, in winape appears more beautiful but in the gx4000 appears a radiactive zone XD.
So even with help of the ia It take much time to Port all levels, but at least was a posible thing that probably alone was tedius very tedius.
So i really don’t care what think at CPC power. I’m doing this because i like the game and even with IA it take a lot of time.
