Hacker News Reader

Almost hilarious

7 patterns, none of them at all correct

The first is the funniest. Three consecutive uppercase letters. Lmao. Yep! Got that one right.

I don't know why this was hidden. I find it interesting.

You find a list of patterns in a string, except none of those patterns are actually in the string, interesting?

Or you find "LLM can't handle characters" interesting?

Those both seem extremely boring (and off topic) to me.

The most interesting thing is that a bunch of apparently smart people keep treating LLMs as if they were truly capable of something resembling problem-solving or logical thought, or even "understanding" a question.

It should surprise no one that they're bad at string processing, and yet.

Why did this seem worth trying, to you?

It's a great way to verify the machine knows nothing.

We already know it knows nothing – but this doesn't verify it. I could produce a similar response, even though I know you can't crack Enigma in O(n) time.

Well, not unless you ignore the 158962555217826360000 constant factor. (Ref: https://the-axiom.uk/how-long-would-it-take-you-to-crack-the...)