Hold the physical dose equal
At each dose, forward and backward used the same number of founder-pair swaps. Only the information-defined direction differed.
Future-blind Flow Lenia dose response
The memory moves into identity.
A stronger information-directed push made the later founder mixture remember which founder acted first—even after another founder action arrived.
At the 20% dose, Φ-forward retained 29% more compositional action memory than Φ-backward.
The three-dose slope was +0.0221, with a 90% family-bootstrap interval of +0.00323 to +0.0436.
Same early state · opposite information directions
6%
Φ-forward local swap
Φ-backward matched swap
12%
Φ-forward local swap
Φ-backward matched swap
20%
Φ-forward local swap
Φ-backward matched swap
Φ-forward / high doseA stays legiblein founder identity
Φ-backward / high doseA is erased morefrom founder identity
2,432new future trajectories
19organic families
3matched doses
29,184decoded future states
01 / The question
Is Φ a direction—or a dial?
The first experiment suggested that a Φ-forward nudge made later spatial dynamics harder to overwrite. This experiment repeated the intervention at three strengths, under a new future, and asked whether causal memory changed with dose.
Change the dose
Each early state received weak, medium, or strong matched swaps: 6%, 12%, and 20%.
Read two memory channels
After actions A and B, we measured how much A remained visible in the spatial field and in the mixture of founder identities.
02 / The response curve
Identity memory rose with the information dose.
The specific medium-dose spatial result did not repeat in this unseen future. But the dose curve exposed a broader response: the forward–backward difference moved from negative to positive in both channels, and founder composition separated cleanly at high dose.
+29%
more founder-identity memory at high dose
At 20%, compositional first-action memory was 0.00678 after Φ-forward and 0.00526 after Φ-backward: a difference of +0.00152.
The early information direction changed how strongly the later ecology remembered who had acted first.
This is a mechanistic follow-up on the same 19 families under an unseen future—not an independent population replication. The frozen medium-dose field-memory endpoint was not reproduced: +0.00042 [−0.00497, +0.00495]. The positive result is the predeclared dose response in founder composition.
03 / What moved
Forward preserved. Backward erased.
The compositional separation did not come from indiscriminate disruption. As dose rose, Φ-forward kept first-action identity memory roughly intact. Φ-backward progressively removed it.
The simplest reading
The tangent controls retention versus erasure.
At high dose, the same second action B overwrote less founder-identity information about action A in the forward state than in the backward state.
Φ-forward0.00678high-dose memory
Φ-backward0.00526high-dose memory
The field channel tells a complementary story: its forward arm tended to strengthen with dose while its backward arm stayed flat, though that decomposition remained uncertain.
04 / Developmental time
The difference grew after the push.
If the effect were only a mechanical scar, it would be largest immediately. It was essentially absent at passage 1, emerged around passages 5–6, and remained dose-ordered late in the trajectory.
Passage 1Immediate mechanical trace
−0.00009
[−0.00057, +0.00028]
[−0.00057, +0.00028]
Passage 5Separation begins
+0.00125
[−0.00091, +0.00341]
[−0.00091, +0.00341]
Passage 6Identity memory separates
+0.00277
[+0.00055, +0.00509]
[+0.00055, +0.00509]
Passage 9Dose slope is clearly positive
+0.0311 / dose
[+0.00895, +0.0568]
[+0.00895, +0.0568]
Passage 12Late dose ordering remains
+0.0344 / dose
[+0.00208, +0.0732]
[+0.00208, +0.0732]
06 / The exact control
No information direction, no difference.
Four family-schedule states offered no directional leverage: the forward and backward selectors chose exactly the same local swap. Those paired futures collapsed pathwise at every dose.
maximum later difference0.0all three doses · all measured curves
Why that matters
The simulator was not quietly labeling one arm “forward” and producing a difference by itself. When the physical state was identical, the later fields and founder compositions were identical too.
Across all other states, forward and backward still used equal swap counts at each dose. The measured manipulation separated the information score most cleanly at 12% and 20%, before the unseen future was run.
Direction created the divergence. Labels did not.
07 / The direct next swing
Put the memory axis back before form.
The result is not yet “Φ makes an organism.” It gives us a sharper instrument: a dose that can preserve or erase causal history. The next experiment should use it before an organic body has formed and ask whether it moves the commitment transition itself.
Intervene across the commitment window
- Take matched Flow Lenia states before, around, and after the first geometry turn toward an organic body.
- Apply the strong Φ-forward and Φ-backward interventions, plus matched blind and untouched controls.
- Follow founder-identity memory, spatial memory, morphology, fission, identity loss, and the later repertoire of possible futures.
- Ask whether the information direction moves when development becomes difficult to overwrite.
The bold question
Is the birth of an organism the moment when a previously fluid ecology begins to preserve its own causal history?
If Φ-directed intervention can move that boundary—earlier or later—then we are no longer merely detecting an information signal. We are controlling a candidate mechanism of developmental commitment.
The present result establishes a dose-ordered change in founder-composition action memory on these 19 families. It does not establish causal emergence, autonomous replication, or a population-wide law.