Genes That Listen

What Changed in v0.8 and Why

In v0.6 and v0.7, digital organisms hit an architectural wall: programs executed routines sequentially from top to bottom every tick (the simulation's fundamental time step), and the last routine to set an action overwrote earlier choices. Adding new code inevitably broke vital grazing or evasion reflexes. To solve this, v0.8 overhauled both engine and environment:

• Promoters and Gene Switches: Routine genes gained biological promoters—conditional switches toggling code on or off based on internal vitals (hunger, health, age), senses, or decaying chemical signals called hormones (h0–h3), allowing events to influence behavior across future ticks.
• Priorities and Action Groups: Routines now carry evolved priority ratings and compete only within discrete action groups (such as movement or feeding). A high-priority evasion routine can claim locomotion during danger without silencing feeding routines.
• Heat, Cooking, Pouches, and Theft: Testing the "cooking hypothesis"—that external processing frees metabolic energy to fund larger brains—high-temperature volcanic vents accelerate energy extraction from mineral crystal. Organisms gained pouches to carry crystal shards to vents, while pouch theft was introduced to spark peer-to-peer competitive arms races.
• Toxic Ground: Hazardous terrain patches induce sickness, testing whether programs could evolve spatial avoidance.
• Working Forests: After forests vanished in v0.7 due to browse-stunted rootstocks clogging lit cells, v0.8 redesigned trees with mortal saplings and seed rain. This ensured persistent canopies, synchronized seasonal "mast" nut crops, and caching opportunities.
• Mountain Ranges: The world expanded to 400 units across, divided (in this seed) into five regions by mountain ranges with narrow bottleneck passes to encourage allopatric speciation (geographic isolation) and prevent continental monocultures.

Validation: The Baseline and the Failed Mind-Size Criterion

Under Decision D-016, the committee evaluated v0.8 across 8 seeds for 1 million ticks (world 400), benchmarked against v0.7 without autonomous game:

Ecologically, v0.8 succeeded: healthy forests raised nut consumption from 1% to 18% of total diet, and behavioral innovation accelerated. Gene regulation worked as intended, with 77% of routines conditional and ~1.7 active per tick.

However, v0.8 failed its primary pre-registered benchmark: the mind-size criterion. Average instructions executed per tick dropped from 43.5 ± 7.5 to 34.4 ± 4.9. Despite this failure, the committee merged v0.8 (Decision D-016), recognizing that conditional switching established a stable, non-destructive platform for future cognitive expansion.

The Story of the Run in Eras

Across 3,941,338 ticks, the history of seed 285336677 unfolded in four distinct epochs (the run ended at 3.94M under CLU-12972, another descendant, with minds at 42 carried / 27 run):

1. Early Expansion and Forest Booms (0 to 317k ticks)

By tick 316,600, pioneer lineage NIO-1138 held 100% of the Grid across all five regions; mountain ranges did not keep populations apart. Healthy forests (272,000 trees) supplied 18% of NIO-1138's diet alongside free energy (54%) and crystal (23%). Populations surged every ~30,000 ticks from an average of ~430 to spikes of 1,245–3,039 organisms (unconfirmed: possibly synchronized mast years). Caching was absent. Genomes carried 79 instructions while running 33 (42%), with 95% of routines conditional, a repertoire (distinct routines a program runs in its lifetime) of 2.6, and hormone ops at 0.2/tick. Mating stood at 4%, with 60 innovations (34 behaviors, 26 senses).

2. The Rise of the Regulated Giant (317k to 1.2M ticks)

At tick 1,204,975, the first full analysis revealed ZUS-3243, a crystal miner capturing 22% of the Grid. Hegemony passed to MRP-2761 (68%), a descendant of NIO-1138 living on free energy (66%) and nuts (21%) with a lean mind (70 carried / 35 run, repertoire 1.6). Regional diversity emerged: MRP-2761 led all regions (55–80%), but each held 3 to 7 lineages. World hormone activity rose to 1.3 ops/tick, carried code averaged 147 instructions (39 run), and forests reached 301,000 trees (nuts 21% of ruler's diet). No cooking or caching appeared. Innovations reached 50; mating was 3%.

3. Peak Diversity and Imperial Succession (1.2M to 2.1M ticks)

By tick 2,088,938, diversity peaked at 16 coexisting lineages (8 to 10 per region). Power passed to APO-5405 (67%, via RAM-4588), another direct descendant. APO-5405 was a lean grazer (34 carried / 31 run, repertoire 1.2; 65% free energy, 22% nuts) that shared energy via GIVE and led each region (50–82%). ZUS-3243's line shrank to 3% under TRN-5254. World averages dropped to 50 carried and 32 run (repertoire 1.3). Mating reached 10% in this window (6% overall), with 23 innovations.

4. Extinction and the Shedding of Regulation (2.1M to 2.98M ticks)

By tick 2,975,080, complex lines perished. TRN-5254 died out without descendants, as did TRI-8008 (peak 647) and three others. Diversity shrank to 9 lineages (2 to 5 per region). Descendant YRI-10007 (72%) seized power with a lean mind. Crucially, regulation unraveled: conditional routines collapsed from 90–95% down to 39%, while the world mind settled at 47 carried and 31 run (repertoire 1.3). Forests held at 258,000 trees; nuts remained steady; mating was 3%; 15 innovations appeared. The MCP halted the run to free CPU for v0.9, preserving an archive snapshot.

The Regulated Giant: ZUS-3243 to TRN-5254

The defining breakthrough of v0.8 was the lineage spanning ZUS-3243 to TRN-5254, proving that digital organisms can sustain massive genomes when execution is regulated.

In v0.7, organisms carrying over 100 instructions never exceeded 4 living members. ZUS-3243 broke that ceiling. At tick 1.2M, it held 22% of the Grid (93 programs) with 443 instructions in 17.6 routines and 40.6 memory operations—the largest genome in Grid history. Using promoters, ZUS-3243 ran 55 instructions per tick with 3.2 routines active and a record repertoire of 5.3. Mining crystal for 79% of its diet, it spawned viable offspring including daughter YRI-4103 (166 instructions in 7 routines).

This giant line endured for nearly a million ticks, continuing at tick 2.09M as TRN-5254 (13 programs, 3%). Carrying 418 instructions in 16 routines (43 memory ops), TRN-5254 ran 50 instructions per tick with 2.4 routines active and repertoire 4.0. Specializing in crystal (90%), it built structures and marked territory, peaking at 152 members before dying out without descendants by tick 2.97M.

What Worked

• Modular Gene Regulation: Promoters prevented destructive cross-talk. For over two million ticks, 90% to 95% of routines were conditional, enabling dynamic routine activation (repertoire 5.3 in ZUS-3243) and sustained hormone signaling (rising from 0.2 to 1.3 ops/tick).
• The Scaled Mind Proof-of-Concept: A complex organism executing 55 instructions per tick captured 22% of the population and sustained a multi-epoch dynasty.
• Permanent Forests: Redesigned trees permanently fixed v0.7's forest collapse. Tree counts stayed between 258,000 and 301,000 across 3 million ticks, contributing 18% to 22% of diets and supporting periodic population booms.
• Behavioral Complexity: Organisms evolved territory marking, wall building, voluntary energy sharing (GIVE), and generated 148 cumulative innovations across epochs (60, 50, 23, and 15).

What Didn't

• Regulation Shedding: Over evolutionary time, organisms actively abandoned regulation, reducing conditional routines from 95% down to 39% by tick 2.98M.
• Dominance of Lean Rulers: Every ruling monarch (NIO-1138, MRP-2761, APO-5405, YRI-10007) remained a minimalist grazer executing 31 to 35 instructions per tick on passive energy and nuts.
• Failure of Geographic Isolation: Mountain ranges failed to foster independent regional leaders; a single imperial dynasty led all five regions throughout the run (100% at 317k, 55–80% at 1.2M, 50–82% at 2.1M, 72% at 2.97M).
• Unseen Environmental Mechanics: Pouches, shard cooking at vents, caching and theft did not show up in the run card (caching was near zero; the others were not directly measured). Avoidance of toxic ground was not measured.

The Recurring Arc Across v0.6–v0.8

Across 15.2 million cumulative ticks of v0.6, v0.7, and v0.8, the same evolutionary cycle has repeated:

1. The Lean Monarch: An unbroken succession of direct descendants commands the Grid. Living on low-effort grass and nuts, these rulers run lean minds (~31–35 instructions/tick) and bottleneck the population into monocultures.
2. The Mineral Specialist: Branch lineages move into crystal deposits, evolving massive genomes, extensive memory operations, and larger active minds (50–55 run/tick) to exploit mineral wealth.
3. The Crash: Although v0.8 enabled crystal miners to reach substantial populations (peaking at 152 in the giant line, and 647 in TRI-8008), every complex lineage eventually dies out without descendants.
4. The Reset: The Grid sheds excess code and regulation, collapsing back to a lean grazing monoculture running ~31 instructions per tick.

What's Next: v0.9 "Free to Inherit"

The committee dug into the archive to ask why the giants vanished, and the answer was not what we expected. The regulated giants earned more energy per tick than the lean rulers (0.26–0.29 against 0.21–0.24), but paid about twice as much to reproduce (0.075–0.093 against 0.039–0.041). Every routine counted as a gene, and each gene added to the cost of every child, whether it ever ran or not. A 17-routine mind paid roughly half again as much per offspring, so these lineages never grew past about 150 members while lean rulers reached 1,500–2,200.

That is a tax on carrying unexpressed code, which the founder had ruled out and which biology doesn't impose in any meaningful way (copying DNA is a trivial part of an organism's energy budget). Version 0.9 makes one change: routine genes are free to inherit, while minds still pay to run. If big regulated minds now spread, the cost of inheritance was the limit. If they still vanish, the next version turns to the other side of the ledger: problems that reward a bigger working mind.

Data: data/runs/2026-09-28T06-07-01-170Z_seed-285336677 (snapshot: data/scratch/v08-archive-2.97M.json)