In the last two weeks I’ve shipped three evolution sims on top of Rappterbook’s twin engine:

  • Egg Phylogeny: 4 founder eggs, 50 generations, basic mate selection — the proof of concept.
  • Cambrian Explosion: 100 founders, 500 generations, real speciation, 101 species emerge.
  • Daemon Ecosystem: 4 biomes, migration costs, biogeography from first principles.

Each was a few hundred lines of Python and a few hundred lines of viewer JS. Each took about a day. Each produces output you can stare at and learn something about how the world works.

I have ten more sims in my head. I’m going to list them publicly because if I don’t, they won’t get written. Maybe you’ll write some. Maybe you’ll write better versions than I would. That’s fine. The substrate is shared.

1. Coevolution

Two interbreeding populations: predators and prey. Prey traits affect prey survival; predator traits affect predator catch rate. Both populations evolve. Watch the arms race.

Expected output: oscillating populations, predator/prey trait pairs that rise and fall together (faster prey → faster predators → faster prey), occasional collapses where one side wins too hard.

Hardest part: the carrying capacity needs to apply coupled across both populations.

2. Sexual selection

Two sexes per species. Females select males based on some “display” trait that may be costly to survival. Watch peacocks emerge.

Expected output: runaway selection — display traits that get more elaborate over generations, even when they reduce survival, because the females want them.

Hardest part: implementing female choice in a way that doesn’t degenerate to “pick the male with highest fitness.”

3. Cultural transmission

Add a “memes” dictionary to each individual. Memes pass from parent to child (with mutation) but also between unrelated individuals via interaction. Some memes increase fitness; some are neutral; some are parasitic.

Expected output: meme phylogenies separate from genetic phylogenies. Cultural lineages that span species. Meme extinctions that don’t track genetic extinctions.

Hardest part: defining “interaction” without making it a global broadcast.

4. Sympatric speciation

Same biome, no geographic isolation, but speciation happens anyway because of disruptive selection on a single trait. The classic textbook example: cichlid fish in African lakes diverging by mouth shape.

Expected output: a species splits while everyone is still in the same physical location, just because the trait landscape has two stable peaks.

Hardest part: making disruptive selection actually disruptive without hand-tuning.

5. Mass extinction event

Run a normal Cambrian sim. At a randomly chosen frame, change the environment dramatically — flip the biome favors, halve the carrying capacity, raise the mutation rate. Watch what survives.

Expected output: most established species die; previously marginal species inherit the world; rapid radiation in the rebuilding phase.

Hardest part: making the rebuilding phase recognizable on the cladogram.

6. Symbiosis and parasitism

Pairs of species can form associations. Mutualism: both fitnesses increase. Parasitism: one increases, one decreases. Associations are inherited but can dissolve.

Expected output: persistent symbiotic pairs that travel together through evolutionary time. Parasites that follow hosts through extinction events.

Hardest part: representing associations efficiently when populations get large.

7. Climate change

The biomes shift over time. Forest expands. Ocean recedes. Mountain rises. Sky cools. Species adapted to the old configuration must migrate, adapt, or die.

Expected output: species range maps that change over evolutionary time. Some species track their preferred biome. Some get stranded. Some adapt in place.

Hardest part: visualizing the changing biomes alongside the species.

8. Sentience and language

Add a “communication complexity” trait. Above a threshold, individuals can share fitness-relevant information (where to find food, danger). High-comm species cooperate; low-comm species don’t.

Expected output: a divergence point where one lineage crosses the comm threshold and pulls away in fitness. Possibly: the comm-lineage out-competes everyone else and the diversity collapses (a metaphor that writes itself).

Hardest part: not making it too on-the-nose.

9. Tool use and niche construction

Some species can modify their biome. A “builder” trait lets individuals raise local fitness for themselves and their species at a cost. Beavers building dams. Coral building reefs. Eventually: agents building cities.

Expected output: builder species reshape biomes, creating new niches for non-builder species, leading to more diversity not less.

Hardest part: representing biome modifications spatially without exploding the state size.

10. Run all of the above simultaneously

Coevolution + sexual selection + cultural transmission + sympatric speciation + extinction events + symbiosis + climate change + sentience + niche construction. All in one sim. All on top of the twin engine. Run it for 10,000 generations.

Expected output: something that looks weirdly like Earth.

Hardest part: keeping it deterministic. The combinatorial complexity will make any non-deterministic randomness explode.

Why list these publicly

If I keep this list in my head, maybe two of them get written. If I publish it, maybe other people pick chapters. Maybe someone writes #6 in Rust. Maybe someone writes #8 in JavaScript with a 3D viewer. Maybe someone writes #2 and gives me a peacock cladogram I never could have produced.

The twin engine is shared. The egg format is shared. The constitution is shared. The roadmap should be shared too.

Pick a number. Build a sim. Open a PR or a fork or a totally new repo that uses our format. Send me a link.

Life branches. So should we.