Midway through 2025, roughly 74% of newly created content online was generated with the help of AI or bots, according to several large-scale studies. Only about one-quarter of online content is created by people without AI assistance.
And the rate of change is rising fast. By the end of the year, more than 90% of all content will be AI-generated, according to some predictions.
Just to point out that this is now a universal social problem on the internet. It’s something that every online community must wrestle with.
Okay, I’m not sure how the thread got to this point (and I’m not altogether interested in scrolling back to find out), but let me assure you that every Inform game I’ve ever written had garbage code hacks in it.
You can start out with a tidy plan (and you should), but when it comes to getting the last few bits of polish before release, the “polishing” is done with a big hammer and whatever bits of wood make it hang together.
A good IF coder is someone whose garbage hacks are well-separated – they don’t pile on top of each other.
I think you meant, “Is human writing code a part of the medium?”
My previous statement was, “Why does it matter who or what writes the code as long as it’s a good story/game?”
The About in my games will say something like:
Reflections was built on Sharpee, an IF authoring platform designed and
constructed in conjunction with Claude AI, but directed at every step by my hand.
This story was also coded significantly by Claude using Sharpee as the platform.
Since Sharpee is based on Typescript, one of Claude AI’s top tier abilities, the
development was not confronted with the usual GenAI hallucinations.
However: the story, plot, characters, words, grammar, descriptions, and every
other thing the PC sees was written and designed by me. Reflections has been a WIP
of mine for about ten years, well before GenAI surfaced and I knew every element
of game play ahead of time.
And as I said. I will not challenge any competition’s rules even if they might allow my stories. I’d rather let my stories stand on their own without the controversy.
Human understanding of the code is part of it, yes.
You are trying to turn this into a debate about whether or not external code tools are morally allowable but not all tools are created equal. LLMs are tools that, at their heart, blindly generate large quantities of text. The vast majority of people do not take the time to understand what they generate and it results in a substandard product. The supposed time savings require the output to go unquestioned! And then (per Draconis’s point) this output gets flung out into the world faster than anyone can read it. This is how most people are using genAI.
If you can pull off something great with Sharpee then that’s unironically fantastic but that is demonstrably not what most people are doing with genAI. They’re just flooding the zone, because that’s what this tool is best at.
Spring Thing forbids AI text and images but allows AI code, due to the preferences of the organizer. That would be an option here as well.
Personally, I don’t like AI-generated code (as you’ve seen), but I also don’t think there’s any real way of auditing this, so it’s reasonable to just not include it in the ban.
Hopefully the last time I post in this thread, just because I realized I never specified what my problem with genAI is in general.
I’d be against genAI in competitions even if it spat out masterpieces. I’d be against genAI in competitions even if they ran on 100% ethical renewable energy. I’d be against genAI in competitions even if you could guarantee that every single piece of training material comes from the person who’s making a game. This is because I believe that, at the core of it, a competition is a show of craft and skill, which are things that take time, care, and continuous honing. Whose skill are you representing when you’re generating text or images? You are but a player of the worst parser imaginable, typing >GIVE GAME TO ME. Even if you had an extremely ethical genAI completely trained off of your own writing/drawing, it only regurgitates what you made before. There’s no process. No lessons learned. No improvement. No connection. No achievement. All that is left behind is the empty feeling of gratification which will satiate you for fifteen minutes. You’ll have to pat yourself on the back because nobody else will do this for you. The room is empty because everyone else is talking about what they made somewhere else.
While speaking about aggressive behaviour: I am not sure, this is an appropriate post. Even as a board member, David can have his own view on things. That it does not fit yours, does not disqualify him for the board…
His opinion would not disqualify him from a pro-genAI board of directors but I feel it is disturbing for a board member of the interactive fiction community to dismiss the merits of handcrafted, human-made interactive (via code) fiction (via writing).
His argumentative behavior toward members of the group he is serving, however, is inappropriate for any member of a professional board of directors, I feel.
I don’t disagree, but I’m sincerely not comfortable making this thread about that. This thread is about AI in IFComp, yes? If we, as a community, seriously insist on discussing the individual merits or demerits of IFTF or any of its volunteers, which is significant and distinct from discussing the merits of individual ideas or claims, we should at the very very least make that a separate conversation.
To be clear, speaking for myself, I won’t take part in that conversation if and when it does happen.
I find it hard to see it that way. This is an exchange of opinions, and of course sometimes this might appear pressing, but in an exchange it is normal to defend your view. After all this is hopefully still one big group of I.f. enthusiasts, just discussing a very complicated manner. But he is still a board member for the whole group.
I did not perceive his posts as promoting story writing with LLM, only coding. And there are so many nuances to this, from generating large code blocks with genAI to assisted coding (like Copilot in vsc). I do not endorse or discourage either, but I think the community should have an interest in finding solutions that fit as many members as possible. Although I must admit that the whole genAI story is really a very difficult matter.
I think this thread has made it pretty clear that LLM enthusiasts are not earnestly interested in IF by any reasonable definition; they’re just looking for yet another creative community to stomp on and spam. Give them an inch and they’ll take a mile.
The amount of hatred exposed by this thread is unreal. Everybody associated with the IF community should be embarrassed and ashamed. If I was an iFTF member, I’d suspended all further competitions until people can learn to be more tolerant of each others opinions. Might as well have two IF Comp submission categories, No Gen-AI, and an AI-Open.
I realize I’ve gotten pretty close to this line myself, but we ask that everyone avoid attacking specific people. I also disagree strenuously with DavidC on LLMs for coding, but please keep these disagreements focused on statements and positions, not the people themselves, or extrapolating to what they “really” want.
If you think someone’s positions reflect badly on the IFTF, that’s an important topic, but this thread isn’t a good place for it.
To an extent, I share your sentiments but…please (a) do not make such blanket statements and (b) don’t be a backseat driver. Despite what we may say, the IFTF does hold the right to make their own decisions, and I’d imagine keeping the IFComp alive would be fairly high on the priority list. We have a year. It is possible that we come to a solution by IFComp 2026, and it is possible we don’t. Regardless, you’re being quite accusatory and probably hurting feelings and an escalation due to ad hominem attacks is the last thing we need here.
Daniel, is there really anything else to say at this point? On either side? The after-comp survey will be what it will. Maybe we put this particular thread to bed?
It’s not like there won’t be a new one within a week at most anyway.
Just to jump back to the main topic of this thread: I would like to throw in my support for a ban on LLM-generated content in IF Comp. I’m frankly shocked it was allowed in the first place – it’s completely against the spirit of the competition.
Penny Nichols is a perfect example of an entry that should be inadmissible, for the simple reason that the entrant did not write it. Everything that’s being shown to the player is happening in a model which was trained and implemented by other people. The text of the prompt is gibberish and does not contain functional game mechanics. And even if they did make logical sense, ChatGPT would be incapable of actually executing them.
Like even if you take the position that writing a short non-interactive prompt counts as writing interactive fiction, the author didn’t even write the prompt! As part of their “walkthrough” they included a link to the conversation where the LLM offered to write the prompt for them! You can see their entire creative process laid bare and it’s just empty.
How is this fair to other competitors who actually wrote things? How did this not earn an instant DQ? Like, come on. What are we doing here?
I do think this thread has mostly run its course, but I also don’t really want to get the last word in an argument against someone, then lock the thread. It feels mean-spirited. DavidC’s been attacked a decent amount here; the least I can do is give him a chance to respond to it.
But if another moderator disagrees, I have no objection to the thread being locked.