Developmental tangent controller · fresh cohort

The future splits.
The transition holds.

We pushed 64 fresh Flow Lenia organisms forward and backward along the information-response direction that emerges during development. The action did not broadly move the birth-like transition. It exposed something stranger: microscopic futures separate while the developmental event stays canalized.

64fresh organisms
234pre-transition states
1,234900-step futures
8physical probe actions
01 · The direct test

The tangent is not a universal accelerator.

Forward versus backward changed the certified transition by +0.32 steps. The 95% interval runs from -0.47 to +1.19. Coarse adult morphology barely moved.

−1.6 steps+1.6 +0.32

The developmental coordinate reads the phase. One tiny action does not simply turn the clock.

02 · The surprise

Different microfutures. The same developmental event.

Across untouched, forward, backward, Φ-high, TDMI-high, and blind futures, the transition-time range had a median of only 3 steps. Yet the mean pairwise final-field distance was 0.309.

51.3%

of pre-transition states kept all six developmental events inside three steps.

0.344

mean forward/backward final-field Hellinger distance; centroid alignment leaves 0.342, so this is not simple translation.

The basin absorbs the direction at the level of timing while remembering it in the microscopic future.

03 · A new lead

Whole-over-parts may mark causal commitment.

At every fixed checkpoint, higher whole-over-parts information went with a narrower transition-time repertoire. After controlling checkpoint and rectangularity, the partial rank association is -0.193 with bootstrap interval [-0.366, -0.006].

q24q32q40q48 whole-over-partsrectangularity
q24q32q40q48 turnednot yet turned

Blue bars at left: whole-over-parts organization versus timing range, negative at q24, q32, q40, and q48. Red bars: rectangularity versus timing range, positive at every checkpoint. The information coordinate and visible morphology are not saying the same thing.

The selected policy panel contained between two and five distinct physical actions. After controlling that count too, the estimate remains -0.184, but its interval just touches zero [-0.355, +0.012]. That does not erase the lead; it tells us exactly how to remove the ambiguity.

04 · The actual creatures

Canalized bodies and rare open basins.

Most states are tightly timed. A few organisms remain radically redirectable. These are not selected “wins”; they are the strongest contrasts in the complete panel.

ethereal-form-1459 · q32

All six transition times fit inside 1 step, yet the final fields separate strongly.

untouched final field
untouched
tangent forward final field
tangent forward
tangent backward final field
tangent backward

radiant-form-1228 · q32

A rare wide basin: transition times spread across 231 steps.

untouched · 392 final field
untouched · 392
forward · 389 final field
forward · 389
blind · 620 final field
blind · 620

radiant-blob-1453 · q48

One tiny action family spans 98 transition steps inside the same organism.

TDMI-high · 199 final field
TDMI-high · 199
forward · 261 final field
forward · 261
backward · 297 final field
backward · 297
05 · Next direct swing

Carry the entire causal repertoire forward.

Does whole-over-parts organization prospectively predict contraction of the complete causal future repertoire before the visible transition?

All 8

Every physical action goes to adulthood for every fresh pre-transition state. No policy-selected action-count confound.

Same clock

Measure transition-time span and future-field span against ordinary morphology, then ask whether whole-over-parts predicts the contraction.

If the result repeats, causal emergence in Flow Lenia may look like this: microscopic possibilities remain broad while the macroscopic developmental event becomes hard to move.