A causal transplant into hidden state

The body stays.
Its future leans.

We copied no visible shape. We changed only how the same matter was divided between two hidden channels. The later body moved part-way toward the donor’s native future.

39organic Flow Lenia organisms
156reciprocal transplant futures
16/16families remained closer to host overall
0.0maximum starting visible-field residual
01 / What we changed

An invisible transplant into an identical visible body.

For every pixel, we held total matter exactly fixed. Then we replaced the host’s channel ratio with the ratio from its paired donor branch.

HOST BODY visible total matter at every pixel DONOR COMPOSITION channel 0 : channel 1 ratio HYBRID START same body · donor allocation channel 0 + channel 1 preserved bit-for-bit at all 16,384 pixels
same visible field + donor channel allocation → hybrid developmental state

The preservation check was exact Float32 equality, not a tolerance: all 156 hybrid starts had the same channel-0-plus-channel-1 field as their host at all 128×128 positions.

02 / The result

The hidden donor pulled the future without replacing the organism.

After 128 untouched steps, the hybrids had traveled 27.9% of the way along their paired donor branch’s native morphological direction.

27.9%mean donor-direction displacement95% CI [12.1%, 47.2%]
14/16families moved in the donor directionreciprocal, both histories pooled
-0.028donor-vs-host full-field advantagenegative = still closer to host · [-0.034, -0.022]
156/156starts with bit-identical visible fieldmaximum residual exactly 0.0

THE DONOR DIRECTION ACCUMULATED OVER THE UNTOUCHED FUTURE

-0.1+0.0+0.1+0.2+0.3+0.4+0.5 163248648096112128 steps after transplant fraction of native donor direction

The signal appears early, survives the middle of the trajectory, and grows strongest by step 128. This is not the donor image being pasted in: it is the hybrid’s own dynamics progressively expressing a bias carried by hidden composition.

03 / What moved—and what did not

A developmental vector moved. Whole-body identity did not.

COMPACT BODY + EXPANSIVE COMPOSITION

39.8% toward its actual donor future

95% CI [3.3%, 86.6%] · 9/16 positive families.

EXPANSIVE BODY + COMPACT COMPOSITION

16.1% toward its actual donor future

95% CI [-18.1%, 44.3%] · 12/16 positive families.

Host morphology dominatedThe full-field Hellinger comparison favored the host in every family.
+
Donor direction was importedThe future nevertheless shifted along the donor branch’s organism-specific trajectory.

The most faithful description is not “body-plan transfer.” It is a partial developmental vector: hidden composition changes where this body tends to go next.

04 / Why the simple label cancels

“Compact” and “expansive” were control directions—not permanent destinies.

UNIVERSAL TARGET AXIS

+0.050

Mean transfer at step 128, 95% CI [-0.258, +0.370]. The pooled label-level effect is weak because many native futures cross or rebound after release.

ACTUAL PAIRED FUTURE

27.9%

Mean movement along each organism’s own donor-versus-host future. That organism-specific direction is the one the transplant reliably carried.

At step 128 the native “expansive minus compact” separation itself averaged -0.226 [-0.670, +0.173]. So a global label can cancel even while the correct organism-specific developmental direction transfers.

05 / What this opens

A Flow Lenia organism is more than its visible matter field.

STATE

The same visible body can contain different internal channel arrangements.

DYNAMICS

Those hidden arrangements alter the direction of later autonomous morphology.

DEVELOPMENT

Part of an organism’s future is distributed through composition before it is obvious in shape.

This connects the information work back to organism emergence. The interesting precursor may not be a scalar alarm before “birth.” It may be an organized hidden state that already constrains which forms are reachable next.

06 / The next direct swing

Find out what the hidden composition is carrying.

The immediate experiment combines dose and spatial scrambling. It asks whether we found a scalar mixture, a spatial code, or both.