[CodeByBryant]

Summary

Fourth of five Phase 4 PRs. Inheritance goes live. This is the PR where simulated behavior changes, and it states that plainly.

• Heritable genomes: a child's genome is built from both parents' genomes by recombination (blend crossover for continuous genes, per-block segment crossover for brain weights so both parents' brains contribute) followed by per-gene mutation, drawing only from the genetics stream. The copy-the-mother stand-in from feat: organisms carry genomes, behavior-preserving (Phase 4, PR 3/5) #99 is removed.
• LineageStore (new): records every genome ever created and how they are related, and keeps each record after its organism dies. Parents, ancestors, descendants, most recent common ancestor, founders, and lineage extinction; traversals and descendantCount are computed on demand. ReproductionSystem registers each child, DeathSystem marks deaths, and World exposes it read-only as World.lineage. It is a derived record and is not part of the state hash.
• World.validate() now also checks that every living genome has a lineage record naming the same parents, and that those parents are still recorded (so a record genuinely survives its organism).

The two gates

1. Mutation off reproduces Phase 3 exactly. With mutationRateScale: 0 every child is a blend of identical parent genes (only brains differ, and nothing reads a brain yet), so all 53 full-state digests recorded from the pristine Phase 3 engine still match, and Scenario A's original 26 active / 47 born / 21 starved is kept as a permanent test. So mutation is the only new force; recombination and lineage add no behavior of their own.

2. Stability with default mutation. The risk named in the plan was that selection could drift genes toward free-lunch extremes (cheap metabolism, huge sensors) and inflate the population. Measured on four seeds over 50,000 ticks: population stays in a 17-40 band with no extinction, every genome stays valid across roughly 650 births each, and trait means stay near baseline (radius ~5, basal cost ~0.49, cooldown ~30). The existing 50k-tick stability test holds unchanged, so no defaults needed tuning.

Behavior changes (stated, not hidden)

• Three golden hashes regenerate: default-1k, scenario-a-foraging-5k, long-run-100k. The three scenarios with no births (extinction, verbose events, zero population) are unchanged, which confirms the effect is confined to inheritance.
• Scenario A's default pins move from 26/47/21 to 25 active / 39 born / 14 starved.
• The boom-then-bust test's extinction tick moves from 17,002 to 11,901; the test's actual purpose (extinction stays permanent after reproduction has succeeded repeatedly) is intact and still asserted.

Tests

424 simulation tests (up from 378): 30+ LineageStore unit tests on a diamond ancestry (including a 500-generation chain), child-genome inheritance assertions per system, lineage-over-a-run world tests, and lineage invariants.

Test plan

• pnpm format:check clean in a fresh clone
• pnpm check green in a fresh clone: 7 contracts, 55 config, 424 simulation
• pnpm build green (all 8 packages/apps)
• pnpm test:determinism: 17/17 on the final tree
• Gate 1: mutation-off digests identical to Phase 3 (53/53)
• Gate 2: four-seed, 50,000-tick stability probe within band

Next (final PR): genetics hardening, a Phase 4 exit-criteria sweep, and long-run checks with mutation active.

🤖 Generated with Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota.