Can evolving organisms reshape inherited geography?
Organisms can now excavate, carry and deposit conserved units of ground substrate, with no construction objective anywhere in the reward or perception model. Two earlier architectures failed for different reasons before a third produced a viable, currently live candidate that adds ecological succession on top. This is the honest, unfiltered record of that path, including the failures.
The question
Given the physical ability to move conserved ground material, with no reward or objective for doing so, can an evolving population produce persistent, measurable terrain change, and can it do so without collapsing the ecology it depends on?
Setup
- Ruleset
- phase-y-mutable-terrain-v1 (superseded), phase-z-voxel-substrate-v1 (superseded), phase-z-voxel-substrate-v2 (current live candidate).
- Seeds
- Multiple exploratory runs across Phase Y, Y2, Y4, Y5, Z and Z2, plus a six-seed, two-arm pre-registered viability gate for Z and for Z2.
- Duration
- From a few thousand ticks (both Phase Y candidates, at extinction) to ongoing on the current live candidate.
What was measured
- Whether organisms attempted and completed terrain-modifying actions at all, and how often.
- Population survival and cause of death, to separate a terrain effect from an unrelated ecological collapse.
- Net land area change and the fraction of the map physically reworked.
- Whether the simulation's authoritative energy invariant held throughout each run.
- Under Phase Z2 specifically: ecological cover recovery on disturbed cells against the pre-registered survival-parity gate.
Result
Two architectures that let organisms change terrain without adequately letting them perceive it produced no directional terrain signal and two independent starvation extinctions. A third architecture that closes the perception gap passed its pre-registered viability gate and is currently live, showing a small but real and persistent net increase in land area.
Limitations
- The Phase Y extinctions are confounded: a matched terrain-off control also had a weak founder population, so terrain's own causal contribution to those specific deaths cannot be cleanly isolated from founder viability.
- The Phase Z v1 energy-invariant halt is preserved as evidence of a real defect the simulation caught rather than silently absorbed, but it means that specific candidate's own long-run terrain trajectory was never observed to completion.
- The current 3% land-area figure is one measurement at one point on one ongoing run, not a converged or final outcome, and not the result of a pre-registered long-horizon test.
- None of this constitutes evidence of intentional construction, planning, or engineering by any organism, and no claim of that kind is made anywhere in this note.
Sources
Every number on this page comes from one of the following, all of which live in this project's own repository and are append-only records rather than summaries written after the fact.
- The Phase Y mutable-terrain and desktop-recovery record, and the Phase Y2 promotion record
- The Phase Y4 candidate autopsy, the Phase Y5 findings, and the Phase Y5 renderer-truth record
- The Phase Z voxel-substrate record, and the Phase Z2 ecological-succession record
- The simulation's terrain and perception layers, and the observer's brain inspector
Everything needed to check this claim is on this page already: the ruleset, the seeds, the duration, what was measured, the result, every limitation and the sources. Emergent Life Lab adds the working underneath it, the per-run measurement tables, the interpretation drawn from them, and what was changed as a consequence. See what Lab includes →