Yeah, it reads more or less exactly like what you'd expect to see if you asked claude to run / investigate the same benchmark, then sat there and read thru its entire chain of reasoning. Beyond making it difficult to trust the veracity of the content, that AI level of verbosity is just a very inefficient way to communicate.

As a developer, writing a 500 line function to express some behavior that could have been expressed in 50 lines is generally a bad look. As a reviewer it makes you wonder whether the author is not competent enough to notice that there's a simpler solution, don't care enough to go back and simplify before publishing, or some combination of both.

It's the same concept with prose. Publishing 50 paragraphs of meandering fluff describing 5 or so paragraphs of actually salient information just shifts the burden of parsing that salient info from the author to all of the readers.

>The same Java code, the same JDK, the same machine

The article has 25 "the same"s. Anthropic has probably spent quite a bit of its R&D budget on figuring out how to watermark Claude's output, but all you really need to do is count the number of times "the same" and "different" appear per 1,000 words :)

I'm genuinely curious: I read through the article and the narrative felt deliberate and didn't trigger my AI-radar. What made you think it was AI written?

It’s rife with “and that’s the important thing. $thing”-type constructions. Human writing uses those much more sparingly and doesn’t separate them into two sentences as often. It has a lot of “subject change: teaser” constructions, too. And overuse of bold and restatement.

Could it be human? Sure. But it doesn’t seem likely to me.

It didn’t read AI generated to me. I also love the irony of then using AI to generate a summary

> This article is the story of chasing that number down to a single machine instruction, and then finding out that the instruction was only half of the answer.

> So the difference has to be in what the JIT generated, and the profiler gives us exactly that.

> That is the whole vocabulary. Let’s read some code.

> Decoding the x86 version instruction by instruction is out of scope here.

> What is not architecture specific is the logic.

Here’s a segment flagged by Pangram: https://www.pangram.com/history/87e25169-30a4-4030-a65a-dba8...

I don't know if it's AI or not, but if it is, it isn't egregious, and certainly not "garbage" or disrespectful to the reader.

I will admit I shouldn’t have said “garbage”, because it is too emotional.

But I absolutely think it is bad for the reader, and deserves to be called out. Authors should know that it’s not good enough.

I don't think the Pangram result is shareable to others.

"If you are sure the link is correct, please make sure you are logged in with the account associated with this result."

Even if that were AI (and I think there are better examples in the article to indicate that it is), why do you care if you just used AI anyway to summarize it anyway?

I don’t really care that it’s AI, the problem is that it’s bad AI writing.

The summary I shared is much more straightforward. It’s mediocre, but just barely good enough to extract the message without making me super annoyed.

If you don't like AI writing, why are you on hacker news? Most of the articles posted are written by AI.

Some of us have been here longer than AI. "To be intellectually stimulated" would be the why, which there used to be more of prior to the frontpage becoming riddled with AI slop. "HN is for conversation between humans" as the guidelines say, and though the guidelines imply that's towards comments, I'd prefer it to be towards submissions too, for the same reasons it was added for comments.

I'm already at the point of "the reward/effort of HN is getting pretty low", but the question is, where to leave for?