Here is a problem physicists take seriously and wish they did not. In a large enough, long-lived enough universe, random fluctuations can assemble anything physically possible — including, very rarely, a fully formed brain that pops into existence with a complete set of memories and a coherent experience of a world that never happened. Such a thing is called a Boltzmann brain, after the physicist whose statistical mechanics implies it.

The trouble is the counting. Producing an entire orderly universe with billions of years of history is astronomically expensive in probability terms. Producing one brain with the false impression of that history is astronomically cheaper. So in certain cosmological models, the overwhelming majority of observers who think they remember a past are not embedded in a real one. They are fluctuations with installed memories.

This is not mysticism. It is a genuine puzzle in cosmology, and theories that predict too many Boltzmann brains are treated as broken for exactly that reason. The argument cuts at the root of trust in memory: if the cheap observer is the common observer, your confidence that your past occurred is unearned.

The simulation framing makes the same point with less drama. A system does not need to compute a full history to give an observer the experience of having one. It only needs to load the memory. The honest reading is that "I remember it, therefore it happened" is an assumption, not a proof — and one the universe is not obligated to honour.