The theory is built from a single starting point: there is an observer, and that observer notices a set of modest, empirically grounded facts about the universe. Everything else — space, time, particles, forces — is derived from those facts, not assumed.
We begin with four rules. These are not axioms in the abstract mathematical sense, but concrete, empirically grounded claims about the world we actually inhabit.
These four rules are the entire foundation. No additional assumptions are made. What follows is derived from them.
Let us run a thought experiment.
We build a computer and use it to simulate a human — call her Alice — based strictly on her digitised DNA profile and the laws of physics. The computer simulates Alice step by step. As the program runs, virtual Alice exists, experiencing her virtual universe in the same way real-world Alice experiences ours.
Now we optimise the software. We replace common calculations with precomputed lookup tables, so the computer no longer spends CPU cycles working out how a protein folds or how a neuron fires under specific conditions. It simply retrieves the answer from memory.
What happens?
We continue optimising. Every calculation, one by one, is replaced with precomputed data. At each step, the output is the same. The only thing changing is how quickly we produce it.
Taken to its mathematical limit, the software disappears entirely. What remains is a static dataset — a massive bitstring on a hard drive — containing the complete record of every state Alice will ever pass through. No processors are running. No electricity is flowing. No clock is ticking. The external runtime is exactly zero.
Here is the question: does Alice’s experience still exist?
If the answer is no, then somewhere in that optimisation process consciousness switched off. There is a specific ratio of code to data below which experience vanishes. But that threshold is a new physical constant — one that does not follow from any of our four rules. Introducing it contradicts Rule 2, which insists that the physics of biology is complete as stated. We are forced to the opposite conclusion.
Conclusion 1: Time is internal. The flow of time does not require an active engine to push it forward. Time is an internal ordering relation within a structure of information. What Alice experiences as the passage of time is simply the ordered sequence of states encoded in the dataset — traversed from the inside, not driven from the outside.
Conclusion 2: The observer is an equivalence class. The optimisation argument produces not one execution trace but a whole family of them — the original simulation, the partially optimised versions, the fully precomputed dataset, and real-world Alice herself. Each is a different description of the same observer. Alice is not any particular trace. She is the invariant that all of them share: the equivalence class [A] of every execution trace encoding the same internal experience sequence. Running the simulation again does not create a new Alice. It produces another description of the same one. Computers do not manufacture observers. They reveal coordinates of a structure that is already there.
The optimisation argument established that the process generating Alice does not matter. A second thought experiment establishes that the arrangement of her data does not matter either.
Imagine our computer running Alice’s simulation while simultaneously running a billion other simulations. To handle the load, it interleaves execution: one step of Alice’s code, then millions of steps from other programs, then back to Alice. From the outside, the raw binary stream containing Alice’s data is completely fragmented. Her information is separated by vast oceans of unrelated noise. It looks, to any external observer, like randomness.
Does Alice notice? Does her experience fragment or degrade?
It cannot. Her internal experience depends entirely on the relational order of her information — which state follows which — not on how densely her bits are packed in a physical medium. You could print her execution trace across a billion pieces of paper, scatter them across the galaxy, and the internal coherence of her experience would be unaffected.
Conclusion 3: Substrate independence. The medium is irrelevant. Neurons, silicon, magnetic domains, interleaved bits embedded in noise — these are all different physical arrangements of the same information. A conscious observer is not a property of any particular substrate. It is a property of a particular information structure, regardless of what carries it.
The wooden computer makes this vivid. Build a computer from wooden spheres and mechanical ports. Step it through the same DNA simulation. The result is identical to the electronic version. Just as 2 + 2 = 4 whether you are counting apples or bananas, the mathematics of an observer is invariant across physical substrates.
These three conclusions, taken together, force a precise answer to a question that is usually left vague: what is an observer?
An observer is not a body. Not a brain. Not a running process. Not a specific bitstring. Each of those is a description — one representative from a much larger equivalence class. The observer is the invariant that all valid descriptions share.
This matters because it dissolves a confusion that derails most discussions of simulation and consciousness. People ask: if I copy Alice’s data and run it on three different computers, have I created three Alices?
No. And physics already has a precise framework for understanding why.
When a physicist samples a quantum wavefunction ψ multiple times, each measurement produces a different outcome — a different bitstring, so to speak. But those multiple samples do not create multiple distinct particles. They are localised glimpses of the same underlying invariant ψ. The Born rule, |ψ|2, gives the probability of each glimpse. The particle is the equivalence class; the measurements are its representatives.
Alice is the same structure. Running her simulation three times produces three descriptions — three samplings — of the same observer. No new information is created. No new Alice comes into existence. The equivalence class was already there.
This analogy is not decorative. It will become load-bearing in later chapters, where we derive the Born rule itself from the same compression principle that produces observers. The structure here and the structure there are the same structure.
It is often assumed that a computer must actively execute code to bring a simulated Alice into existence. However, this simulated Alice is not created within our physical universe. We cannot measure a spatial distance between the computer and Alice, nor can we measure any temporal delay between a bit-flip in the computer and its corresponding effect on her virtual state. Therefore, one cannot argue that one created the other. The computer executing the simulation and the simulated agent are not causally linked; they are two distinct representations of the same underlying information.
If a computer running a simulation of a physical system and the physical system itself are two descriptions of the same information, then every property of the physical system must appear in the execution trace of its simulation. The geometric description and the computational description are representationally equivalent.
This equivalence is the foundation on which everything that follows is built. We may not fully understand the universe from the inside. But we understand computers. And if the two are the same thing viewed differently, then studying the computational description gives us direct access to the physics.
The universe that emerges from this foundation has three properties that we did not assume but were forced to accept: