Emergencewhy these subjects, and why the camera has none
Many of these subjects are the examples the word was invented for. A de Jong attractor takes fourteen chaotic iterations per point and settles into a shape nothing in the iteration announces. A Penrose tiling turns local matching rules into order that never repeats. Dielectric breakdown, drainage finding its channels, a bifurcation on its way to chaos: a few lines of arithmetic, and structure nobody put there.
Not all of them. The Hopf fibration is found but not emergent - it is a clean geometric fact with nothing iterating toward anything, and the same is true of the polytopes and the E8 projection. The atlas holds both kinds, and they photograph differently.
What the emergent ones have is structure at every scale at once. Lean in anywhere and there is more, not because detail was added but because that is what an unbounded rule looks like from the inside. A photograph of one never runs out - which is the property that makes them worth photographing at a hundred gigapixels rather than at a comfortable size.
The camera is built to be the opposite
An instrument for photographing emergence has none of its own. Every arithmetic operation the Eidograph performs is pinned: deposits accumulate in fixed point through integer atomics so the order threads arrive in cannot matter, expressions are bound so no compiler may reassociate them, and the handful of library functions that drivers are free to disagree about are computed from the exact subset instead. Most of the registry now returns one hash on every card the census has reached, across three driver stacks and five architectures. The plates that still split between vendors are counted and named on the census page, which reads from the instrument rather than from this sentence - and they are named rather than rounded away, because a claim of “bit-identical” that quietly excluded its exceptions would be the kind of sentence this practice exists not to write.
That reads like a contradiction and is the reverse. Chaos means sensitive dependence: fourteen iterations amplify a difference in the last bit into a visibly different point. So for these subjects there is no such thing as a small arithmetic disagreement between two machines. Two people whose cards round differently have not made the same photograph slightly differently - they have photographed different objects. Bit-identity is not fussiness about pixels. It is the only condition under which two strangers can be said to have photographed the same thing.
And one place it is allowed
The public works are emergent on purpose. Nobody schedules them. A donor claims one supertile, renders it, delivers it; other donors re-render it and vote; the region folds when quorum agrees. No coordinator, no queue, no plan - the picture assembles out of local decisions by people who never speak to each other. The first work to close this way finished all 576 of its supertiles with no recorded dissent, checked by four independent groups.
Three relationships with one idea, pointing three ways: the subjects have it, the instrument refuses it, and the commons that computes them runs on it.
Sources
Bedau, “Weak Emergence”, Philosophical Perspectives 11 (1997) - the nearest published statement of what the emergent plates do: a pattern that follows from the rule, but that can be got out of the rule only by running it. Bedau reads that as a fact about the derivations rather than about anybody watching - on his account a perfect calculator would still have to run it.
Chalmers, “Strong and Weak Emergence” (2006), in The Re-emergence of Emergence - the weak and strong senses of the word, worked through the gliders in a cellular automaton, which is this atlas's own case. His test is whether the behaviour could be deduced in principle, a looser bar than Bedau's.
Anderson, “More is Different”, Science 177(4047), 1972 - the ancestor of the three-levels reading above. Each level of complexity has its own fundamental questions, which is why a rigid instrument and an emergent subject are not a contradiction.
Wolfram, A New Kind of Science (2002), ch. 12 - computational irreducibility: for some systems there is no way to know what they do except to compute it. It rests on a conjecture rather than a theorem, and it is a claim about systems, so it fits a deep zoom and not the Hopf fibration.