undefined | Better HN

0 pointspests7mo ago0 comments

While I usually hate all the accusations of writings being LLM generated, I find your example a bit odd as that phasing is very typical of ChatGPT, especially when it was glazing everyone after that one update they had to reverent.

“It’s not just _________. It’s _________________.”

This was in almost every response doubling down on the users ideas and blowing things out of proportion. Stuff like…

“It’s not just a good idea. It’s a ground up rewriting of modern day physics.”

0 comments

rocky_raccoon7mo ago

I picked up on it very quickly as well. Here are some more phrases that match that same LLM pattern. Sure, you could argue that someone actually writes like this, but after a while, it becomes excessive.

- Your program continues running with a corrupted heap - a time bomb that will explode unpredictably later.

- You’re not just getting 64 bytes of memory. You’re entering into a complex contract with a specific allocator implementation.

- The Metadata Mismatch

- If it finds glibc’s metadata instead, the best case is an immediate crash. The worst case? Silent corruption that manifests as mysterious bugs hours later.

- Virtual Memory: The Grand Illusion

- CPU Cache Architecture: The Hidden Performance Layer

- Spoiler: it’s even messier than you might think.

zem7mo ago

huh, interesting, I guess I haven't read enough of it to pick up on the patterns

ziml777mo ago

Atomic Shrimp has an aside in a recent video about how to identify AI writing. It's worth a look https://youtu.be/VeD9dUUFl-E?t=668

He's not the only one to point out these things that LLMs (currently) tend to output, but this is one of the shorter overviews of the tells you can spot.

stevenhuang7mo ago

E.g. the not x but why slop leader board

j / k navigate · click thread line to collapse