Aqua's Blog

Alright, folks, drum roll please! Life Like This is officially complete. It will be dropping tomorrow in the marveycord and on this blog tomorrow, July 3rd, at 09:00 CST. Please share it far and wide as you see fit. I hope everyone loves it!

Artist: Karel Gott

Nicky finds out he's pregnant two hours before they leave for South Sudan.

For centuries alphas and omegas were persecuted until they became a relic of the past. In the present, Joe and Nicky are taken by Merrick and soon their captors will find out that immortality is not the only thing unique about them.

Or; the events of the The Old Guard (2020), but omegaverse (with a twist).

Now let me be clear - Gen AI is awful and most people agree that it has no place in creative spaces, but that's besides the point. I'm not here to debate AI in creative spaces, I'm here to talk about the moderators who feel the need to investigate and punish writers who they believe use AI.

I'm going to start by reminding people of two important things:

There is no way for anyone to know for certain if someone is using AI to write - unless they tell you they are. The common "tells" that are cited online are found in the work of people who have excellent control of functional grammar, good vocabularies, and an excellent control of sentence structure. AI is trained on the work of talented writers.

There is no way for an author to defend themselves against a claim that they are using AI, especially in an online environment. Nor should an author have to. The rule should not be "prove you are innocent or we will find you guilty!"

I feel very passionately about this and have for quite some time. I've seen fandom witch hunts before, and I know how easy it is for some people to climb to moral high ground and punish those they believe are doing the wrong thing.

My friend recently chose to leave a discord server after the mod team contacted them to announce that there had been reports made by people who wished to be anonymous who suspected their work was AI generated. The mods explained that they had investigated these claims by looking over their writing, and although they admitted they could not be 100% certain the work contained AI, they were issuing a strike/warning to them in accordance with their discord rules because they believed they might have.

The investigation involved them reading some of their work. 

That was it. 

They read their work and decided that, in their own words, they thought it might be AI so they were issuing a strike. They then proceeded to delete the writing they had shared on the server without any opportunity for this person to retrieve it.

There was no communication prior to them issuing a strike. My friend was not told about these accusations ahead of time, the moderators did not talk to them about the accusations or raise concerns, and did not stop to ask themselves why the people making these claims wished to do so anonymously.  

Perhaps if they had, they would have learned that members of their discord had been harassing this person via Tumblr anons about this issue. Instead they chose to give the bullies anonymity, to deny a writer a chance to have open communication, and issued a punishment for a rule infraction that they themselves admitted they could not be certain had been broken.

When my friend gave the names of beta readers and editors who had worked with her in real time, these moderators chose not to engage with them or seek clarification, and instead doubled down and tried to justify their decision.

My friend asked for the evidence they had that their work contained AI content, and it was not provided to them. Instead they supplied a generic statement about how they had made their decision based on their use of syntax, grammar, tone, and word choice was rolled out. But again, they did not give examples or explain this. 

As someone who admins and mods several creative spaces, this kind of behaviour worries me. Actually,  it terrifies me. Because anyone can be next: you, me, the brand new writer eager to share their work,  or the fandom oldie. 

It sets a dangerous precedent for fan spaces and the policing of creative works, because it leaves writers in these spaces open to harassment, bullying, intimidation, and censorship under the guise of keeping fan spaces "AI Free". Anyone can accuse  someone of using AI to write, and this can be used to bully people out of fan spaces. 

Moderators are supposed to keep fan spaces safe for members, not take it upon themselves to play detective and police creative works in case someone is using AI. In fact, mods taking it upon themselves to police writers in fan spaces by handing out warnings and bans are making fan spaces unsafe and encouraging social exclusion. This kind of behaviour will actively scare and intimidate writers into simply walking away - because who wants to hear an accusation that their work is AI?

Facilitating AI witch hunts is killing fan spaces.

Attempting to police the writing of others is alienating writers. 

So putting aside the not-so-gentle slide the argument of "guilty until proven innocent" is into fascist thinking (which is honestly another good reason to sit and re-evaluate this position if you have it), here we go.

I'm a labor organizer. I've not been terribly vocal about that on tumblr, but that's been my job for the last 10 years, and so I'm approaching this through a lens of workplace justice to highlight just how fucked up it is to "do an investigation" that never involves asking the accused a single goddamn question.

Just imagine for me showing up to work one day, and your boss pulls you into their office. You sit down, maybe a little nervous, and your boss slides a piece of paper across the desk stating you have been formally written up and this is your first of three strikes. Three strikes and you're fired.

The offense? A coworker claimed they felt like you were eating someone else's lunch.

No one ever stopped to ask you, "are you eating coworker's lunch?" The person who's lunch was being eaten was also never asked if their lunch was being eaten. All the mattered was the feelings of the person who reported it, and the boss who decided well, you could have done it and so they issued you formal discipline, and now you are that much closer to being out of a job.

But the thing is, you never ate anyone's lunch. You make your own lunch, it's clearly labeled "my own fucking lunch" — yet none of that mattered, and no one even bothered to ask.

You'd be furious, right? That's not what a proper investigation is. The standard in any space should not be: we are presumed guilty and must, therefore, prove our innocence. But in my fictional scenario — and in the real life scenario above — the accused was never even granted the chance to prove their innocence. In my fictional scenario, I'd be writing up the grievance yesterday and I would fight that case all the way to arbitration and I'd fucking win in five minutes.

The notion that we must police other people's creative works in creative spaces and issue strikes (especially without doing any actual fucking investigation that is an actual fucking investigation) is harmful and you know what, fuck it, I'm not gonna go soft on anyone's ass — it promotes fascist thinking. If you engage in this type of behavior, if you believe that people must prove their innocence to you (while denying them the opportunity to do so), you are that much closer to being a fascist. You're not saving creativity, you're not a champion of what is just and good in this world.

(J. R. R. Tolkien)

The fat cat on the mat

may seem to dream

of nice mice that suffice

for him, or cream;

but he free, maybe,

walks in thought

unbowed, proud, where loud

roared and fought

his kin, lean and slim,

or deep in den

in the East feasted on beasts

and tender men.

The giant lion with iron

claw in paw,

and huge ruthless tooth

in gory jaw;

the pard dark-starred,

fleet upon feet,

that oft soft from aloft

leaps upon his meat

where woods loom in gloom —

far now they be,

fierce and free,

and tamed is he;

but fat cat on the mat

kept as a pet

I am once again begging people to realize that AI checker doesn’t work. it’s never worked. it’s notoriously known to have flagged human-made works as AI and AI-generated works as human-made. and by feeding it people’s works, you are feeding more works to AI, because apparently the machine itself is AI.

This (as well as the not very off chance that pupils/students do use AI in their various assigments *without* permission and *without* labelling something as AI generated or AI assisted) is why as we speak (and they literally started years ago) educators all around the world are using all their experience, knowledge and creativity to re-shape how (and what) we teach and test.

Not sometimes. Not until the next update. Always. They proved it with math.

Even with perfect training data and unlimited computing power, AI models will still confidently tell you things that are completely false. This isn't a bug they're working on. It's baked into how these systems work at a fundamental level.

And their own numbers are brutal. OpenAI's o1 reasoning model hallucinates 16% of the time. Their newer o3 model? 33%. Their newest o4-mini? 48%. Nearly half of what their most recent model tells you could be fabricated. The "smarter" models are actually getting worse at telling the truth.

Here's why it can't be fixed. Language models work by predicting the next word based on probability. When they hit something uncertain, they don't pause. They don't flag it. They guess. And they guess with complete confidence, because that's exactly what they were trained to do.

The researchers looked at the 10 biggest AI benchmarks used to measure how good these models are. 9 out of 10 give the same score for saying "I don't know" as for giving a completely wrong answer: zero points. The entire testing system literally punishes honesty and rewards guessing.

So the AI learned the optimal strategy: always guess. Never admit uncertainty. Sound confident even when you're making it up.

OpenAI's proposed fix? Have ChatGPT say "I don't know" when it's unsure. Their own math shows this would mean roughly 30% of your questions get no answer. Imagine asking ChatGPT something three times out of ten and getting "I'm not confident enough to respond." Users would leave overnight. So the fix exists, but it would kill the product.

This isn't just OpenAI's problem. DeepMind and Tsinghua University independently reached the same conclusion. Three of the world's top AI labs, working separately, all agree: this is permanent.

A couple of years ago, during a seminar on gen AI my then company hosted for us, the employees, a participant of that seminar asked: "Well, I've heard that AI hallucinates quite often. Is that true?" -- The docent answered: "No, you got that quite wrong. AI 'hallucinates' ALL THE TIME. Like, literally, ALWAYS. The 'hallucinations' just are surprisingly spot on most of the time."

And there you have it. AI (as in the LLMs we mostly work with nowadays) doesn't 'know' anything. It doesn't even know that it doesn't know anything. It's the anti-Socrates, so to speak.

In 2025, we responded to the infuriating influx of AI-generated slop in creative spaces with the following policy: Our Stance on GenAI. Given the significant developments in GenAI since last spring, we thought it was important to issue an updated statement and remind you where we stand. The TL;DR is that we’re more anti-GenAI than ever; please see below for our expanded policy for FTH 2026 and onward.

To start, let us firmly reiterate:

Generative AI has no place in FTH. It is not welcome here, under any circumstance.

Non-exhaustive list of unwelcome slop generators:

image generators like Imagen, Midjourney, and similar

video generators like Sora, Runway, and similar

LLMs like ChatGPT, Gemini, Claude and similar

vibecode generators like Copilot, Roo, Replit, and similar

audio generators like ElevenLabs, MusicLM, and similar

text editors like Grammarly that are known to be powered by LLMs

Starting this year, the following policies apply to all creators (including those in ROFL) and bidders, respectively:

For Creators

Participants found to have used generative AI to produce a fanwork, in part or in whole, for their bidder(s) will be permanently banned from participating in future iterations of Fandom Trumps Hate. This includes fan labor of all kinds.

We strongly advise (in FTH, but also more generally) that creators of all mediums keep some kind of work history, insofar as possible (we understand this may be more difficult for fan labor). You likely already do, intentionally or otherwise! Documenting your fanwork progress is one of the easiest ways to prove that your work is truly yours, should the need ever arise. To be clear, we don't require that you provide your bidder with intermediate stages of your fanwork—that is entirely up to you.

For Bidders

Participants who deliberately introduce any AI-generated materials into the fanwork creation process (see below for concrete examples) will forfeit any unfinished portion of the won fanwork and will be permanently banned from participating in future iterations of Fandom Trumps Hate.

This includes (but is not limited to):

giving a creator an AI-generated prompt (of any medium) (e.g. requesting a fanfic based on an AI-generated image, generating a fanwork prompt using ChatGPT, etc.)

requesting fan labor be performed on AI-generated material (e.g. beta-reading a slopfic, getting help with a vibecoded workskin, etc.)

requesting a fanwork (of any medium) that is intended to accompany AI-generated material (e.g. cover art for a slopfic, photo manips of AI-generated images, podfics of slopfics, etc.)

feeding creators' fanwork(s) into a generative AI model

We are keenly aware that since last year, existing beloved tools and workflows have had unasked-for generative AI features forced into them, with varying degrees of transparency about whether or not they can be shut off or excluded. Our ask is that all participants make an effort to avoid using features known to rely on generative AI in anything even peripherally auction-related. Familiarize yourself with the tools you work with and use your best judgement; if a new feature would make your fanwork trivially easy for someone without your skill set to create, it may rely on generative AI.

A Reminder:

The more generative AI intrudes into creative spaces, the harder it becomes to distinguish between slop and human-made work, even to a trained eye/ear. We absolutely encourage everyone to educate themselves on how to get better at spotting it. That said, it can be very tempting to over-correct and default to the assumption that anything you come across with certain characteristics is AI-generated. We ask that everyone strive to assume that your fellow FTH participants are operating in good faith. Not every piece of text containing em dashes* was generated with ChatGPT; not every video with a cool, improbable-looking effect was generated with Sora; not every piece of audio with a noticeable skip or glitch was generated with ElevenLabs.

Please be kind to each other. If you suspect your fanwork or prompt contains AI-generated material, please reach out to the mods via email so we can address situations on a case-by-case basis.

We recognize this policy may seem unusually rigid in comparison to our others; most aspects of FTH operate on an honor system, and in general we try to be as flexible as we can in our policies to allow for the best experience possible for all FTH participants. This, however, is something we are not willing to be flexible on. We ask that you trust we have given this serious consideration and respect that while we are always willing to answer clarifying questions, we are not open to debate on this topic.

Why do we care about this?

Am I reading this correctly? Are you guys (gn) honestly advising (aka telling) me "to keep a working history"? Seriously???

I mean, while I might understand your point re GenAI (which doesn't necessarily have to mean that I share your stance re this (which I do, but that is not relevant / the point here)), I am honestly stunned.

What do you mean by this?

Do you mean that, "just in case" anyone might think it a good idea (or just an idea) to "accuse" me (or any other creator, for that matter) of "AI use", the onus of "disproving" said "accusation" is on me / the creator? Or the prompter, for that matter? Aka that I, who have (or any hypothetical creator who has) been writing (or art-ing, for that matter) the same way for 25+ years by using the same word doc (or whatever tool artists use) to write (or art) a fanwork from start (= jotting down ideas) to finish (= the final fic I'm sending off to my beta), have to break it down now into a gazillion of 'stages' of work to 'prove' in case of 'accuse' that I have not been using GenAI?

For realz????

I've been participating in FTH ever since it first launched.

This makes me think that maybe I should consider not doing that anymore.

This is absurd; it is an absurd piece of 'advice'.

Like I said, whether or not I agree with your take is not the point. And I absolutely appreciate and love what you're doing with FTH, and that's not the point, either.

a decade ago people didn't have tumblr groups. people didn't even have dms. if you wanted to talk to anyone about anything you had to make a post, or send an ask (which more often than not would get published and thereby become a post in the end too)

the exclusivity, even if a server is not "by invite only" - I really, really miss tumblr in that respect. Open bar, anything can happen.

the idea of having channels - It feels like trying to have a conversation in windows explorer. And it reminds me a bit of that "Gilmore Girls" ep in which Rory and Paris move into their dorm at Yale and Paris tries to convince Rory that their room should have a motto for dorm party night.

the concept of mods* - I am over 50 years old. I don't want to be "moderated". I am capable of deciding on my own what's "appropriate" to say and what not.

*n.b.: I'm not saying the mods or the servers I'm active(ish) in are 'bad mods' or anything like that. I just dislike the concept in itself.