Early in development, a sharp reordering of information-response directions tends to constrain the future. Later, the same kind of turn fans the future outward.
The experiment
At ten developmental checkpoints we already knew how each organism responded to the same eight microscopic actions. For each of 64 organisms, we found its sharpest early reordering of the action landscape—then ran all eight actions to the end from that peak and its two neighboring ages.
8 scoresΦ-R + whole prediction for the same 8 actions
→
largest rank rotationthe organism’s own causal-compass turn
→
8 futuresmean pairwise final-field distance
organic organisms64disjoint intervention panel
future trajectories1,536all prefixes mechanically identical through action time
overall peak advantage+0.002395% -0.0028 to +0.0073
The average hid a developmental transition
The compass peak was not uniformly expansive. At q24 it leaned constrictive. By q40 and q48 it leaned expansive. The causal meaning of the same geometric event changed with development.
early peaks · q24/q32-0.003130 organisms
late peaks · q40/q48+0.007095% +0.0014 to +0.0130
late minus early+0.010195% +0.0004 to +0.0204
The all-organism prediction was fixed before these futures existed. The age split was found after opening them, so this is the new lead—not yet its independent repeat.
It is a population shift, not one showcase
Every dot is one organism’s future breadth at its compass peak minus the mean of its neighboring ages. The distribution moves upward across the developmental sequence even though individual forms remain heterogeneous.
What “constrictive” and “generative” look like
q24 compass peak
radiant-blob-0658
-0.0661
a1
a2
a3
a4
a5
a6
a7
a8
Eight inventory-preserving actions, eight final states. Number = peak repertoire minus neighboring ages.
q40 compass peak
harmonic-entity-0167
+0.0512
a1
a2
a3
a4
a5
a6
a7
a8
Eight inventory-preserving actions, eight final states. Number = peak repertoire minus neighboring ages.
These endpoint grids are illustrations, not the estimator. The measurement uses all 28 pairwise distances among the eight futures for every organism.
The emerging picture
Information geometry acquires developmental semantics.
Before the organism consolidates, a compass turn tends to canalize: action directions reorganize without opening many additional futures. Later, a compass turn becomes generative: the same class of reorganization identifies a state from which small conserved actions reach more distinct forms.
That is a plausible causal-emergence transition: not merely “more information,” but a change in what information-bearing organization allows the system to do.
Next direct swing
Freeze a q40 classifier on a new organic cohort. Before opening any long futures, label organisms whose q40 compass rotation is locally maximal. Then run the same eight actions at q40 for peak and non-peak organisms. The clean question: does the late compass peak identify broader future maps at the same age?