Hacker News Reader

Comment by username135 | Hacker News Reader

Was not expecting to read the whole thing. Very interesting.

Was it really interesting? To me, it has certain hallmarks of an AI-generated article. In particular, it introduces the same concept several times, in different sections. For example, fare classes, nested booking, and the SABRE system each get two different introductions.

The content seems legitimate, but I felt like my time was being wasted through at minimum a lack of editing.

Yeah, there was a lot of repetitive information which made me lose interest

Should have passed it through AI for a summary!

> To me, it has certain hallmarks of an AI-generated article.

I wondered that too.

I don't want to offend anyone, and have no idea how it was written, and I already know most of this stuff so am not the audience. But respectfully I feel like it had a lot of words for a fairly shallow overview, which feels AI-ish, plus the "delve" at the beginning got my radar up. This is sort of what I expect from Manus or one of those ersatz "research" LLMs. Anyway, it's got lots of upvotes, hopefully people are finding it useful.

(Edit to add: it's actually content marketing for some kind of [questionable, subscribe to access some hidden refund thing] travel company so I don't feel bad criticizing anymore)

What's funny about it is that even though it follows the instructions your writing instructor gave you to always write a conclusion for your articles (and that probably are the instructions in the prompt for an LLM) it does not follow the instructions that I got from everybody I ever did content marketing with: to always end an article with a "call to action", which is why this article was up for hours before anybody noticed the site it was on.

Article can be skipped here.

"Deep-dive" is the call-to-action.

"Airlines don't just sell seats - they manage a dynamic inventory of fares, divided into booking classes (fare buckets)"

that "They don't just _____ -- they ________" construction! It's definitely a "once you see it" thing that you start to see constantly in AI-generated content! I wonder why the model loves that so much

Training on content with parallelism

"delves"

dashes

an explicit "conclusion" section at the end

Humans do not write conclusions? As someone who went to college, that is a natural way to end a long essay. True mark of higher education would be writing the conclusion at the top.

It's called an executive summary in that case, ain't it?

I was thinking an academic Abstract, but sure.

Exactly, it's a natural way to write a college essay. I've never not cringed reading an article/blog post that is structured that way, it comes across very contrived. I've also noticed that LLMs tend to prefer it, and humans tend to avoid it in general.

For a 500 word blog post, sure it may be a bit much. This article was a decently hefty read, for which the summary callouts are not out of place.

None of which are hard AI tells.

> dashes

I am forever angry that LLMs have ruined em-dashes. They’re a wonderful part of punctuation.

I think this is an example of above average but not great AI writing. I still read it to the end because the subject is interesting and there is enough focus (and, seemingly) expertise on the topic.

I think the telltale for me that makes me count as heavily AI-assisted is the lack of inclusion of real, inline examples of actual fares & their restrictions. I know I've seen them broken down before in other content. But not once here was there a full readout of an actual fare bucket & its rules. I think a human writer would have been tempted to include even one of those as an artifact, but an AI as a topic reviewer/summarizer/collator won't unless explicitly instructed.

Reply on news.ycombinator.com