RootNet Research Series: v5.8.0 to v6.8

Status date: 2026-08-22
Latest packaged assay: v6.8 confirmed +0.035 water-field repair
Latest completed confirmatory evaluation: v6.8 fixed-candidate evaluation
Frozen ecological baseline: v5.8.0

Read This First

RootNet is now a series of related but distinct artificial-ecosystem models. A higher version number does not always mean that the earlier model has simply been replaced.

There are two research lines:

v5.8.0  frozen ecological transport baseline

v6.0  evolutionary kernel
  └─▶ v6.1  persistent spatial mycelium
        └─▶ v6.2  local module metabolism and partial necrosis
              └─▶ v6.2.1  calibrated patch and observer pacing
                    └─▶ v6.3  preregistered transport comparison
                          └─▶ v6.4  fixed versus adaptive conductance
                                └─▶ v6.5  dynamic-environment comparison
                                      └─▶ v6.6  continuous coordinates and rotation audit
                                            └─▶ v6.7  water-field/necrosis cause separation
                                                  └─▶ v6.8  fixed +0.035 water-capacity repair

v5.8.0 remains the reference for the original carbon/water/nitrogen transport ON/OFF and drought results.
v6.0 starts a separate model line from a common ancestral population. It is not an in-place upgrade of v5.8.0.
v6.1 through v6.2.1 progressively add a persistent spatial body, local metabolism, partial necrosis, and fragment identity. Their fixed-seed runs are mechanism validations, not replicated biological findings.
v6.3 is the completed transport-presence experiment, derived from v6.2.1. It tests three resource-translocation rules.
v6.4 and v6.5 test whether history-dependent adaptive conductance adds a detectable benefit over fixed conductance. v6.4 found no clear difference in the normal artificial environment. v6.5 also found no clear difference across its dynamic environments, but its formal status is COMPLETED_WITH_PROTOCOL_DEVIATION because 13 early-extinction runs were not padded with zeros through tick 360 in one endpoint calculation.
v6.6 is a continuous-coordinate engineering audit. Its three primary geometric endpoints passed all 24 preregistered rotation-equivalence comparisons, but 274/300 runs became extinct before tick 360 and active-patch uptake occurred in only 10/300 runs. Rotation equivalence is not evidence of ecological health.
v6.7 is a completed 160-run development assay, not a confirmatory evaluation. It classified the v6.6 water field as the dominant tested cause of the observed collapse. Its sealed result is retained as a reference-runtime result; cross-platform portability of its byte-exact dynamic-state digest remains unresolved.
v6.8 is a completed fixed-candidate confirmatory evaluation. After exploratory calibration, water_capacity_lift = 0.035, the acceptance criteria, and 20 evaluation-critical files were sealed before 30 completely unused confirmatory seeds were generated. All 120 control/candidate runs (60 pairs) completed, all eight predeclared checks passed, and the verdict was V6_8_LIFT_0_035_CONFIRMED.

All versions are artificial systems. None is calibrated to predict a real fungal species, forest, mycorrhizal network, or evolutionary history.

Version Map

Version	Role	Main biological unit	Resource representation	Evidence status
v5.8.0	Frozen ecological transport baseline	Predefined `fungus`, `alga`, and `symbiotic` individuals	Individual carbon, water, and nitrogen pools; conservative transfer between connected individuals	Replicated ON/OFF and drought experiments completed
v6.0	Minimal evolutionary kernel	Organisms in non-overlapping generations	Inherited traits, energy acquisition, costs, sharing, mutation, and dispersal	Apparatus mechanically validated; no sealed replicated diversification result
v6.1	Persistent spatial mycelium	Colony made of cells, tips, and edges	Uptake is spatial, but energy and water are pooled at colony level	Deterministic mechanism validation only
v6.2	Local module metabolism	Module, ramet, genet, and lineage	Local energy, water, vitality, starvation, fixed-conductance transfer, necrosis, and detritus	Deterministic mechanism validation only
v6.2.1	Patch to v6.2	Same as v6.2	Same transport/death framework, with reduced post-extension water reserve and event-paced display	Calibration and observer-usability validation; seed 620 is not an evaluation seed
v6.3	Paired transport experiment	Same as v6.2.1	`none`, `equalize`, and tip-oriented `demand` transport	Preregistered 30-seed paired evaluation completed; both transport conditions classified `SUPPORTED` against `none`
v6.4 r2	Adaptive-conductance comparison	Same module/ramet/genet system as v6.3	Fixed equalization versus history-dependent edge-conductance allocation	Preregistered 30-seed, 90-run evaluation completed; both primary comparisons were `NO_CLEAR_DIFFERENCE`
v6.5 r2	Dynamic-environment stress test	Same core body and transport network as v6.4	Fixed versus adaptive conductance under four environmental schedules	240-run evaluation completed with a protocol deviation; all eight primary comparisons remained `NO_CLEAR_DIFFERENCE` in the prespecified analysis and zero-padded sensitivity analysis
v6.6 r2	Continuous-coordinate rotation audit	Modules and edges in continuous 2-D coordinates	Euclidean geometry, bilinear field coupling, and fixed/adaptive transport	300-run confirmatory engineering audit completed; 24/24 primary rotation comparisons were equivalent within ±5%, but the assay showed severe ecological collapse
v6.7	Causal-separation development assay	Same continuous-coordinate system as v6.6	Two water fields crossed with immediate versus 5-tick-delayed necrosis response	160 development runs completed; `WATER_FIELD_DOMINANT` in the sealed reference runtime; no v6.7 confirmatory evaluation was designed or run
v6.8	Fixed water-field repair	Same continuous-coordinate system as v6.7	Exact `clamp(v66_water + 0.035, 0.08, 1.00)` environmental water-capacity field versus the v6.6 control	120-run, 30-seed paired confirmatory evaluation completed; all eight predeclared checks passed; `V6_8_LIFT_0_035_CONFIRMED`

Which version should I use?

Use v5.8.0 to reproduce the original ecological transport and drought study.
Use v6.0 to inspect the deliberately minimal evolutionary kernel.
Use v6.1 to observe persistent colonies with overlapping parent and offspring lifetimes.
Use v6.2.1 rather than v6.2 for the patched local-metabolism demonstration.
Use v6.3 to reproduce or inspect the completed network-translocation comparison.
Use v6.4 r2 to inspect the normal-environment fixed-versus-adaptive conductance comparison.
Use v6.5 r2 to inspect the dynamic-environment comparison, while retaining its protocol-deviation label.
Use v6.6 r2 to audit continuous-coordinate rotation equivalence, not as evidence that the ecological balance or nutrient-patch assay is healthy.
Use v6.7 to inspect the completed water-field/necrosis development diagnosis. Do not treat its diagnostic water field as a final model choice or its development seeds as confirmatory evidence.
Use v6.8 to inspect the confirmed +0.035 artificial-system water-field repair and its paired control/candidate result. Do not interpret 0.035 as a real-fungus water calibration or the fuller terminal image as morphological validation.

Results and counts should not be compared across versions as though only one parameter changed. The models change their life cycle, biological unit, resource accounting, death definitions, and research question.

Numerical portability and operating-system differences

RootNet uses floating-point geometry and math-library operations. From v6.4 onward, audits showed that a full raw simulation-state JSON can produce a different byte-level SHA-256 digest on another operating system or Python build even when the package files themselves are unchanged. In the v6.7 audit, Windows 11 runs with CPython 3.13.5 and 3.14.5 reproduced the sealed digest, while another Python 3.13.5 platform produced a different digest. The Python version number alone therefore does not define an equivalent numerical runtime; the OS, Python build, and underlying math library can matter.

Interpret verification outputs separately:

Artifact integrity: ZIP, manifest, source, protocol, seed-list, and result-file hashes must match exactly on every OS. A mismatch here can indicate a changed or damaged artifact.
Mechanical and numerical validation: unit tests, conservation laws, event/accounting checks, and float comparisons should use their stated tolerances.
Reference-runtime replay: an exact full-state digest is guaranteed only for the declared reference runtime. A different digest elsewhere is a portability warning, not by itself proof of package corruption or scientific falsification.

A portability warning is not automatically harmless either. If the claim depends on cross-platform replay, compare discrete IDs, events, counts, and topology exactly; compare continuous values with prespecified tolerances; and classify any discrete trajectory divergence separately from harmless last-bit numerical drift.

For exact sealed replay, use the runtime declared by the package. The v6.8 reference validation environment is Windows 11 x86-64 with CPython 3.14.5. Other platforms may run the model, but byte-exact dynamic-state reproduction is not promised. v6.7 should still be reported as:

WATER_FIELD_DOMINANT — sealed reference-runtime result; cross-platform portability audit pending

This qualification does not erase the sealed Windows result. It limits the claim about cross-platform replay until the full assay is independently rerun or compared semantically on another platform.

The v6.8 package reports artifact integrity, mechanical validation, the inherited v6.6 reference-runtime regression, and confirmatory-result reproduction separately. Its clean-extraction verification passed in the reference runtime.

What v6.3 Established

v6.3 asked a narrow question:

In the tested artificial environment, does mass-conserving resource translocation over the existing hyphal graph contribute to persistent growth and spatial exploration?

Thirty previously unobserved deterministic seeds were paired across three conditions for 360 ticks each. Seed 620, which had already been used during development, was excluded.

Endpoint, median across runs	No transport	Equalization transport	Demand transport
Living-module tick average	7.50	22.50	23.01
Distinct cells ever occupied	12.0	38.5	54.5
Ticks with extension	2.22%	7.22%	11.94%
Longest extension-free interval	128.0	60.0	77.5
Total extensions	10.5	34.5	50.5
Necrosis per 100 extensions	40.00	25.95	20.59
Final living modules	9.0	30.0	45.0

The preregistered rule classified both equalize and demand as SUPPORTED relative to none.

equalize improved mean living modules and explored cells in 30/30 paired runs. Growth-tick fraction was favorable in 28/29 non-tied pairs. Its longest-gap result did not remain significant after correction.
demand improved mean living modules in 30/30 pairs, explored cells in 28/28 non-tied pairs, growth-tick fraction in 28/30 pairs, and longest growth gap in 27/30 pairs.
Whole-ecosystem extinction was 0/30 in every condition.
Maximum observed absolute balance error was 3.55e-15 for energy and 4.44e-16 for water.

The supported interpretation is:

In this RootNet environment, conservative resource translocation over the hyphal network strongly supports sustained growth and spatial exploration.

The result does not show that transport is necessary for survival. The no-transport condition retained the graph itself, and all no-transport ecosystems survived. It also does not establish demand as universally superior to equalize; their direct comparison was secondary, post hoc, and mixed by endpoint.

What v6.4 Established

v6.4 r2 asked whether history-dependent allocation of a fixed total edge-conductance budget improves on uniform fixed conductance in the normal artificial environment. It compared adaptive-equalize with fixed-equalize; none was retained only as context.

Thirty new paired evaluation seeds were run in three conditions for 360 ticks, giving 90 complete runs. The two preregistered primary comparisons were:

Adaptive minus fixed	Median paired difference	Holm-adjusted p-value	Classification
Mean living modules	−0.198611111	0.855535552	`NO_CLEAR_DIFFERENCE`
Newly occupied cells	+0.5	0.848712444	`NO_CLEAR_DIFFERENCE`

All three conditions had 0/30 ecosystem extinctions. Resource-conservation and conductance-budget errors remained within tolerance.

The bounded result is that adaptive conductance was not clearly better or worse than fixed uniform allocation under this tested normal environment. This is a valid negative result; it does not show that adaptive conductance can never matter under other environmental schedules or damage regimes.

What v6.5 Established

v6.5 r2 retained the v6.4 conductance rules and compared fixed-equalize with adaptive-equalize under four schedules: a long-term fixed patch, a slowly moving patch, an abrupt relocation, and environmental edge damage. The design used 30 new paired seeds, two transport conditions, and four environments, for 240 runs and eight preregistered primary comparisons.

All eight comparisons were NO_CLEAR_DIFFERENCE. One unadjusted fixed-patch uptake comparison had p = 0.0428, but its Holm-adjusted value was p = 0.342, so it did not support a benefit. The maximum recorded conservation error was at most 1.42e-14.

The formal status is COMPLETED_WITH_PROTOCOL_DEVIATION. Thirteen runs ended early in ecosystem extinction, and mean living modules after extinction were not padded with zeros through tick 360 as the protocol required. A post hoc sensitivity analysis with zero padding left all eight classifications at NO_CLEAR_DIFFERENCE. That supports robustness of the qualitative observation but does not erase the deviation.

The abrupt-relocation uptake endpoint was also weak: fixed had no new-patch uptake in 27/30 runs, adaptive in 25/30, and 23 pairs tied. The experiment therefore did not establish adaptive conductance as either a benefit or a protective reserve under the tested schedules. The 30 evaluation seeds are used and frozen; a corrected confirmatory test would require a new version and new seeds.

What v6.6 Established and Exposed

v6.6 r2 changed the spatial representation to continuous two-dimensional coordinates while preserving the ecological rules. It added Euclidean edge lengths, continuous spatial indexing, bilinear environmental-field coupling, arbitrary-angle rotation, and a circular evaluation region within the square arena.

Thirty locked evaluation seeds were run across five angles and two transport modes, for 300/300 completed records. For the three primary endpoints—living modules, total network length, and convex-hull area—all 24 preregistered comparisons against 0° were classified EQUIVALENT_WITHIN_5_PERCENT. This is a successful engineering rotation-equivalence audit.

It is not a successful ecological-health demonstration:

274/300 runs, or 91.3%, became extinct before tick 360;
active-patch substrate uptake was positive in only 10/300 runs;
all ten positive observations came from the same evaluation seed across two transport modes and five angles;
the active-patch endpoint was auxiliary and was not used to produce the primary rotation verdict.

Accordingly, v6.6 supports the narrow claim that the tested geometric outputs were stable under rotation. It does not support the claim that nutrient exploration, survival balance, or the overall ecology was healthy. The frozen v6.6 result should not be retroactively tuned; its collapse became the question addressed by v6.7.

What v6.7 Established

v6.7 was a predeclared development cause-separation assay, not a confirmatory evaluation. It crossed two water-field profiles with two necrosis responses, using 20 new development seeds at 0°, both fixed and adaptive transport, for 160 runs:

Condition	Water field	Necrosis response	Extinct by tick 360
A	v6.6 radial field	Immediate	32/40 = 0.800
B	v6.5-equivalent depth-stretched field	Immediate	6/40 = 0.150
C	v6.6 radial field	5-tick grace	31/40 = 0.775
D	v6.5-equivalent depth-stretched field	5-tick grace	6/40 = 0.150

Changing only the water field reduced extinction by 26/40 runs, or 0.65. Adding only the necrosis grace reduced it by 1/40, or 0.025, and B and D did not differ. Under the preregistered rule, the sealed reference-runtime classification was WATER_FIELD_DOMINANT.

The supported interpretation is that chronic water limitation in the v6.6 water field was the dominant tested driver of the collapse, with necrosis largely following that shortage. It is not evidence that necrosis is irrelevant in general.

The v6.5-equivalent water field was a diagnostic condition, not a selected final repair. Only one founder had a positive initial water balance under that field. No v6.7 confirmatory evaluation was designed or run, and no real-organism calibration was attempted. The correct current label, including the runtime limitation described above, is:

WATER_FIELD_DOMINANT — sealed reference-runtime result; cross-platform portability audit pending

What v6.8 Established

v6.8 fixed one candidate before confirmatory evaluation:

v68_water = clamp(v66_water + 0.035, 0.08, 1.00)

This is a lift to the environmental water-capacity field and the environmental water stock initialized from it. It is not an increase in an individual module's internal water-storage limit. Substrate, patch geometry, growth, branching, transport, conductance, immediate necrosis, detritus, spores, fusion, rotation, and keyed randomness were held fixed.

The candidate value, acceptance criteria, protocol, model, tests, and evaluation code were sealed before 30 completely unused confirmatory seeds were generated. Each seed was paired across the v6.6 radial control and the v6.8 candidate under fixed and adaptive equalization transport, for 120 completed runs.

Confirmatory endpoint	v6.6 control	v6.8 `+0.035`
Ecosystem extinction	50/60 = 0.833	12/60 = 0.200
Median mean living modules	8.96	12.80
Median mean total hyphal length	7.40	10.71
Median mean tips	3.65	5.03
Median mean branch modules	0.95	1.68
Pooled cumulative necrosis fraction	0.978	0.822
Median water-limited module-tick fraction	0.233	0.231

Candidate extinction was 7/30 under fixed transport and 5/30 under adaptive transport. The candidate/control ratios of medians were 1.447 for hyphal length, 1.380 for tips, and 1.768 for branch modules. Transport-specific paired improvement counts were 30/30, 28/30, and 28/30 for fixed length, tips, and branches, and 30/30, 29/30, and 27/30 under adaptive transport. All exceeded their predeclared thresholds. The maximum mechanical error was 1.78e-15. The confirmatory verdict was V6_8_LIFT_0_035_CONFIRMED.

The supported interpretation is narrow:

In this fixed artificial system, a small increase in local environmental water capacity reduced whole-ecosystem extinction and cumulative necrosis while increasing living-module, hyphal-length, tip, and branch endpoints, without erasing partial necrosis or reversing the result across the two transport modes.

The nearly unchanged median water-limited module-tick fraction shows that the result cannot be summarized as simply eliminating exposure to water-limited ticks. The confirmatory summary did not establish whether shortfalls became shallower, occurred at different modules or times, or were converted into different growth and death histories through path dependence.

It also did not test whether preserved graph connectivity mediated the extinction reduction. Largest-component fraction, bridge loss, fragmentation timing, post-necrosis reconnection, and related topology-mediated pathways were not predeclared acceptance endpoints. "Preventing spatial holes" is therefore a plausible follow-up hypothesis, not an established mechanism.

Relationship to the v5.8.0 drought result

An insurance interpretation is useful only if kept explicitly hypothetical. v5.8.0 showed that resource transport reduced deaths under all tested rainfall regimes, but under natural rainfall its reduction in births and growth was larger, leaving fewer final survivors. Under the 120-day drought, reduced water deficiency and deaths coincided with more final survivors. This supports a stress-dependent tradeoff in redistribution, not a demonstrated mechanism that stored resources during normal times for later use.

Likewise, v6.8 supports a benefit from a small local environmental water-capacity margin, but it does not yet demonstrate "spatial insurance" through connectivity preservation. Because v5.8.0 and v6.8 use different biological units, life cycles, and resource accounting, their results do not establish one unified two-layer insurance mechanism.

The bounded cross-version hypothesis is:

Resource redistribution in v5.8.0 and local environmental water margin in v6.8 may both alter the probability of crossing catastrophic failure thresholds. A temporal-insurance plus spatial-insurance interpretation is a useful research hypothesis, not a result already demonstrated across the series.

How the v6 Line Developed

v6.0 — Minimal evolutionary kernel

v6.0 began a new research line without modifying v5.8.0. It removed predefined fungus, alga, symbiont, species, and ecotype states from the dynamics.

It added:

one common-ancestor clonal initialization;
five inherited continuous parameters;
deterministic phenotype derivation and explicit costs;
habitat-dependent acquisition without habitat-dependent mutation;
conservative local resource redistribution;
non-overlapping survival and reproduction;
random inherited mutation and adjacent-patch dispersal;
homogeneous and heterogeneous habitat presets;
subject-keyed random streams;
full ancestry and per-generation machine-readable traces.

The environment cannot directly rewrite the genome during an organism's life. Any adaptation must arise through differential survival and reproduction of inherited variants.

v6.0 provides an apparatus for a diversification experiment. A single default run is not evidence that ecological diversification occurred.

v6.1 — Persistent spatial mycelium

v6.1 replaced whole-population generational turnover with persistent, overlapping colony lifetimes. A colony became an explicit graph of cells, tips, and edges that could grow, branch, release spores, and form contact-mediated fusion links.

Important accounting changes:

releasing a spore does not remove the parent;
established offspring can coexist with their parent;
actual colony death is separated from spore establishment failure;
failed or expired spores are not counted as dead colonies;
the observer is read-only and cannot change the simulation.

The v6.1 environment is spatial, but absorbed resources still enter a colony-wide energy/water wallet. This limitation motivated v6.2.

The fixed seed-610, 80-tick mechanism validation observed four living colonies, 333 hyphal cells, one living parent-child pair, no actual colony deaths, two establishment failures, and three conservative fusion links. This was a deterministic validation run, not a replicated ecological result.

v6.2 — Local metabolism and partial necrosis

v6.2 replaced the colony-wide wallet with local state on each hyphal module:

energy;
water;
vitality;
starvation duration.

It also added:

simultaneous donor-capped, fixed-conductance transfer;
gradual weakening and module-level necrosis;
limited recycling, visible detritus, and delayed spatial reuse;
explicit module, ramet, genet, and lineage identity;
survival of disconnected fragments;
explicit same-genet re-anastomosis;
a separate non-self boundary for contact with another genet.

Mere adjacency no longer creates an edge automatically. Additional loops require an explicit anastomosis event.

The model records four distinct outcomes:

module necrosis;
ramet death;
genet extinction;
spore establishment failure.

These must not be combined into one generic "death" count.

The fixed seed-620, 360-tick mechanism validation observed six necrotic modules, two fragment ramets, two ramet deaths, one genet extinction, and one successful offspring establishment while the parent genet remained alive. This single run demonstrated that the mechanisms operated; it was not a general ecological or evolutionary result.

v6.2.1 — Patched local-metabolism demonstration

v6.2.1 preserved v6.2.0 and made two bounded changes:

it reduced an unnecessary post-extension water reserve that had produced long visual plateaus in the default demonstration;
it made the terminal observer skip redraws when the visible state had not changed, while still simulating every tick.

--every-tick restores one frame per tick. The transport equation, mutation rule, death definitions, and claim boundary were not changed.

Seed 620 influenced this adjustment and is therefore a development/calibration seed, not independent evidence. v6.3 excludes it from the evaluation set.

v6.3 — Separate transport comparison package

v6.3 preserved v6.2.1 and added three experimental modes:

none: no energy or water moves between modules;
equalize: fixed-conductance movement down local stock differences;
demand: a fixed-conductance source-sink rule that reserves two ticks of maintenance and adds one extension cost for tips.

All proposed outflows are aggregated per donor, proportionally capped, and applied simultaneously. Energy and water are independently conserved. Conductance does not adapt.

The evaluation question, endpoints, seed derivation, and decision rule were fixed before the evaluation runs were inspected. The current result is described in What v6.3 Established.

v6.4 — Fixed versus adaptive conductance

v6.4 was deliberately reduced to one mechanism question after an earlier extended prototype proved too broad. The released simple r2 package added adaptive-equalize, in which edge conductances strengthen or decay from realized transport history while remaining under a fixed total budget. Its comparator, fixed-equalize, retained uniform conductance. A two-round flux aggregation issue was corrected before evaluation, and raw cross-platform state hashes were removed from the sealed validation in favor of semantic regression plus exact replay within the same runtime.

The 90-run evaluation found no clear adaptive benefit or harm on either primary endpoint. The result is described in What v6.4 Established.

v6.5 — Dynamic environment and edge damage

v6.5 froze v6.4 r2 and changed the environment rather than the adaptive rule. It introduced scheduled patch motion and explicit environmental edge damage, with environmental events separated from growth randomness. Environmental edge deletion remained distinct from module necrosis, ramet death, genet extinction, and establishment failure.

The completed evaluation found no clear fixed/adaptive difference in any of eight primary comparisons, but the early-extinction zero-padding error requires the permanent COMPLETED_WITH_PROTOCOL_DEVIATION label. See What v6.5 Established.

v6.6 — Continuous coordinates and rotation audit

v6.6 converted positions and environmental coupling from a four-neighbor lattice to continuous geometry. The purpose was to test whether coordinate-axis artifacts remained, without presenting geometric equivalence as biological validation. Zero-valued endpoints were prevented from becoming artificial log-ratio equivalence through pseudocounts; inactive patch uptake was retained only as an auxiliary diagnostic.

The primary engineering audit passed, but the ecological diagnostics revealed widespread extinction and nearly inactive patch use. See What v6.6 Established and Exposed.

v6.7 — Water-field/necrosis causal separation

v6.7 preserved the v6.6 r2 baseline and used a new seed namespace to separate water-field effects, necrosis-grace effects, and their interaction. A five-tick grace tolerates five consecutive shortfall ticks; the existing vitality-decay rule begins on the sixth, and a fully paid tick resets the counter. It does not weaken the decay coefficient. The package also added founder-level water budgets, resource-limited module-ticks, first-necrosis time, cumulative necrosis, patch-distance, and uptake diagnostics.

The development assay classified the tested water-field change as dominant. This result diagnoses v6.6 but does not select a final repaired ecology. See What v6.7 Established.

v6.8 — Fixed `+0.035` water-field repair

v6.8 retained the v6.6/v6.7 continuous-coordinate dynamics and changed only the environmental water-capacity field to exact clamp(v66_water + 0.035, 0.08, 1.00). An initial +0.15 development candidate produced fuller networks but zero whole-ecosystem extinctions in 40/40 candidate runs and failed its predeclared 10–50% extinction window. Separate exploratory seed namespaces located the transition near +0.03 to +0.04; these calibration results were not used as final evidence.

The +0.035 value, confirmatory criteria, and critical files were then fixed before a new 30-seed paired evaluation. The final 120-run evaluation passed all eight criteria and classified the candidate V6_8_LIFT_0_035_CONFIRMED. See What v6.8 Established.

Frozen v5.8.0 Ecological Baseline

Overview

RootNet v5.8.0 is an artificial ecosystem inspired by fungi, algae, and symbionts—individuals combining both functions.

It is not a predictive model calibrated against measured data from any specific real organism. It is not intended to estimate the population size or mortality of real forests or mycorrhizal networks. It is an experimental world for studying what kinds of resource distribution, survival, growth, reproduction, and fragmentation/reconnection arise from defined rules.

Each individual holds three internal resources: carbon, water, and nitrogen. RootNet—the underground network connecting individuals—transports resources between connected individuals.

Transport has the following properties:

the exact amount subtracted from the supplier is added to the receiver;
nothing is lost to the outside during transport;
the total amount of each resource is conserved across the whole population during transport.

RootNet does not create resources. It redistributes resources that already exist among connected individuals.

Biological Model

Three functional groups

fungus: does not photosynthesize and obtains carbon from organic substrate.
alga: obtains carbon from light through photosynthesis.
symbiotic: uses both photosynthesis and substrate utilization.

These predefined groups belong to v5.8.0. They were deliberately removed from the v6.0 evolutionary dynamics.

Three resources

Each v5.8.0 individual holds three pools, each in the range 0.0–1.5:

carbon: energy for maintenance and growth;
water: hydration state;
nitrogen: a growth requirement.

The model also records how many consecutive days each resource has been insufficient. Adequate days gradually repair the accumulated deficiency state.

One-day sequence

Each day is computed approximately in this order:

resource acquisition from light, substrate, rain, storage, and soil;
conservative RootNet transport between connected individuals;
maintenance consumption;
deficiency and death checks;
growth;
environmental load, biosecurity, reproduction, and fragmentation/reconnection updates.

Reproduction and inheritance

Eligible individuals produce seeds, and seeds can germinate into new individuals. Offspring inherit parental genes and physiological traits with small mutations.

Values such as resource_sharing are physiological parameters derived from genes and morphology, not personality labels.

For the exact variables and equations, see BIOLOGICAL_MODEL_SPEC.md in the v5.8.0 package.

v5.8.0 Experiment

The study compared resource transport ON and OFF under three rainfall conditions:

natural rainfall;
a 60-day drought;
a 120-day drought.

The OFF condition retained connections, signaling, biosecurity, and fragmentation/reconnection. Only carbon, water, and nitrogen transport was disabled. The experiment did not dismantle the network itself.

v5.8.0 Main Results

All results below are internal to the v5.8.0 artificial ecosystem.

Natural rainfall

Final survivors: ON 39.54, OFF 49.45
Total deaths: ON 90, OFF 205
Mean paired difference in final survivors, ON minus OFF: −9.91

Transport reduced deaths, but it reduced births by more, leaving a smaller final population.

60-day drought

Final survivors: ON 32.01, OFF 33.40
Total deaths: ON 72, OFF 177
Mean paired difference in final survivors: −1.39

The gap narrowed relative to natural rainfall. This does not establish a tipping point; it is only an intermediate observation.

120-day drought

Final survivors: ON 16.60, OFF 14.27
Total deaths: ON 123, OFF 405
Water-deficiency individual-days: ON 29.23, OFF 115.01
Mean paired difference in final survivors: +2.33

Under strong drought, transport sharply reduced water deficiency and deaths, and the final survivor count was larger.

[Natural rainfall]
Transport ON ─▶ deaths decrease ─▶ births decrease even more ─▶ smaller final population

[120-day drought]
Transport ON ─▶ water deficiency decreases ─▶ deaths decrease ─▶ larger final population

As a supplementary observation, the final number of symbionts was larger under transport ON in all three rainfall conditions. This indicates that the modeled benefit was not uniform across functional groups.

Path of the v5.8.0 Birth Difference

Under natural rainfall, 200 paired trajectories were used to trace where the final population difference began.

Metric	ON mean	OFF mean
New seeds	43.00	69.93
Germinations	31.44	42.51
Total growth	2274.82	2992.37
Reproduction-eligible individual-days	162.97	259.59

The median first-appearance order was:

transport, pre-reproduction resources, and growth: day 2;
reproduction-readiness state: day 24;
new seed count: day 52;
germination and survivor count: day 57.

This supports the bounded interpretation that the birth difference arose before germination, through earlier growth, reproduction preparation, and seed production stages.

It does not identify resource leveling, donor cost, or population pressure as a sole direct cause. Within the ON condition, net exporters averaged 1.31 seeds and net receivers averaged 0.95 seeds. That descriptive tally does not estimate the causal effect of being an exporter or receiver.

Relationship to Existing Theory

The broad direction overlaps with existing ecological and evolutionary theory, including stress-gradient arguments, models in which cooperation matters more under harsh conditions, models of reproductive cost and population resilience, and experiments showing water movement through mycorrhizal connections.

RootNet did not discover a new general theory of evolution. Its v5.8.0 contribution was a controlled audit inside one artificial ecosystem: conservative transfer of three resources, a reversal between natural rainfall and strong drought, and the measured path by which a birth difference appeared.

References retained from the v5.8.0 README

See rootnet_existing_theory_audit_ja.md in the v5.8.0 materials for the detailed point-by-point comparison.

v5.8.0 Verification

Resource-transport ON/OFF experiment

seeds 0–99;
natural rainfall, 60-day drought, and 120-day drought;
transport ON and OFF for each condition;
600 trajectories and 144,000 simulation-days;
transport under OFF exactly 0;
all mass-balance errors below 1e-10;
overall verdict: PASS.

Birth-path audit

seeds 0–99;
natural rainfall, paired ON and OFF;
200 trajectories and 48,000 simulation-days;
weather and solar radiation identical within each seed pair;
maximum absolute mass-balance error 5.684341886081e-14;
overall verdict: PASS.

v5.8.0 file guide

BIOLOGICAL_MODEL_SPEC.md — variables, formulas, and deficiency-death rules;
ROOTNET_TRANSPORT_ABLATION_100SEED_SUMMARY.md — observations and differences by rainfall condition;
ROOTNET_TRANSPORT_ABLATION_100SEED_VERIFICATION.md — verification record for the ON/OFF experiment;
BIRTH_PATH_AUDIT_PROTOCOL.md — fixed procedure and stopping rules for the birth-path audit;
BIRTH_PATH_AUDIT_REPORT.md — birth-path results;
BIRTH_PATH_AUDIT_VERIFICATION.md — verification record for the birth-path audit;
WEATHER_PAIRING_AUDIT.md — weather and solar-pairing audit.

Per-parent, per-day, and per-resource details are stored in the accompanying JSONL records.

Evidence and Claim Boundaries Across the Series

Supported at the stated scope

v5.8.0 supports its transport ON/OFF findings only inside the v5.8.0 artificial ecosystem and tested conditions.
v6.0–v6.2.1 validations support that their specified mechanisms execute reproducibly and satisfy their mechanical checks.
v6.3 supports a benefit of conservative graph-based resource translocation for sustained growth and exploration under its tested environment.
v6.4 records that its tested adaptive allocation did not clearly outperform or underperform fixed allocation in the normal artificial environment; this is not an equivalence claim.
v6.5 records that no clear adaptive effect was detected under its four tested schedules, subject to the permanent protocol-deviation qualification; this is not an equivalence claim.
v6.6 supports engineering rotation equivalence for three geometric endpoints within the preregistered ±5% bounds; it simultaneously documents an ecologically collapsed assay state.
v6.7 supports WATER_FIELD_DOMINANT only as a sealed development result in the reference runtime. Cross-platform replay remains pending, and the diagnostic water field is not a final repair.
v6.8 supports V6_8_LIFT_0_035_CONFIRMED inside its fixed artificial system: candidate extinction remained within the predeclared 10–50% window, extinction fell by more than 25 percentage points relative to control, partial necrosis remained present, mechanical tolerances passed, and predeclared hyphal-length, tip, branch, paired-improvement, and transport-direction criteria all passed.

Not established

prediction of real organism, forest, soil, or mycorrhizal-network population sizes;
reconstruction of real fungal evolution or real evolutionary history;
emergence of fungi from unconstrained physics or chemistry;
real-world optimality of the implemented transport rules;
universal superiority of demand-based transport;
universal benefit, harm, or uselessness of adaptive conductance;
necessity of transport for survival;
evolution of resource sharing because of drought;
long-term evolutionary-fitness advantage of transport, an optimal sharing rate, or the drought frequency required for sharing to evolve;
operational speciation, coevolving symbiotic partners, or evolution of adaptive conductance;
ecological health inferred from rotation equivalence alone;
cross-platform byte-identical simulation trajectories for v6.4–v6.8;
biological correctness, real-world optimality, or species calibration of the v6.8 +0.035 water-capacity lift;
connectivity preservation as the mediator of the v6.8 extinction reduction;
a demonstrated temporal-insurance plus spatial-insurance mechanism linking v5.8.0 and v6.8;
the magnitude, duration, timing, or module identity of water shortfalls as the established mediator of the v6.8 result;
calibration of any model parameter to a real fungal species.

In the later spatial v6 packages, growth locations, topology, resource histories, necrosis, fragmentation, germination, and lineage outcomes are generated from local state and keyed stochastic events. The author still supplies fungal-like primitives such as tips, edges, extension, spores, and anastomosis. The viewer remains observational and does not feed back into the dynamics.

Running the Packages

Each version is a separate directory or ZIP. Open the intended version's directory before running its commands.

v5.8.0

python main.py --days 10 --seed 42
python -m unittest discover -v
python validate_v5_8_0.py

The default save target is fusa_rootnet_v5_8_0.json, and the lineage tree is fusa_lineage_tree.html.

v6.0

python watch.py --generations 40 --seed 600
python main.py --generations 20 --seed 600 --output smoke_output
python -m unittest discover -v
python validate_v6_0.py
python verify_package.py

v6.1

python watch.py --ticks 200 --delay 0.05
python main.py --ticks 80 --seed 610 --output run_output
python -m unittest discover -s tests -v
python validate_v6_1.py
python verify_package.py

v6.2.1

python watch.py --ticks 360 --delay 0.04
python watch.py --every-tick
python main.py --ticks 360 --seed 620 --output run_output
python -m unittest discover -s tests -v
python validate_v6_2.py
python verify_package.py

v6.3

python watch.py --transport none
python watch.py --transport equalize
python watch.py --transport demand

python main.py --transport demand --seed 620 --ticks 360 --output run_output_demand

python -m unittest discover -s tests -v
python validate_v6_3.py
python verify_package.py

To reproduce the sealed 30-seed evaluation:

python experiment_v6_3.py --ticks 360 --replicates 30 --output reproduced_evaluation_v6_3.json

v6.4 through v6.8

Each later package includes its own version-specific commands and sealed settings. Start with the non-mutating package checks from the extracted package directory:

python -m unittest discover -s tests -v
python verify_package.py

Do not launch or alter a sealed evaluation merely to test installation. Follow the package protocol for the exact experiment command, dimensions, ticks, conditions, and locked seed list. A v6.5 replay must retain the published protocol-deviation label, and v6.7 has no confirmatory evaluation to reproduce. The v6.8 package contains its sealed critical-file record, locked confirmatory seeds, 120-run result, and reproducing acceptance logic.

Python 3.10 or newer is sufficient for v6.1–v6.3; the packages use the standard library. v6.0 recommends Python 3.11 or newer. For v6.4–v6.8, use the exact runtime declared inside the package for byte-level sealed replay and read Numerical portability and operating-system differences before interpreting a digest mismatch.

Package Guide

The v6 line is distributed as separate packages:

rootnet_v6_0_evolutionary_kernel.zip
rootnet_v6_1_spatial_mycelium.zip
rootnet_v6_2_module_metabolism.zip
rootnet_v6_2_1_module_metabolism.zip
rootnet_v6_3_transport_experiment.zip
RootNet_v6_4_simple_adaptive_conductance_r2.zip
RootNet_v6_5_dynamic_environment_r2.zip
RootNet_v6_6_r2_continuous_coordinates.zip
RootNet_v6_7_causal_separation.zip
RootNet_v6_8_lift_0_035_confirmed.zip

Each ZIP has a companion _SHA256.txt file. Inside the packages:

README.md gives version-specific operating instructions;
MODEL_SPEC.md defines the implemented mechanics;
RESEARCH_BOUNDARIES.md defines the allowed claims;
RELEASE_NOTES_*.md lists changes from the preceding package;
validation_*.json records deterministic mechanical validation;
verify_package.py checks the package manifest and reproducibility.

The v6.3 package additionally contains:

V6_3_PREREGISTERED_PROTOCOL.md — question, conditions, endpoints, seeds, and decision rule fixed before evaluation;
EVALUATION_REPORT_V6_3.md — paired results and bounded interpretation;
evaluation_v6_3.json — machine-readable evaluation data;
experiment_v6_3.py — evaluation reproducer.

The v6.4–v6.8 distributions additionally provide version-specific preregistered protocols, locked seed lists, validation records, machine-readable evaluation or development results, and companion seal/hash records. Keep these paired with the exact package revision. In particular:

v6.4 r2 is the evaluated simplified adaptive-conductance package; the earlier extended prototype is not the evaluation package;
v6.5 r2 must remain labeled COMPLETED_WITH_PROTOCOL_DEVIATION;
v6.6 r2 separates three primary rotation endpoints from auxiliary active-patch uptake;
v6.7 is a causal-separation development package with 160 completed runs and zero confirmatory evaluation runs.
v6.8 fixes water_capacity_lift=0.035, retains a 20-file pre-seed critical seal, and contains 120 completed confirmatory control/candidate runs with verdict V6_8_LIFT_0_035_CONFIRMED.

The packaged v6.3 ZIP SHA-256 is:

3E4E0DC7B666DD4069E52D4D9AA2712419FB49021A6E97F4CC00F2667526AEE3

Later sealed package ZIP hashes recorded in the development logs are:

Package	ZIP SHA-256
v6.4 simple r2	`E9CC0FD6A2EF3C22759FBFCC3865BA88F79920E3789DC113F2FE86296F826D6C`
v6.5 r2	`7769910BD7408F393B240E074DCC7FF52785FDA5F1404D08C6D07AE3ACFD7C50`
v6.6 r2	`AC6CFB094D5C14A263E1C9A39B3C946B7E6ECA61B022239C8F3AA1643DB804E2`
v6.7 causal-separation package	`BD21398C6F262257CB3E5B9FDEC6A84339ADADDCC974E4AD3F82861D406B5F49`
v6.8 confirmed `+0.035`	`771A0033AE13994990F9F53DA0FB84B68DB9ABEE04EE90F13D2A627BCE0DD932`

The later outer archive RootNet_v6_7_Causal_Separation_SEALED_2026-08-21.zip has SHA-256 D7FB473A68AE780FF661856644C5398006D857A190D3A727CF320EC2C7054F31; it contains the unchanged inner v6.7 package. The v6.8 hash above is for RootNet_v6_8_lift_0_035_confirmed.zip. Package hashes establish artifact identity, not cross-platform equality of regenerated floating-point state digests.

Do not overwrite the frozen v5.8.0 baseline or an earlier v6 package when working with a later version. Extract each version into its own directory and retain its matching hash and validation records.

Name		Name	Last commit message	Last commit date
reports		reports
zip		zip
LICENSE		LICENSE
README.md		README.md

Latest commit

History

Folders and files

Repository files navigation