I tried it, and the llm gave me an absurd home lab scenario about servers shooting each other in the head to determine which was the "master server". So I told it that it was not an actual problem that it had, and sure enough it admitted it made it up. When you press an llm you will always find there is no internal state behind the thinking. It's just output.