This tree has an unintended property: it remembers where you looked, and it cannot unlook.
Nothing in the code was designed to produce this. It emerges from three features of the CRR architecture interacting:
The result: wherever you direct your attention first becomes the region of deepest growth. When you look away and return, the fine structure has been garbage-collected — but the fork-points remember they already fired. That branch of possibility is spent.
This is precisely how morphogenesis works. A stem cell accumulates developmental coherence, commits to a lineage at rupture (differentiation), and cannot easily reverse the decision. The cell's children may die, but the parent's commitment is permanent. Waddington's epigenetic landscape — the ball rolling down valleys of decreasing potentiality — is exactly C → δ → R with irreversible branching.
Root tips do the same thing underground. The finest rootlets are constantly being shed and regrown, but the main root architecture, the decision points, are fixed. Explore, commit, prune, regrow. The tree you see here is doing real tree things for real tree reasons.
In neuroscience, attention is typically modelled as gain modulation — you increase the signal-to-noise ratio wherever you're looking. In CRR terms, attention is literally the driving function L(x,τ). Attending to something provides the coherence input that pushes a percept toward its rupture threshold.
This has a strange consequence: the order in which you attend to things constrains what you can perceive later. The tree makes this visible. Your first area of focus gets the deepest structure. Later areas get what's left. This matches the psychological finding that attentional set — what you look at first — shapes what you can subsequently notice.
Under the FEP, organisms minimise surprisal by building generative models of their environment. In CRR's language, a generative model is a coherence structure — a pattern of accumulated C that predicts the next rupture. The tree's branch architecture is its generative model of "where growth is possible."
When you zoom in (reduce Ω, increase precision), you're performing active inference — sampling the world at higher resolution to reduce uncertainty. But each observation commits the system. The surprisal is resolved, the coherence is spent, and the structure is fixed. You can't unsee it.
Friston's FEP says organisms act to confirm their predictions. Here, the tree grows to match where you're looking — and then can only grow more in places that are consistent with what it's already built. The model constrains the modelling. That's the FEP in forty lines of JavaScript.
Whitehead argued that reality is composed of actual occasions of experience — momentary events of prehension (grasping the past) that perish the instant they reach satisfaction. Each occasion is a C → δ → R cycle: the past is prehended (coherence), satisfaction is reached (rupture), and the occasion contributes its data to future occasions (regeneration).
The critical point: actual occasions are irrevocable. Once an occasion has achieved its satisfaction, it becomes an objective datum — it can be felt by future occasions but cannot itself change. This is exactly the irreversible rupture in the tree. A forked branch-point is an objective datum: permanently decided, available to its children as initial data, but closed to revision.
The observer-dependence is equally Whiteheadian. For Whitehead, there is no experience without a subject. The tree doesn't grow "in itself" — it grows where it is experienced. Your act of observation isn't separate from the tree's becoming. It is the occasion's subjective form — the manner in which coherence is accumulated.
What you're watching is not a simulation of a tree. It is a process — a real pattern of coherence, rupture, and regeneration — that happens to look like a tree because trees are what CRR does when you point it at a branching geometry.
§CRR · All geometry derived from Ω = 1/π, C* = π, σ = 1/2, CV = 1/(2π). No free parameters. The emergent observer-dependence was not designed. It follows from the irreversibility of δ(now) and finite computational resources. Alexander Sabine · temporalgrammar.ai · cohere.org.uk