morganmuffle

Pedestrian traffic lights

sometimes I’m reminded that there are still people who don’t know ao3 was literally created by incest shippers — and the site’s sole purpose is to 1. be completely against censorship and 2. host all kinds of dark, taboo fics that are banned on other platforms — and the first ever fic that was posted on ao3 was a fic about an incest ship from supernatural.

you are in the house that was created by freaks. for freaks (affectionate). every disgusting thing you can think of is rightfully allowed and welcomed on ao3, because they are exactly the reasons why ao3 was created in the first place.

ao3 was created because its creators got tired of censorship, they got tired of dark and taboo fics getting banned on pro-censorship platforms, and they wanted a place that was safe for ALL FICS THAT WERE DARK AND TABOO.

ao3’s main principle is being against censorship and being proship / profic.

there are some things in fiction that make me uncomfortable, but instead of shaming people who are just minding their own business and not harming anyone in real life, I choose to curate my own internet experience by blocking/muting what I don’t want to see. ao3 has excellent tagging system, so instead of being a bitch, use their tagging system properly and you won’t see the things you don’t want to see.

it’s your job to curate what you see. it’s not other people’s jobs or responsibilities to censor themselves for your personal comfort. the world does not revolve around you.

also you cannot censor “only the things you personally hate” without expecting everything else, that isn’t of conservative beliefs, to be censored too. because censorship is a slippery slope and a fascist tool. I promise you there are people who think “why do tags for queer love even exist on ao3? they’re grooming children”.

if you allow the things that you hate to be censored — because someone with enough power gets to control what other people can and cannot create/consume, it will not stop at the things that you hate.

Two limits! You can have a maximum of 4,096,000 characters in 1 [one] tumblr post. I would work out how many combinations this is, but 26^6,000 is already considered to be "Infinity" by most calculators, and a program I wrote threw an error code.

26^95,000 is already over 134,000 characters long - which would take 33 different text blocks to convey via tumblr. Whenever somebody says we're running out of posts, don't forget that tumblr is needlessly designed for MASSIVE amounts of information [no matter how detrimental it may be for mobile phones].

Though I'm not certain if that character count includes spaces, unfortunately, as I got that figure by googling "how many letters are in lord of the rings" and came upon a TikTok that counted the number of letter characters in LotR in order to figure out how many Spaghettios cans would be needed to re-write the entire thing, if one were to cut and paste each individual letter from the cans blackmail-letter style.

For those curious, the numbers are 2,261,081 letters in LotR, which calculates out to 8,795 cans of Spaghettios needed, which would cost about $12,225.

So, those are the theoretical limits, yeah. But in practice, the editor is not even close to be optimized to handle these kinds of huge posts: there is a point, far far away from the size of the lord of the rings, that your browser would just crash.

So if you are planning to post long fanfiction, or anything, you better work on something that's optimized for long form (locally, or some alternative to Google docs) and then post in chunks.

For the first post (first book only) there was no issue. Edit loaded, I could add the tag, press Save and done. Note: this was on a PC using Chrome

For the second post (all three books) it worked the same way as the API. I could load up the page and edit it, but when pressing save I'd get an error message. However the edit did go through actually and the tag is now there.

Haven't tried to edit the text itself, but don't think it would make a difference - if your browser can load it'd just work.

(On mobile on a low/mid-range Android the App could open up the post without problems, but haven't tried editing. Using a mobile browser - Firefox in my case - the page could load but it was super slow, and buggy, and it also did crash once)

Also as a public service announcement: if you want to filter the post out from your dash the tag is #full-lotr-test

Actually, it's the polar opposite. We are throwing crazy shit at Tumblr here, and it's taking it as a champ. I can't understand how people are seeing this post and going "ooh, crappy Tumblr am I right?", when we haven't manage to break anything even while trying to go up to the theoretical limits.

Let me tell you: Tumblr codebase was the most tidy and well kept I've had worked with. While I was there, the front end had a 100% test coverage rule, which is bananas. And the backend is a marvel of optimization that works so well that's scary.

Does Tumblr have hiccups from time to time? Of course. Are there parts of the code that are so optimized the only a few folks actually understand what's going on there? Yup.

Hedgehog-shaped jar, Neolithic period (3500-3000 BCE)

-------------------------------------------------

Also prev tags:

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.