Hacker News Reader

Comment by alwa | Hacker News Reader

I’m surprised at the length of this gag, drawing on what seemed to me to be a particularly robust Wikipedia entry. Is the point that the article strikes you as juvenile, or that you feel like it unfairly makes touristic spectacle of something more complex than the treatment here reflects?

And without intending offense—is this a work of LLM? I’m curious because of the earnestly, completeness, and the slight flavor of “how do you do, fellow kids” in lines like “Talk about a plot twist!”

View on HN

> And without intending offense—is this a work of LLM?

I suspect so, as ChatGPT generates a fairly similar output[1] from the prompt "put this into dialog form between two American 20 year olds: https://en.wikipedia.org/wiki/Bunkers_in_Albania". As an example (this was part of the first output it produced):

OP: "Jake: So get this - there was this leader named Enver Hoxha who was super paranoid about getting invaded. We're talking an average of almost 6 bunkers per square kilometer. They're literally everywhere - beaches, mountains, cities, even in hotel gardens!"

ChatGPT: "Jake: Okay, so get this: back in the Cold War, Albania’s leader, this guy Enver Hoxha, went totally paranoid. He thought everyone was gonna invade—like, NATO, the Soviets, literally everyone.

I retried the prompt twice - the first time[2], "so get this" was used again, and the second time[3] used the name "Jake" again. Hardly proof of LLM use, but it'd be a weird coincidence if not IMO.

[1] https://chatgpt.com/share/678dcdaa-bda4-8011-a3a9-a5d03e7c9d...

[2] https://chatgpt.com/share/678dcf87-869c-8011-8727-8e370bde6f...

[3] https://chatgpt.com/share/678dcfd5-f87c-8011-b1ff-dca4dfd5eb...

I started reading the wikipedia article and just couldn't get into it. It all read like blah blah blah to me but why was it on the HN homepage? There must be something interesting about it. I've used the prompt "put this into dialog form between two American 20 year olds" many times. The output I was able to read no problem and actually enjoyed it. Any sort of dry text this works wonders for my brain being able to comprehend it. But judging by the downvotes I'm guessing this is just me!

It was https://claude.ai pro plan.

I thought the article was entertaining because of how absurd and obviously bad the idea of constructing the bunkers was.

See, now that seems like a pretty cool use case, I’m glad to hear this technique is serving you well! I guess it’s cases like these—where I’d be inclined to do the reverse, taking “20-year-old dialog” and asking Claude to give it to me straight—where the promise of LLM personalization shines…

I suspect the downvotes have less to do with your good intentions here, and more a mix of confusion over the context and HN’s hostility toward LLM content (which tends to have a low effort-to-quantity ratio).

Personally, I probably would have reacted more positively if you’d introduced your original comment with this explanation (although I’d suggest just leaving a link to the LLM conversation rather than pasting the whole thing).

Thank you for engaging on it!

But yeah, to me the LLM version strips out the most entertaining part: the pictures of these crazy things! Who but an autocrat could look at those and think “good idea”…

Reply on news.ycombinator.com