2.png)
Revision 2026 - The Power Within - April 3rd to 6th 2026
category: general [glöplog]
Quote:
I've asked the team to deactivate the compo rules page, as this is indeed quite heavily discussed by the team at the moment, as well as a general rules overhaul.
:thumbs_up:
Quote:
Furthermore I find it not that easy to draw the line at what constitutes problematic AI usage, in my opinion even before the current AI hype there were questionable technologies and behaviors by competitors in all categories.
I am sure they can work out a nice system and definition of whats not allowed.
Like "You must not use any system based on a LLM, like agentic coding and generative AI in any way, in any form. Doing so will render the demo disqualified and the author banned from future Revision events."
People managed to successfully draw the line when it came to sizelimited coding and the rules in whats allowed or not, like API usage, etc. I am sure this is a similar situation where cool heads will prevail.
Na dann ¯\_(ツ)_/¯
The topic of using AI-support for coding touches upon using examples or given code from tutorials, documentation and utilized frameworks. And what constitutes documentation at all and how it can be presented?
I reckon people would be surprised as to how many Amiga productions used given code like Promax' Blitter-line routine from ASM-One, let alone the music replayers.
I reckon people would be surprised as to how many Amiga productions used given code like Promax' Blitter-line routine from ASM-One, let alone the music replayers.
Quote:
You must not use any system based on a LLM, like agentic coding and generative AI in any way, in any form.
Doing so will render the demo disqualified and the author banned from future Revision events.
I'm not sure if some people really don't understand that, so some "news" from the coding side...
V-Tune is giving us optimization hints since the late 90s.
We use optimizing C compilers and don't write all code in Assembler.
We use editors with code inspection, spell check and syntax highlighting.
We use cross-compilers and linkers which output ROM cartridges and run our code in emulators.
And of course we are now using LLMs for coding assistance, reading the docs for us and producing uninspired example code which is sometimes even "good enough".
Now feel free to disqualify my entries and ban me from parties ;)
Last not least, I heard that some people choose techbro operating systems which made their owners billionaires, spy on users, harvest data, waste energy on bloat, copyright enforcement and gimmicky appearance...
To process data that is publicly available, was even made available by creators themselves, can never be theft. Go side with the DRM and content industry, and say so. It's anti demoscene.
To process data that is publicly available, was even made available by creators themselves, can never be theft. Go side with the DRM and content industry, and say so. It's anti demoscene.
Quote:
We use optimizing C compilers and don't write all code in Assembler.
That's actually really on point: Is an optimizing C compiler not equal to a sort of a specialized AI ? Will you ban it from being used...sure not, I guess...we simply got used to it...
...luckily I'm a 256 Byte assembler coder not dealing with that topic...though it might be interesting for you that we happily dissect each others code since forever and checkout sizecoding.org for reference and publish whatever secrets anyway...
...for me it's practically almost impossible to invent rules regarding code. Where does it start where does it end, where and how will you find the truth about an entry ? E.g....did the coder just ask ChatGPT what general idea he should do out of laziness, stole some algo idea elsewhere by AI or web search or ask for the breakthrough algorithm to speed up a square root or did he let the AI do the whole shit...and who wants to check all that in what time and who is judging it ?
The topic of AI-generated code for demos was discussed at length in this thread. While I won't claim that there was anything like a consensus, if I were to summarize the prevalent sentiment from that thread and from people I talked to (including party organizers) it would be something like:
- Including AI-generated code in a demo comes with legal issues (mainly: it is dubious whether you have the right to grant the party organizers the necessary redistribution license) and is broadly considered unethical in a competition setting (as a form of doping).
- Using AI-generated code for tooling is fine and likely unavoidable.
- Including AI-generated code in a demo comes with legal issues (mainly: it is dubious whether you have the right to grant the party organizers the necessary redistribution license) and is broadly considered unethical in a competition setting (as a form of doping).
- Using AI-generated code for tooling is fine and likely unavoidable.
It's more of a form of doping on generated images but not so much on code, in terms of winning compos with zero skills. You can easily with zero artist skills generate an image and if nobody knew win compos. But I don't see the same with code, nothing can insta-generate a complete demo. You have to babysit the LLMs 1000s so much that it's better to write it on your own. And what you would get would be so lame anyway that won't be winning any places. I don't see anyone managing to successfully cheat in code with the present tools. But I guess it's more of an ethical issue this one.
Maybe we should discuss all of this in another thread and don't spam the Revision-thread anymore.
Quote:
and who wants to check all that in what time and who is judging it ?
You'd be quite surprised.
I did some forensic analysis on a commercial product's firmware which found out to use GPL code I contributed to, under the GPL....Eventually the company admitted indeed they used my shit and made all code public.
Was done purely by disassembly and manual analysis, by me to match firmware compiler's basic blobs to ARM generated GCC code made for elsewhere. And matching function names, etc.
So yes, finding code generated by LLM is not too far off. Just takes someone persistent enough.
Quote:
Maybe we should discuss all of this in another thread and don't spam the Revision-thread anymore.
OK who's already planning a Revision demo? I wasn't really, but just seeing the thread got me hyped and maaybe just a small one? Just sofascening this year though.
Quote:
OK who's already planning a Revision demo?
Always planning, but haven't even decided on the platform yet :)
Quote:
Always planning, but haven't even decided on the platform yet :)
let’s keep the DOS-season going :)
(my DOS demo is in progress)
There is only one platform :)
Looking forward to see my demoscene family again :-)
Started slowly working on something. Apologies in advance…
