this post was submitted on 21 Feb 2026
533 points (98.7% liked)

Memes

54614 readers
589 users here now

Rules:

  1. Be civil and nice.
  2. Try not to excessively repost, as a rule of thumb, wait at least 2 months to do it if you have to.

founded 6 years ago
MODERATORS
 
you are viewing a single comment's thread
view the rest of the comments
[–] Axolotl_cpp@feddit.it 42 points 1 day ago* (last edited 1 day ago) (7 children)

Tested on GPT-5 mini and it's real tho?

Edit: Gemini gives different results

[–] Ephera@lemmy.ml 37 points 1 day ago (5 children)

Man, I really hate how much they waffle. The only valid response is "You have to drive, because you need your car at the car wash in order to wash it".

I don't need an explanation what kind of problem it is, nor a breakdown of the options. I don't need a bulletpoint list of arguments. I don't need pros and cons. And I definitely don't need a verdict.

[–] AlfalFaFail@lemmy.ml 21 points 1 day ago

I'll also accept sarcasm.

"Unless you've successfully trained your car to follow you like a loyal golden retriever, you're probably going to have to drive."

[–] Axolotl_cpp@feddit.it 8 points 1 day ago (1 children)

Yeaaah they waffle a lot, i hate that

[–] LordKitsuna@lemmy.world 1 points 4 hours ago

You can actually fix this in the settings there's an option for permanent prompt tunings and you can add things like "focus on concise answers" or my favorite " i don't need to be glazed , I don't need to be told that it's an insightful question or reaches the heart of the matter. Just focus on answering the question"

[–] TropicalDingdong@lemmy.world 4 points 1 day ago

It's the illusion of reason

[–] Jakeroxs@sh.itjust.works 4 points 1 day ago

I've found some success in ussing system prompts or similar to ignore explaining things lol

[–] Xavienth@lemmygrad.ml 3 points 1 day ago

They are trained to yap because it gives them a higher likelihood of giving the correct answer. If they don't go on and on in user presented text, it at least does it in hidden text.

[–] AlfalFaFail@lemmy.ml 11 points 1 day ago* (last edited 1 day ago) (2 children)

That first gemini response is legitimately hilarious.

[–] user224@lemmy.sdf.org 1 points 1 day ago

Yeah, Gemini is... interesting. It said it's a "one-way route" for the human.

[–] Axolotl_cpp@feddit.it 1 points 1 day ago

Yeah, he yapped a lot for no reason but at least it make you giggle a little

[–] Spezi@feddit.org 3 points 1 day ago* (last edited 1 day ago)

Tried it in GPT 5.2 (although in German) and it also says, that walking is better.

It also made a completely illogical sentence at the third point with two words in lowercase that should be uppercase.

[–] VincentComfy@lemmy.world 7 points 1 day ago

Gemini’s responses were surprisingly humorous

[–] xthexder@l.sw0.com 2 points 1 day ago

Bold of Gemini to imply any sort of liability for what it says. Google's lawyers really don't want that to be the case.

[–] thedeadwalking4242@lemmy.world 2 points 1 day ago* (last edited 1 day ago)

I used paid models which will be the only ones the LLM bros will care about. Even they kinda know not to glaze the free models. So not surprising

( I have to have the paid models for work, my lead developer is a LLM nut )

[–] troglodytis@lemmy.world 1 points 1 day ago* (last edited 1 day ago)

Gemini got jokes, but why does it think walking emits zero carbons? Humans are carbon emitters, more so with exercise. Hell, I farted while giggling at its humor.

Much less carbon than the car? Yep. Zero? Nope