Got deepseek stuck on an infinite loop lol

Asked it which natural number has the letter “k” in it without thinking mode. After some initial overthinking and finally reaching the conclusion that maybe no English natural number has a “k” in it, it counters with “That can’t be” and seems to gets stuck onwards in an essentially infinite loop lol. That “That can’t be” thing is kinda interesting, I guess that’s the kind of behaviour that it might have been incentivized to do, which could allow it to think more deeply and try to find less evident solutions to problems, but could also explain a part of its high hallucination rate, if it can’t seem to accept that no can be an answer? Also a slight tendency to brute force finding a solution I guess, which could explain the ample token consumption and very lengthy and verbose thinking newish models tend to do… idk Turning thinking on still yielded immense overthinking but at least it didn’t get stuck, and I don’t know if it’s something that can be directly reproduced but it’s a funny failure case nonetheless lol