Hacker News Reader

Comment by trwhite | Hacker News Reader

Any proof?

Subsection titles, and how the text flows. When I read these kinds of LLM-Powered posts, I feel the same thing. Like remembering the taste of some canned food which was not great, but I had to consume to stay alive. It feels like junk food, not optimized to taste great, but doesn't reek of machine oil either. A price-optimized, slightly unhealthy substance which is somewhat compatible with my body because of food engineering magic.

These posts are not only a waste of time, but it also makes me feel deceived and betrayed. Why this is important? Because it's the honest take and bears the load of the reality of the post.

Please, don't make me read something you didn't write.

Disclosure: This post, incl. llm-sounding paragraph is typed by a human. All of the keystrokes are 100% human and made possibe by burning ATPs and firing tiny lightning bolts across some brain cells. No humans are harmed or GPU hours are burned in this comment.

Total water consumption: 50ml of cold spring water, because I'm thirsty.

I just don't see the "LLM-speak" here at all, for example. And I'm very tired of people who throw around the LLM-power critique as an easy criticism when there's no clear basis for it. It's lazy and stupid.

There are certain styles for a post like this that some people may not like but things like subheads and what would be, by historical standards somewhat clipped writing are just how writing in this content is mostly encouraged theese days. I'd just add that when people do post long-form New Yorker pieces there's plenty of whining about getting to the point.

I think this mostly boils down to experience with writing styles. I am an avid reader, albeit mostly read in Ukrainian/Russian, but I can confidently tell whether some paragraph was written by Platonov or Nabokov or Zhadan or Sorokin; even when the guys are stylizing their texts (Sorokin does this all the time) there are still subtle things in how text is composed that tell you who wrote it. If my reading menu was mostly contemporary blogs, probably I would be less sensitive to distinct writing styles.

The first tech book I did with a publisher, I was pretty much told that I needed to break things down more with subheads and the like. And I was already doing some writing of whitepapers and blogs with my then-employer which had a somewhat similar structured style. Certainly very different from Dickens, Trollope, Collins, etc.

I'm not making a value judgement, but modern writing on technical topics has generally been steered toward digestible formats whether or not LLMs are involved.

Until I left, I was on social media and style committees at my former employer and there was a lot of data-based evidence around keeping things (in various multimedia formats) shorter and easier to consume. In fact, we mostly got away from writing white papers at all in favor of shorter pieces.

I don't disagree with you that contemporary technical writing, editing and technical blogs were converging to what is known as llm style even before provisioning of llms. In the end, llm were trained on these recent writing and recent blogs and this is why similarity exists. Still one can find some technical writing that stands aside and has very distinct style without compromising on clarity/digestability, i.e. Martin Fowler, Armin Ronacher, Michell Hashimoto write very differently from each other and Claude, but their content is always sharp and very much digestable. Take this paragraph by Claude

> Scientific Linux gave the scientific community its own implementation of a RHEL-compatible computing environment. More importantly, it kept alive the people, processes, infrastructure, and institutional knowledge needed to maintain one.

> That capability had option value. You might not need to exercise the option this year, or even this decade. It becomes valuable when the assumptions underneath your primary platform change. Those assumptions changed surprisingly quickly.

Can't imagine this lengthy style with paraphrases coming from Hashimoto or Fowler. Yes, it can appear not only in llm writing, but also in some heavily edited o'reilly book, sure, but I miss the times when tech blogs and o'reilly books were different.

I might have avoided the option value phrasing. But both sentences seem pretty ordinary to me and something I would write. But agree to disagree.

Personally, I make at least some things I write more personal but, yes, that tends to be somewhat frowned upon in corporate blogs.

Why would a person use RHEL in the first paragraph, making the assumption that people already know what that means. Then a few paragraphs later, write the full name?

Theoretically a mistake a person could make, but it feels a bit unlikely in this context to me.

Also, the repeating pattern of Prompt:List through multiple paragraphs.

Headlines, that are sentences

Headline that goes "A was this" and next headline "Then b was that".

You obviously can't get proof. But there are enough stylistic tells to confidently say that AI was substantially involved in writing this. It's annoying to read. That's enough to dislike AI writing IMO.

"Scientific Linux was never simply "RHEL with a different wallpaper." It gave the scientific community an institutionally independent implementation of the Enterprise Linux platform."

Any proof that you're a human and not a sealion?

Reply on news.ycombinator.com