EvoSphere

EvoSphere · Build dossier

A pass over the project’s own repository: what is implemented, what is partial, what stayed on paper.

A browser application without a bundler: ES modules, Canvas 2D across three layers and Howler 2.2.4 for audio. The output is a static index.html and index.ru.html; no GPU is required.

3 implemented · 2 not implemented · 1 partial

Concept illustration: EvoSphere. Not a product screenshot

What the code says

  • Rendering is three stacked Canvas 2D layers (background, world, ui), not WebGL

    Source js/renderer/core/CanvasManager.js

  • Start: 16 agents (4 families × 4), hard ceiling 100

    Source js/constants.js

  • Genome: ~45 scalar genes plus a tendency vector; mutation rate 0.055

    Source js/core/genome.js

  • 38 action types, 22 brain evaluators, 10 emotions, 10 global events, 3 diseases

    Source js/constants.js, js/entities/agentBrain.js

  • 43 audio files and a single runtime dependency, Howler 2.2.4

    Source assets/sounds/, package.json

  • Build: 42 files, 3.03 MB including a 1.56 MB sourcemap

    Source dist/

Module state

  1. implemented

    Genetics and inheritance

    A random genome, one parental value per gene and mutation; there is no crossover or gene duplication.

    Source js/core/genome.js

  2. implemented

    Evaluator-based agent brain

    Normalised needs (health, energy, food, oxygen) and a sorted action list; there is no neural network in the code.

    Source js/entities/agentBrain.js

  3. implemented

    Ecology: families, leaders, lairs, corpses, scavenging

    Leaders issue attack/defend/gather/explore commands; global events shift resources and health.

    Source js/sim_modules/

  4. not implemented

    Neural networks, state export, analytics, Electron

    README items with no matching code; only a results summary is persisted, in localStorage.

    Source README.md

  5. not implemented

    Automated tests at all

    The test script prints an error and exits non-zero; the CI test step is commented out.

    Source package.json, .github/workflows/ci.yml

  6. partial

    Bilingual entry: EN and RU

    Two entry pages are built; the project has no Arabic localisation.

    Source index.html, index.ru.html

How this was checked

  • The runtime makes no network calls

    No fetch, XMLHttpRequest or api_key occurrences were found under js/.

  • CI builds and publishes but does not test

    The workflow contains build and a GitHub Pages deploy; the test step remains a comment.

  • The population ceiling is a constant, not a setting

    MAX_POPULATION = 100 is fixed in js/constants.js.

    Source js/constants.js

Limits of these claims

  • The README guidance of "100–200 agents smoothly, 500+ needing a GPU" contradicts the hard ceiling of 100 agents and is unmeasured.
  • The "tested browsers" table is not backed by a single test run.
  • The scientific framing (epigenetic landscapes, game theory) is descriptive analogy: the code implements genes and selection, not those models.
  • The README JS API examples (a Simulation with environment parameters, extinction events) have no implementation.
  • The project is distributed under CC BY-NC 4.0: commercial-use rights stay with the copyright holder, that is, its author.
  • A stub pause path still sits in js/main.js next to the working handler.

Live build

Live buildevosphere.moonsugar.space ↗

Revision
41c9a6a
Build directory
dist

The shipped interface exposes no simulation parameter controls: start, pause, restart and result views are what is present. The population has a hard ceiling of 100 agents. Rendering is Canvas 2D, so no GPU is required.

The page itself does not confirm availability: the host is checked when the build is published.

Back to the project overview ↗