logoalt Hacker News

Departed7405yesterday at 3:00 PM0 repliesview on HN

I tested Gemini 3 Flash (no visible reasoning trace). It gave me a choice matrix. Said that unless it was getting soap and a sponge, I should drive.

Kimi 2.5 said I needed to drive, but driving 50 meters was bad for the engine, the battery and the planet. it then recommended me to push the car, if safe.

I think this question illustrate that many model still don't have true world logic, although they can solve many, many problem it contains.

Also interestingly, the two models I tested didn't consider EVs.