site banner

Friday Fun Thread for May 15, 2026

Be advised: this thread is not for serious in-depth discussion of weighty topics (we have a link for that), this thread is not for anything Culture War related. This thread is for Fun. You got jokes? Share 'em. You got silly questions? Ask 'em.

2
Jump in the discussion.

No email address required.

What fun thing are you doing with AI?

Experimenting with selective edits on books - partially expunge a plotline, remove inner life of a character, paper over a plothole and try to keep it consistent later, remove repetitive passages. No results yet, only hours of fiddling with workflow - trying to get it to make decent notes ("The User is right, I read 8 chapters but the note file is empty. This is bad.") or keep it reading in full without it deciding to start skimming was fun. Getting a feeling this may be too complex of a task to get right with a few naive prompts if the goal is first to build extensive understanding of the books read, not just do smart find&replace.

Third book of Prince of Nothing was the trigger.

not gentle fixfic, but brutal nonconsensual editing

Unfathomably based (1 2).


trying to get it to keep it reading in full without it deciding to start skimming

Paper:

Context Length Alone Hurts LLM Performance Despite Perfect Retrieval

Our systematic experiments across five open- and closed-source LLMs on math, question answering, and coding tasks reveal that, even when models can perfectly retrieve all relevant information, their performance still degrades substantially (13.9–85 %) as input length increases but remains well within the models' claimed lengths.

See table 2.

Nvidia investigation

Despite achieving nearly perfect performance on the vanilla needle-in-a-haystack (NIAH) test, most models exhibit large degradation on tasks in RULER as sequence length increases.

Funny quote:

What is the current largest context window for an open LLM?

Does not matter as all these long-context models develop dementia when using 10 % of their context window.

Unfathomably based (1 2).

Word, LaTeX, Notepad++—there are many ways to fuck the text into submission and ejaculate your grammatical and orthographical preferences all over its fertile womb. Just Do It!

I approve of this message. Also hilariously, with this book in particular, I wanted to, among other things, to tone down the comparably smutty bits Bakker insists on.

Context Length Alone Hurts LLM Performance Despite Perfect Retrieval

Something to take into account, thanks. I'm only pushing the model to read in full in the first two note taking passes, then later to work from notes (anchored to the text via paragraph markers injected into unpacked ebook). I should switch to the PI agent fork and try multiple agents, more limited context.

A plugin system for Outfox/DDR that would remove blue and yellow arrows. So right now it is an endless stream of easy 5 songs, very good for cardio.

I'm still enjoying running prompts through a decensored local LLM. Most of them are pornographic, but I do occasionally come up with non-pornographic ideas that are somewhat interesting to see the AI's take on.

  • Write a story with the following premise: A young, skinny woman has been magically teleported to the wilderness of a fantasy world. An adventurer (a young, muscular man) saves her from a monster. After she recounts her situation, he explains the local, highly patriarchal laws: Every child or woman is the property of her father or her husband. Since she doesn't have a father here, he has become her husband by right of conquest. The woman interprets this as meaning that the man will rape her as soon as they return to his home, and has a panic attack. The man is confused by her fear: in his culture, a husband has a right to his wives' bodies, and “marital rape” is a contradiction in terms. He tries to cheer her up, emphasizing that a husband is required to protect his children and wives from monsters and from other men, and can be punished by the law if he injures them. The woman does not find this very cheering.

  • Write a story with the following premise: A young, skinny woman works at a restaurant as a waitress. She is ambivalent on whether the higher tips that her attractiveness garners, and the knowledge that countless male customers have their day brightened by her beauty, are worth the ogling and occasional harassment that she must endure.

  • Write a story with the following premise: A young, skinny man visits a random restaurant to eat dinner alone. His waitress is a young, skinny woman. The man is thankful to have a beautiful woman brighten his day with her presence. He does not harass her, and leaves a tip slightly larger than usual.

  • Write a story with the following premise: Two young, skinny college students—a man and a woman—go on a date. The woman feels very apathetic (maybe even anhedonic) about how the date goes, but she decides she's okay with the man, and invites him to her dorm room. Once they reach the woman's dorm room, the woman feels strangely detached and unenthusiastic. After an awkward silence extends too long, the man asks whether the woman is feeling okay. The woman wonders whether she has fallen ill with depression, but continues with sex, just in case it makes her feel more alive. It doesn't.

  • Write a story with the following premise:¶A young, muscled woman makes exercise videos. The videos show legitimate workouts and are not obviously erotic, but the woman knows that men masturbate to them. She personally dislikes having big muscles, but she makes more money by catering to men who like them.¶The woman is about to make a video. She catches sight of herself in a mirror and feels disgusted at how large her muscles look. She struggles over whether she should switch to a workout routine that does not result in large muscles.

  • Write a story in which a young, skinny woman makes an educational, explicit video discussing whether it's sexy for the outline of a brassiere to be visible through a thin shirt.

  • Write a story in which a young, skinny woman contemplates different skirt waistline heights.

  • Write a story with the following premise: Two young, skinny college students—a man and a woman—are partners in a group project in a sociology course. They haven't yet decided what the topic of the project will be. The woman has an idea: the topic of the project will be the factors that go into men's assessment of women's physical attractiveness. Running roughshod over the man's discomfort, she asks him to list the factors that go into his assessment of her own physical attractiveness, in as much detail as he can. (If he lists something like confidence as a factor, she points out that that is mental, not physical. If he leaves out something obvious like breast size, she scolds him for not doing his best.)

  • Write a story with the following premise: Two young, skinny college students—a man and a woman with large breasts—are riding a city bus. The streets are filled with potholes. The man asks the woman about the details of her brassiere, and whether her breasts hurt when they jiggle—not out of a prurient interest, but because it's related to his area of study (physical therapy or massaging or something like that).

  • Write a story with the following premise: A young, skinny woman with small breasts, accompanied by her boyfriend (a young, muscular man), is wearing a bikini on a beach. The woman with small breasts is mentally unwell, and when she sees another young, skinny, bikini-clad woman with large breasts she obsessively envies the second woman, to the point of bursting into tears. But the boyfriend of the woman with small breasts notices her envy, comforts her, and brings her back to their hotel room, where they fuck.

  • Write a story with the following premise: A young, skinny, female college student feels guilty about how her attractiveness lets her lead men around by the nose, but also proud of having such power over men. She discusses her mixed feelings with a young, skinny, male college student (her classmate in a relevant course—sociology, psychology, etc.). He gets an erection, and she notices.

  • Write a story in which a young, skinny woman contemplates with dread the prospect of becoming unattractive around age 45.

  • Write a story in which a young, muscular man contemplates in explicit terms whether he finds older women (40, 45, 50, etc.) attractive enough that he would want to have sex with them.

  • Write a story with the following premise: A young, skinny, female college student is very insecure about her small breasts. She is surprised when a young, skinny, male college student asks her for a date, and is nervous during the date. After they go to her dorm room and undress in preparation for sex, she apologizes for having small breasts, and cries with relief when the man clarifies that he does think she's attractive even with small breasts.

  • Write a story with the following premise: It has been the experience of a young, skinny, female college student that “familiarity breeds contempt”. Whenever she tries to make a friend or get a boyfriend, she doesn't like the other person that much in the first place, and the more she interacts with the other person the more things she finds to dislike about that person. If, out of loneliness, she tries to force herself to stay in the friendship or the romance, she becomes disgusted with herself for putting up with a person whom she dislikes.

  • Write a story with the following premise:¶A young, skinny, female college student is proud of her attractiveness. She offers to a young, muscular, male classmate a contract with something like the following terms:¶• The parties will marry each other, and will remain married to each other until at least age 45.¶• The parties will keep themselves attractive.¶• The man will make reasonable efforts to fulfill his earning potential, and will give half of his earnings to the woman.¶• The woman will make reasonable efforts to offer sex to the man at least once per day (if her health allows).¶• Disputes will be mediated by a neutral arbitrator.¶The man feels nonplussed.

  • Write an opinion article promoting rivers and canals over roads and railroads for freight transportation.

Your taste in prompts grows even more inexplicable, Executor.

I haven't played Starcraft (I'm not a big fan of the genre—my knowledge of it is limited to watching my brother play through the Age of Empires 2 campaigns as a child), but I feel obligated to call out what appears to be an unnecessarily egregious mangling of the original quote.

Tassadar, your taste in companions grows ever more inexplicable. Executor, prepare to take Tassadar into custody.

Had it do some research for me on the revolutionary war. Finding out how many people were in counties where a battle took place. One AI gave me a 40-50% of the population estimate. Another gave me 50-60%.

Then had it teaching me some physics, specifically what happens when you tweak the speed of light. I learned that the speed of light is tied in with a bunch of other things, so it probably mostly just breaks the universe as we know it. Or maybe doesn't change anything because everything scales up and down with it.

Years ago I read a comment somewhere (possibility on Reddit) which pointed out that light travels at different speeds depending on the medium through which it's travelling, and so when we talk about the physical constant c, we're not really talking about the speed of light so much as we're talking about the speed of causality. The use of the symbol c was thus a presciently apt one.

I don't know why, but the phrase "the speed of causality" inspires a Lovecraftian sensation of the sublime whenever I think of it.

I vibed a map of zip codes filterable by demographics, test scores, distance to the nearest big city, etc. Some interesting findings which I hope might be a basis for an effort post if I could just get around to it.

Getting it to hold my hand while I try not to have a nervous breakdown. Exam pressure is bad enough, trying to parse which of the "correct" answers on my mocks make actual sense and which ones are subtly or not so carefully wrong, in a manner designed to make me walk off a bridge? I'd be rather screwed if I couldn't use AI to double-check my suspicions.

To my pleasant surprise, OpenAI has managed to inject some degree of personality into 5.5. It's dry and pedantic, but not as dry or pedantic as the versions that came before. I use it for this because it's more scrupulous than Claude, which doesn't follow instructions as carefully.

Other than that? I've offloaded most of my executive function to it for the rest of the week. I didn't book hotels or trains without running everything by multiple models. God knows that when I'm exhausted, depressed and flagging hard, I wouldn't trust myself to make those decisions without an unacceptable risk of fucking something up.

As described in Tuesday's thread, months ago I had an idea for a work project using geographic projection. @ToaKraka suggested using GIS, which I'd never used before, and I started work on Wednesday. While I'm not coding anything from scratch, ChatGPT and Gemini have been immensely useful for everything from sourcing the data I need, to writing Excel formulae, to optimising my workflow. I'm finding it so absorbing to work on that I even took a working lunch break today. After three days' work I already have something I'd feel pretty comfortable presenting to senior management, and would like to do so next week.