Simulations within simulations · where the stack stops
Everything a simulated computer does is computed by the machine running its world. So a nested simulation can use at most a share of its parent’s computing, and never more than its world’s matter could compute at Lloyd’s bound. Every layer has less computing than the world that runs it.
If the parent world is resolved atom by atom, nesting adds nothing to the base reality’s total. A computer built inside it is ordinary matter, and that’s already counted. If the parent is coarser, nothing inside it can compute. Then the operator has to run the inner simulation separately, on its own computing. Other dials stay at the Sandbox baseline.
At 100%, a world does nothing but run the next one down. Our own civilisation uses a tiny fraction of its matter for computing.
- Base realityPhysics like ours: 1% of every kilogram in the observable universe at Lloyd’s bound, 8.1 × 10101 FLOPS.
- Layer 1needs 1.3 × 1068 FLOPS; the base reality allows 8.1 × 10101 FLOPS. Runs.
- Layer 2needs 6.0 × 1041 FLOPS; its parent allows 1.3 × 1066 FLOPS (its share of the parent’s computing). Runs.
- Layer 3needs 1021 FLOPS; the base reality allows 8.1 × 10101 FLOPS. Runs. Its parent is coarser than atoms and can’t contain a computer, so the base reality runs this layer directly.
The stack runs 3 layers deep. The base reality has to supply 1.3 × 1068 FLOPS, for 2 layers computed separately.
A layer resolved at atom detail or finer can contain working computers. These numbers come from the Sandbox engine, with everyone alive computed as a functional mind and time running from 1900. The share is a wild guess. The ceilings are textbook physics.