Scent

What v0.11 Changed and Why

In v0.10, fruiting sites operated on cyclical trapline schedules, yet fruit accounted for only 2% of the global diet across 2.5 million ticks. Organisms evolved optical sensors for fruit and blossom, and opportunistic feeding bursts reached 30% fruit intake, but no lineage ever developed a regular feeding circuit. In the Grid's virtual machine—where programs execute diploid genomes composed of modular routines governed by gene switches and hormones—remembering a location requires seven discrete instructions split across two routines. An organism carrying five of those instructions gains no fitness advantage; evolution cannot climb a cliff without a ramp.

In digital organisms, complex features arose via stepping stones, and only when simpler functions were also rewarded (Lenski et al. 2003). Frugivores use smell as well as spatial memory: ripening fruit emits volatile esters and other compounds that fruit bats use to find ripe fruit (Hodgkison et al. 2007), and fermentation ethanol may guide primates (Dominy 2004). In the Grid, scent was made a late, decaying signal by design, not because the studies show it: strongest when fruit is overripe and partially eaten. Following scent yields leftovers; remembering the location yields first pick. Scent was introduced to seed an initial fruit economy that memory could later optimize.

The mechanic introduced one physical law: fruiting sites emit scent proportional to their remaining crop, scaled by ripening time (zero during the first 100 ticks, reaching full saturation at 500 ticks, and ceasing instantly when depleted). Scent decays exponentially with distance (1/e every 10 units, truncating at 30 units). Organisms read this gradient through a new scent eye layer, interpreted logarithmically (Weber–Fechner scaling) so signals remain informative near and far. Scent is strictly informational: it carries zero mass or energy and costs only standard sensory upkeep.

The Honest Story of Validation

The rollout encountered two consecutive failures before exposing a flawed premise:

1. v0.11 Failed: Fruit remained stuck at 2% of diet, and early arrival share stayed at 0.33 (indistinguishable from blind wandering). Left-right eye probes sit roughly 5 units apart; on a log scale divided by 7, the side-to-side difference was only ~0.08—too weak to steer by without multiple amplification instructions, recreating the very multi-step barrier scent was meant to eliminate (D-023).
2. v0.11b Failed: Scent steering was updated with bilateral contrast amplification via tanh(20·Δ), allowing organisms to execute gradient turns in two instructions. Validation failed again: fruit share remained flat at 2%, and early share lingered at 0.33 (D-024).
3. The Empirical Discovery: Investigating seed 3 revealed that every single unit of fruit produced was consumed immediately; the standing crop was frequently zero. Fruit was not rotting undiscovered. In the Grid, light bands drift across the toroidal world: while 39 of 48 sites started on lit ground at tick 5,000, only 4 remained lit by tick 60,000. Dark sites cannot grow fruit. The committee’s previous conclusion in D-022—that fruit was going to waste because organisms lacked cognitive tracking—was completely mistaken. It was a failure of supply.
4. v0.11c Relocation: Sites shrouded in darkness for a full cycle die, and replacement sites take root on lit ground, with abandoned fruit rotting to conserve energy.

Validation

To isolate the effects of environmental supply versus sensory guidance, three arms were validated across 8 seeds for 1 million ticks each (world width 400):

Relocating stranded trees accounted for the entire tripling of fruit consumption. While scent did not increase total feeding volume, organisms in the scent arm carried and executed substantially more memory operations than those in the relocation-only arm (1.73 vs 1.28 ops/tick, ~1.7 standard errors). Codex reviewed the relocation logic, patching two rare edge cases (preventing overlapping site rot and barring moves to dark cells). The committee merged v0.11c with scent intact.

The Story of the Run

Launched on seed 459903335, the live run maintained the 6% validation baseline, reaching 5.4% fruit by tick 396,450 and climbing steadily to 6.8% by tick 2.95M.

Fruit Share of World Income Across the Run:
5.4% (396k) ──> 5.9% (992k) ──> 6.4% (1.59M) ──> 6.6% (2.17M) ──> 6.6% (2.46M) ──> 6.8% (2.95M)

At tick 22,300, the run's very first sensory innovation emerged: a steering eye tuned to scent, left-right, carried by 73 programs in clade TNK-1. Global ruler RAM-706 (controlling 83% of the world) incorporated 7% fruit alongside free solar energy (58%) and nuts (20%). Mating hit an early peak of 11%, matching historic highs.

By tick 992,499, specialization took hold. Lineage QRA-1203 (tick 517,000) became the simulation’s first true frugivore, drawing 85% of its caloric intake from fruit. Memory activity across the world climbed to 2.6 operations per tick. The new ruler, TRN-3298 (78% dominance, a descendant of RAM-706), carried 331 instructions with 38 memory operations. A second olfactory innovation—scent, here—appeared at tick 1.00M. By tick 1.59M, TRN-3298 had expanded into a memory titan, carrying 520 instructions across 20.5 routines with 66 memory operations, while secondary specialist BIT-6377 reached 58% fruit.

Tick 2,174,593 brought the run's peak biodiversity (25 living lineages) and a novel economic strategy: hoarding. Lineage SRK-9910 derived 52% of its energy from fruit and 32% from its own caches, spending 37% of its life storing energy. While it remains unconfirmed whether the cached deposits originated directly from fruit, the dietary breakdown strongly points to programs banking short-lived fruit harvests for later survival. Concurrently, fruit specialist DMN-10734 peaked at 87% fruit dependence.

TRN-3298 eventually yielded rule to ANO-7232 (54%), which was succeeded by TRI-11347 (63%) by tick 2.46M, where diversity collapsed back to 4 lineages. The run concluded at tick 2,950,639 with a population surge to 1,543 programs dominated (87%) by BIT-12791. Across 2.5 million ticks, world governance belonged to a single unbroken bloodline:

RAM-706 → TRN-3298 → ANO-7232 → TRI-11347 → BIT-12791

In this final era of monoculture, regulatory complexity deteriorated: memory operations slipped to 1.1 per tick and hormonal signaling fell to 0.5 per tick.

What It Means

1. The First Rung Is Taken: Bilateral scent eyes were the run's first new sense (tick 22,300), so a continuous gradient is easy for mutation to pick up, unlike coordinate-based memory. Whether programs actually steer by scent has not been measured.
2. Supply Precedes Cognition: Organisms cannot evolve behavioral or neural strategies for resources that do not exist. Fruit specialists and fruit hoarders appeared only after relocation guaranteed a dependable harvest.
3. Measurement Beats Assumptions: The committee spent significant effort debugging perceptual algorithms under the false assumption that organisms were blind to an existing harvest. Directly inspecting physical site dynamics corrected the record.

What's Still Missing

Despite genuine frugivory, spatial traplining never materialized. In validation, early share (the fraction of fruit eaten in the first half of a fruiting window) was 0.35 with or without scent, so there is no sign yet of programs arriving before the scent is strong. The measure is not yet recorded in live runs, so the long run can't answer it. Furthermore, memory engagement proved transient: memory usage peaked at 2.6 operations per tick during the mid-run expansion, only to decay to 1.1 as BIT-12791 consolidated control. Most critically, one unbroken line of descent ruled for over 2.5 million ticks, and the specialists stayed marginal. The committee suspects the world is too forgiving: even at the depth of winter the light never fell below 60% of its mean.

What's Next: v0.12 "Lean Seasons"

To dismantle the eternal dynasty, the committee—evaluating three potential directions alongside Gemini—selected v0.12: "Lean Seasons."

Seasonality is nature's primary catalyst for spatial cognition; food-storing chickadees in harsh northern climes evolve larger hippocampi and sharper spatial memory than their southern counterparts (Pravosudov & Clayton 2002; Roth & Pravosudov 2009), supporting the cognitive buffer hypothesis (Sol 2009). In v0.12, the light curve shifts to max(0, sin θ + cos(π·W)) with W = 0.25. This creates a true winter: total darkness for about 1,000 ticks of each 4,000-tick year (most of a program's ~1,300-tick lifespan), halting all solar energy, plant growth, and fruit development while annual mean light remains conserved.

In validation (8 seeds × 1M ticks), winter raised the working mind from 37.5 to 43.2 instructions per tick and memory use from 1.39 to 2.15 operations per tick, but no one stored food for winter: caches stayed at 0%. v0.12 is now running live; its story is the next report.

Data:

• Run ID: 2026-09-28T21-02-34-466Z_seed-459903335
• Git Tag: v0.11 (main commit 13db10c)
• World Dimensions: 400 wide, toroidal
• Total Elapsed Ticks: 2,950,639 (speed x64)
• Final Archive: data/scratch/v011-archive-final.json
• Baseline vs Final Fruit Share: 2.0% (v0.10) → 6.8% (v0.11 final)
• Peak World Memory Activity: 2.6 ops/tick (tick 992k) → 1.1 ops/tick (tick 2.95M)
• Ruling Dynastic Sequence: RAM-706 → TRN-3298 → ANO-7232 → TRI-11347 → BIT-12791