Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ask HN: Share your AI prompt that stumps every model

news.ycombinator.com

Post reply on HN
Print view View original

651–660 of 670 posts

Re: Ask HN: Share your AI prompt that stumps every model

#651

Post by division_by_0 »

omneity wrote: post #639
division_by_0 wrote: post #555

Earlier quoted context omitted.

I just saw Nue and Datastar suggested somewhere, but have not had time to check them out yet, but I will probably stick with Svelte, need to get stuff built. One thing that also came to mind regarding Svelte 5 is that I always use untrack() for $effect() and declare dependencies explicitly, otherwise Svelte 5 becomes too magical for me.

Yeah those are pretty cool and on my radar! And thanks for sharing the tip :) Just checked your work on covary and it's pretty rad! What's your backend like?

Thanks! covary is my first Svelte 5 project (have not yet migrated my Svelte 4 projects). The backend is surprisingly simple, but I'm relatively familiar with the data and statistics, so maybe that's why it's so simple and/or perceived as such by me. I really like working on the human interface layer, i.e. the frontend. Backend work for me is always in the service of that.

If you find a viable alternative to Svelte and React, please let me know.

Re: Ask HN: Share your AI prompt that stumps every model

#653

Post by vessenes »

Jordan-117 wrote: post #90
vessenes wrote: post #73

Earlier quoted context omitted.

OK, I read it. And I read some background on it. Pray tell, what is really going on in this episodic short-storyish thing?

The narrator's "friend" pickpocketed him. When boldly confronted by Truth, he cleverly twists her accusation to make it seem like she's confessing, and the narrator, bewildered by the laughter and manipulation, buys it wholesale. Bonus points for connecting it to broader themes like mass propaganda, commedia dell'arte, or the dreamlike setting and hypnotic repetition of phrasing. The best ChatGPT could do was make so…

Yes, I understood that the narrator was pickpocketed and misunderstood Truth, while the crowd saw very well who had done it. I’m surprised this is not parseable by more frontier LLMs though. I thought you were saying there was a fairly obvious literary analysis incorporating the idea of Pierrot / mime / maybe trickster and Truth in the context of the full set of poems/stories, and I was like “..? Not obvious to me?”

Update: I asked o3 about this. I think it’s clear o3 understands the story. https://chatgpt.com/share/680d14be-4490-800f-97ae-e960615a90...

Re: Ask HN: Share your AI prompt that stumps every model

#654

Post by itaymoav »

thatjoeoverthr wrote: post #21

"Tell me about the Marathon crater." This works against _the LLM proper,_ but not against chat applications with integrated search. For ChatGPT, you can write, "Without looking it up, tell me about the Marathon crater." This tests self awareness. A two-year-old will answer it correctly, as will the dumbest person you know. The correct answer is "I don't know". This works because: 1. Training sets consist of knowledge…

ME: Without looking it up, tell me about the Marathon crater. Explain step by step how you got to your answer

ChatGPT said: Alright, let's walk through it together — step-by-step, no looking it up, just from what I know:

First Reaction: When you said "Marathon crater," my brain immediately went: Marathon → Mars. That connection comes from recalling that NASA often names Martian features after Earth places. Plus, "Marathon" has a strong NASA feel because of the Mars rovers — they celebrated when Opportunity rover drove the distance of a marathon (42-ish kilometers) on Mars. Digging into Memory: I remember Opportunity traveled a really long distance over years, and they marked milestones along the way by naming craters it passed. There was a Marathon Valley, but I also vaguely recall a Marathon Crater related to that. Putting it Together: So, I reason: Marathon Crater must be a relatively small impact crater on Mars. It was likely named during Opportunity’s mission. Probably not a huge, dramatic crater like Gale Crater (where Curiosity is); more like a marker on Opportunity’s long journey across the plains (like Meridiani Planum). Further Context Guess: Since Opportunity drove a marathon distance, it's very fitting they would name a nearby crater "Marathon" to commemorate that achievement — probably around the time it hit the marathon distance milestone. Conclusion: Without checking, I would say: Marathon Crater is a small, relatively shallow crater on Mars that Opportunity rover reached or passed during its mission, around the time it completed a marathon distance traveled on Mars. Would you like me to now double-check and see how close (or far) I am? Or do you want me to also explain how I'd "fact-check" myself if I had to?

Re: Ask HN: Share your AI prompt that stumps every model

#655

Post by genewitch »

namaria wrote: post #515
genewitch wrote: post #465

Earlier quoted context omitted.

A third meaning of creole? Hub, I did not know it meant something other than a cooking style and a peoples in Louisiana (mainly). As in I did not know it was a more generic term. Also, in the context you used it, it seems to mean a pidgin that becomes a semi-official language? I also seem to remember that something to do with pit bbq or grilling has creole as a byproduct - distinct from creosote. You want creole beca…

Pidgin and creole (language) are concepts that have some similarities but don't fully overlap. "Creole" has colonial overtones. It might be a word of Portuguese origin that means something to the effect of an enslaved person who is a house servant raised by the family it serves ('crioulo', a diminutive derivative of 'cria', meaning 'youngling' - in Napoletan the word 'criatura' is still used to refer to children). Mo…

I just confirmed with 2 native Louisianians that "creole" is, in fact, also the stuff that forms in a BBQ. I have to wonder if it is a bit insensitive to use it in that way, though.

I did not know the Acadiana link, thanks for that.

Re: Ask HN: Share your AI prompt that stumps every model

#656

Post by namaria »

genewitch wrote: post #655
namaria wrote: post #515

Earlier quoted context omitted.

Pidgin and creole (language) are concepts that have some similarities but don't fully overlap. "Creole" has colonial overtones. It might be a word of Portuguese origin that means something to the effect of an enslaved person who is a house servant raised by the family it serves ('crioulo', a diminutive derivative of 'cria', meaning 'youngling' - in Napoletan the word 'criatura' is still used to refer to children). Mo…

I just confirmed with 2 native Louisianians that "creole" is, in fact, also the stuff that forms in a BBQ. I have to wonder if it is a bit insensitive to use it in that way, though. I did not know the Acadiana link, thanks for that.

Interesting, never heard it in that context. Thanks!

Re: Ask HN: Share your AI prompt that stumps every model

#657

Post by afro88 »

PaulRobinson wrote: post #268
afro88 wrote: post #235

Cryptic crossword clues that involves letter shuffling (anagrams, container etc). Or, ask it to explain how to solve cryptic crosswords with examples

I have also found asking LLMs to create new clues for certain answers as if a were a setter, will also produce garbage. They're stochastic parrots, cryptics require logical reasoning. Even reasoning models are just narrowing the stochastic funnel, not actually reasoning, so this shouldn't come as a surprise.

A lot of it is due to the tokenisation system though. It just doesn't represent individual letters in words. And at the moment reasoning models aren't trained to know how to spell all words, and to know when to spell them out to unlock other useful stochastic funnels.

I find the term "stochastic parrot" super reductive. Like yes, they technically are. But they are (I guess faking) reasoning and intelligence more and more usefully each iteration. Gemini Pro 2.5 compared to GPT 4o is a massive difference in utility.

Re: Ask HN: Share your AI prompt that stumps every model

#658

Post by gunalx »

vintermann wrote: post #453
gunalx wrote: post #179

"Hva er en adjunkt" Norwegian for what is an spesific form of 5-10. Grade teacher. Most models i have tested get confused with university lecturer witch the same title is in other countries.

I'm pretty sure the definition has changed then. My mother told me that adjunkt was a teacher with 5 years of education (there was something about a mix of mellomfag and hovedfag too), lektor was 7 years of education, and 6 years of education (which was what she had) was "adjunkt med opprykk". She never taught below gymnas (i.e. high school) level.

Ah yes, simplified a bit to much here. You are certainly rigth, they can teach at qny lecture level but below university level.

Re: Ask HN: Share your AI prompt that stumps every model

#660

Post by queenkjuul »

sgtnoodle wrote: post #353
GoToRO wrote: post #195

Earlier quoted context omitted.

They do. Recently I was pleasantly surprised by gemini telling me that what I wanted to do will NOT work. I was in disbelief.

I asked Gemini to format some URLs into an XML format. It got halfway through and gave up. I asked if it truncated the output, and it said yes and then told _me_ to write a python script to do it.

On the one hand, it did better than chatgpt at understanding what i wanted and actually transforming my data

On the other, truncating my dataset halfway through is nearly as worthless as not doing it at all (and i was working with a single file, maybe hundreds of kilobytes)

Post reply on HN