Hacker News Reader

To me that makes it worse. Why would two people get wildly different answers to a simple factual observation query.

Because of the interplay of how tokenizers work, temperature, and adaptive reasoning? These models aren't fact generators.

~stochasticity~