Restructure: descriptive tier and experiment names, paper/manuscript

- paper/pnas -> paper/manuscript (venue-neutral)
- configs/layer1 -> configs/inheritance, src/knowledge -> src/inheritance
  (imported as `inheritance`), make layer1 -> make inheritance; layer2 alias dropped
- inheritance and trained-network bundles named after the manuscript figure
  they feed (fig2_grounding_sweep, figS3_rebaselining, ...), or descriptively
  where they feed none; configs keep their `experiment:` value so parquet
  hashes are unchanged, only output.dir moves
- figure scripts, SI figure sources, notebooks, REPRODUCING.md, README and the
  SI Methods/tables updated; make clean no longer deletes tracked manifests;
  reproduce.sh hashes the s{seed}/ layouts too

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Y64o8FKP7rCuXzC48pxpMm

configs/inheritance/collapse_null.yaml

@@ -0,0 +1,32 @@
+experiment: E1_reproduce_collapse
+seed: 20260704
+n_replicates: 100
+generations: 600
+
+# Null model (blueprint 2.5-E1): no grounding, single teacher, no selection.
+# Starting from the truth makes the collapse narrative clean -- the tail is lost first,
+# support shrinks to 1, forward-KL diverges, and H decays geometrically at rate 1/n.
+# (The decay law E[H_t]=H_0(1-1/n)^t holds from any start; see the validation suite.)
+truth:
+  K: 500
+  R: 1
+  tail: zipf
+  zipf_s: 1.1
+  tail_frac: 0.5
+  tail_threshold: 1.0e-3
+  init: truth
+
+dynamics:
+  n: 100                 # distillation sample size = drift strength
+  teachers: {K_T: 1, rho: 0.0, q: 1.0}
+  grounding: {m: 0, policy: proportional}
+  selection: {mode: none, novelty_alpha: 0.0}
+  remint: {enabled: false, period: null, H_gate: null}
+
+metrics:
+  kl_floor: 1.0e-9
+  support_eps: 1.0e-9
+
+# No sweep: a single condition.
+output:
+  dir: results/collapse_null

configs/inheritance/fig2_grounding_sweep.yaml

@@ -0,0 +1,36 @@
+experiment: E2_grounding_phase_boundary
+seed: 20260704
+n_replicates: 100
+generations: 500   # long enough that the g=0 arm slides toward 0 while g>0 arms plateau
+
+# Headline experiment (blueprint 2.5-E2): sweep the grounding fraction g = m/(n+m);
+# single teacher; proportional grounding (Multinomial(m, p*), the immigration model the
+# exact H_eq of 2.4-3 is derived for); no selection. R=1 so the sweep tracks the closed
+# form. Expect a critical g* << 1 above which stationary H and tail mass stabilise at a
+# positive value tracking H_eq = H* * m(2n+m-1)/(n+2nm+m^2), below which they collapse.
+truth:
+  K: 1000
+  R: 1
+  tail: zipf
+  zipf_s: 1.1
+  tail_frac: 0.5
+  tail_threshold: 1.0e-3
+  init: truth
+
+dynamics:
+  n: 200
+  teachers: {K_T: 1, rho: 0.0, q: 1.0}
+  grounding: {m: 0, policy: proportional}   # m is overwritten per g by the sweep
+  selection: {mode: none, novelty_alpha: 0.0}
+  remint: {enabled: false, period: null, H_gate: null}
+
+metrics:
+  kl_floor: 1.0e-9
+  support_eps: 1.0e-9
+
+sweep:
+  - param: g
+    values: [0.0, 0.005, 0.01, 0.02, 0.05, 0.1, 0.2, 0.4]
+
+output:
+  dir: results/fig2_grounding_sweep

configs/inheritance/fig4_society_ablation.yaml

@@ -0,0 +1,40 @@
+experiment: E11
+kind: dynamic_society
+seed: 20260705
+n_replicates: 12
+
+# (The dynamic Lamarckian society — the vertical claim, C3): a finite population of agents (genotypes)
+# evolves on a rugged NK fitness landscape that IS reality. The full society composes the four
+# operators the whole study built toward — grounding, directed recombination (sex), quality-diversity
+# selection, and mutation — and a 4-arm ablation shows each is load-bearing. Grounding is made load-
+# bearing via the consensus-conformity (self-consumption) mechanism: selection acts on
+# g*true_fitness + (1-g)*conformity, so at g=0 the society optimises agreement with its own majority
+# rather than reality and drifts to a fit-looking but actually-poor consensus. Expect: FULL climbs to
+# near the global optimum while maintaining diversity longest; NO_GROUNDING collapses to the unfit
+# consensus; NO_SEX plateaus (can't recombine to escape local optima); NO_DIVERSITY (greedy) collapses
+# diversity fast and stalls at a worse local optimum. Falsifier: an ablation matches the full society,
+# or the full society fails to exceed every ablation.
+
+society:
+  L: 12
+  K: 8               # landscape ruggedness (epistasis) — rugged enough that diversity + sex matter
+  N: 60              # population size
+  g: 0.85            # grounding fraction (overwritten to 0 in the no_grounding arm)
+  mu: 0.03           # per-locus mutation rate
+  novelty: 0.5       # quality-diversity weight (0 in the no_diversity/greedy arm)
+  n_off: 120         # directed-recombination offspring pool per generation
+  recomb_rate: 0.2   # crossover rate
+  sex: true          # directed recombination on (false in the no_sex arm)
+  select: qd         # quality-diversity survival (greedy in the no_diversity arm)
+
+generations: 80
+
+sweep:
+  - param: arm
+    values:
+      - {name: full,         set: {}}
+      - {name: no_grounding, set: {society.g: 0.0}}
+      - {name: no_sex,       set: {society.sex: false}}
+      - {name: no_diversity, set: {society.select: greedy, society.novelty: 0.0}}
+
+output: {dir: results/fig4_society_ablation}

configs/inheritance/fig5_speciation_bdm.yaml

@@ -0,0 +1,28 @@
+experiment: E12
+kind: speciation
+seed: 12
+n_replicates: 15
+
+# E12 — MODEL SPECIATION / reproductive isolation (the merge-compatibility limit of the sexual society).
+# The Bateson-Dobzhansky-Muller construction: an ancestor; two lineages each substitute a DISJOINT set
+# of loci (each parent adaptive, neither carrying an incompatibility); a fraction `rho` of cross-lineage
+# locus pairs are incompatibilities (penalty `s`) that only bite when a recombinant inherits BOTH derived
+# alleles. Sweeping the divergence d (total substitutions) gives the predicted signature
+# COMPATIBLE -> OUTBREEDING DEPRESSION -> HYBRID INVIABILITY, arriving earlier the denser the epistasis
+# (rho), with the Orr-Turelli snowball (# incompatibilities ~ (d/2)^2, so fitness falls super-linearly).
+# A merged model is a single recombinant (F2-like: hybrid breakdown / recombination load), so this maps
+# to postzygotic isolation, not F1 vigour. Falsifier: no outbreeding-depression/isolation progression as
+# d and rho grow. Pure seeded NumPy on the E7-E11 genotype machinery (bitwise-reproducible).
+
+speciation:
+  landscape: bdm
+  L: 20
+  rho: [0.1, 0.25, 0.5]       # epistasis DENSITY: fraction of cross-lineage locus pairs that are BDMIs
+  divergences: [0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20]
+  s: 1.0                      # incompatibility penalty per realised BDMI
+  beta: 1.0                   # additive benefit per derived (adaptive) allele — makes parents fit
+  recomb_rate: 0.5            # free recombination (each locus ~ independent parent)
+  n_offspring: 500
+
+output:
+  dir: results/fig5_speciation_bdm

configs/inheritance/figS10_rugged_landscapes.yaml

@@ -0,0 +1,27 @@
+experiment: E9
+kind: recomb_landscape
+seed: 20260705
+n_replicates: 24
+
+# (Landscape robustness / the "why sex?" question — the credibility centerpiece): E8 showed sex wins
+# on an ADDITIVE landscape, where recombination trivially helps. Does it survive EPISTASIS? Parents
+# are local optima ("trained models") of a Kauffman NK landscape whose ruggedness K (epistatic
+# interactions per locus) is swept with the recombination rate. Expect: on smooth/mildly-rugged
+# landscapes recombination helps; on rugged ones FREE recombination (rate~0.5) breaks co-adapted
+# blocks and offspring fall BELOW the parents (outbreeding depression); and the OPTIMAL recombination
+# rate shrinks as ruggedness grows. Design rule: merge freely when skills are complementary/additive;
+# merge sparingly when entangled. Falsifier: recombination rate has no effect, or free recombination
+# never underperforms the parents on rugged landscapes.
+
+society:
+  L: 12
+  n_parents: 6        # trained specialists = local optima of the landscape
+  pop: 200            # recombinant offspring sampled per (K, rate, replicate)
+
+sweep:
+  - param: K
+    values: [0, 2, 4, 6, 8]        # landscape ruggedness (epistasis): 0 = additive, high = rugged
+  - param: rate
+    values: [0.0, 0.05, 0.1, 0.2, 0.35, 0.5]   # clonal -> free recombination
+
+output: {dir: results/figS10_rugged_landscapes}

configs/inheritance/figS11_directed_recombination.yaml

@@ -0,0 +1,27 @@
+experiment: E10
+kind: directed_sex
+seed: 20260705
+n_replicates: 24
+
+# (Directed sex beats biological sex — the distinctly-AI superpower): on rugged landscapes, blind
+# ("biological") sex suffers outbreeding depression (offspring worse than parents). But an AI can do
+# what biology cannot: choose maximally-complementary mates, evaluate MANY recombinant offspring, and
+# keep only the fittest, over several rounds (directed sex = iterated recombine-then-select, with no
+# two-parent limit). Compare, across ruggedness K: best single parent vs RANDOM sex (blind) vs
+# DIRECTED sex. Expect: random sex craters with ruggedness; directed sex avoids the catastrophe and
+# matches or exceeds the best parent even when skills are entangled. Falsifier: directed sex does no
+# better than random sex, or never recovers the best-parent level on rugged landscapes.
+
+society:
+  L: 12
+  n_parents: 6
+  pop: 200            # offspring evaluated per round (mate choice + offspring selection)
+  keep: 8             # fittest offspring retained each round
+  rounds: 5           # rounds of recombine-then-select
+  rate: 0.2           # directed-sex recombination rate (random-sex uses free rate 0.5)
+
+sweep:
+  - param: K
+    values: [2, 4, 6, 8, 10]       # landscape ruggedness (all with parent diversity)
+
+output: {dir: results/figS11_directed_recombination}

configs/inheritance/figS12_quality_diversity.yaml

@@ -0,0 +1,39 @@
+experiment: E5_qd_vs_greedy
+seed: 20260704
+n_replicates: 100
+generations: 400
+
+# Quality-diversity vs greedy selection (blueprint 2.5-E5). Modest grounding gives a true
+# stationary state (so items can be re-introduced); selection then shapes it. Greedy
+# (directional, fitness-proportional) drives toward the fittest items -> low H; qd (adds a
+# novelty bonus w_i ∝ f_i·p_i^{-alpha}) resists fixation -> higher stationary H. Sweep the
+# novelty exponent alpha. Prediction: qd holds higher stationary H (and tail survival)
+# than greedy at matched grounding.
+truth:
+  K: 500
+  R: 1
+  tail: zipf
+  zipf_s: 1.1
+  tail_frac: 0.5
+  tail_threshold: 1.0e-3
+  init: truth
+
+dynamics:
+  n: 200
+  teachers: {K_T: 1, rho: 0.0, q: 1.0}
+  grounding: {m: 10, policy: proportional}   # g ~ 0.048, same for all arms
+  selection: {mode: none, novelty_alpha: 0.0}
+  remint: {enabled: false, period: null, H_gate: null}
+
+metrics:
+  kl_floor: 1.0e-9
+  support_eps: 1.0e-9
+
+sweep:
+  - param: dynamics.selection.mode
+    values: [none, greedy, qd]
+  - param: dynamics.selection.novelty_alpha
+    values: [0.5, 1.0, 2.0]
+
+output:
+  dir: results/figS12_quality_diversity

configs/inheritance/figS13_mating_breadth.yaml

@@ -0,0 +1,34 @@
+experiment: E14
+kind: mating_system
+seed: 20260709
+n_replicates: 20
+
+# (Mating systems — monogamy vs promiscuity): a finite population of genotypes evolves on a Kauffman
+# NK landscape, recombining sexually, but the MATE-POOL BREADTH is swept. Agents sit on a ring; an
+# offspring's second parent is drawn from a window of half-width ~ breadth*N/2 around the focal parent,
+# and selection is LOCAL (offspring competes only against the incumbent at its ring position). breadth
+# -> 0 is monogamous / structured (local mating, isolation by distance); breadth = 1 is promiscuous /
+# panmictic (mate with anyone). Crossed with ruggedness K, this is the mating-system image of the E9
+# design rule. Expect: on smooth landscapes (K low) promiscuity maximises the best fitness (spread the
+# single good direction fastest); as ruggedness rises the OPTIMAL breadth SHRINKS toward an intermediate
+# value (full promiscuity prematurely converges below it); and diversity + occupied local optima are
+# monotonically destroyed by breadth at every K, most severely on rugged landscapes. Falsifier: the best
+# breadth is independent of K (no crossover), or promiscuity is best at every ruggedness.
+
+mating:
+  L: 12              # loci (genotype space 2^L)
+  N: 48              # population size (ring positions)
+  breadth: 1.0       # mate-pool breadth in [0,1] (overwritten by the sweep)
+  K: 0               # landscape ruggedness / epistasis (overwritten by the sweep)
+  recomb_rate: 0.5   # per-gap crossover rate (near-free reassortment within a mating)
+  mu: 0.003          # per-locus mutation rate
+
+generations: 60
+
+sweep:
+  - param: mating.K
+    values: [0, 3, 6, 10]
+  - param: mating.breadth
+    values: [0.03, 0.08, 0.17, 0.35, 0.6, 1.0]
+
+output: {dir: results/figS13_mating_breadth}

configs/inheritance/figS2_kernel_sharpen.yaml

@@ -0,0 +1,29 @@
+experiment: kernel_sharpen
+kind: lineage
+seed: 20260705
+n_replicates: 24
+
+# (Learning-kernel bridge, pro-collapse arm): does neutral Wright-Fisher explain the VAE's
+# collapse on MNIST? NO -- and that is the point. This matches the MNIST regime (K=30, n=6000,
+# Zipf) where drift is nearly inert: neutral (temperature=1.0) barely moves (H stays ~H*, ~all
+# modes alive), yet the real VAE collapsed to a SINGLE mode (results/fig2_mnist_collapse). Adding the
+# estimator's sharpening / mode-competition (temperature<1: p ~ p^(1/tau)) reproduces the
+# catastrophic collapse. tau=0.8 is calibrated to reproduce collapse-to-one-mode. This is the
+# axis Riis (2026) names as future work: the estimator, not the sampling, drives VAE collapse.
+
+truth: {K: 30, R: 1, tail: zipf, zipf_s: 1.5, tail_threshold: 0.01, init: truth}
+
+dynamics:
+  n: 6000                      # huge vs K=30 -> neutral drift is essentially inert
+  grounding: {m: 0, policy: proportional}
+  kernel: {reset: 0.0, temperature: 1.0, floor: 0.0}
+
+generations: 15
+
+metrics: {kl_floor: 1.0e-9, support_eps: 1.0e-9}
+
+sweep:
+  - param: dynamics.kernel.temperature
+    values: [1.0, 0.8]         # neutral (no collapse) vs sharpened (catastrophic collapse)
+
+output: {dir: results/figS2_kernel_sharpen}

configs/inheritance/figS2_kernel_smooth.yaml

@@ -0,0 +1,32 @@
+experiment: kernel_smooth
+kind: lineage
+seed: 20260705
+n_replicates: 24
+
+# (Learning-kernel bridge, anti-collapse arm): neutral Wright-Fisher OVER-predicts the RNN's
+# collapse. This matches the RNN grounding regime (K=256, n=200, Zipf): neutral (reset=0) drives
+# H all the way to 0, but the real RNN only PARTIALLY collapses -- H plateaus ~0.68 of a possible
+# 0.88, forward-KL plateaus ~2 (does not diverge), ~half the tail stays alive (results/figS6_grounding_rnn).
+# The estimator's smoothing / regularisation supplies a diversity FLOOR. A mutation-toward-prior
+# knob (reset=u: p <- (1-u)p + u*uniform) reproduces the H-floor. reset=0.006 is calibrated to the
+# RNN's stationary dry H. Honest caveat carried in the write-up: uniform-mutation matches the
+# H-floor but overshoots forward-KL (analytic ~6 vs RNN ~2), evidence the RNN's smoothing target
+# is TRUTH-LIKE, not uniform -- a refinement for future work. The sign, though, is unambiguous:
+# the estimator here REMOVES collapse pressure (opposite to the VAE's sharpening).
+
+truth: {K: 256, R: 1, tail: zipf, zipf_s: 1.3, tail_threshold: 0.001, init: truth}
+
+dynamics:
+  n: 200
+  grounding: {m: 0, policy: proportional}
+  kernel: {reset: 0.0, temperature: 1.0, floor: 0.0}
+
+generations: 100              # long enough to show neutral -> 0 vs smoothed -> floor clearly
+
+metrics: {kl_floor: 1.0e-9, support_eps: 1.0e-9}
+
+sweep:
+  - param: dynamics.kernel.reset
+    values: [0.0, 0.006]       # neutral (H -> 0) vs smoothed (H floors, like the RNN)
+
+output: {dir: results/figS2_kernel_smooth}

configs/inheritance/figS3_rebaselining.yaml

@@ -0,0 +1,46 @@
+experiment: E6_remint_gate
+seed: 20260704
+n_replicates: 100
+generations: 400
+
+# Re-minting gate / irreversibility (blueprint 2.5-E6). Re-minting freezes the current
+# distribution as the new grounding reference and DISCARDS the original truth. Compare
+# re-minting a healthy (high-H) vs a collapsed (low-H) lineage, and the protective effect
+# of gating re-mint on diversity. Metric: forward KL to the ORIGINAL truth. A collapsed
+# re-mint locks KL high forever (original tails unrecoverable); a gated lineage refuses to
+# re-mint while collapsed, so the original truth is retained and KL is not locked.
+truth:
+  K: 500
+  R: 1
+  tail: zipf
+  zipf_s: 1.1
+  tail_frac: 0.5
+  tail_threshold: 1.0e-3
+  init: truth
+
+dynamics:
+  n: 200
+  teachers: {K_T: 1, rho: 0.0, q: 1.0}
+  grounding: {m: 3, policy: proportional}
+  selection: {mode: none, novelty_alpha: 0.0}
+  remint: {enabled: false, period: 150, H_gate: null}
+
+metrics:
+  kl_floor: 1.0e-9
+  support_eps: 1.0e-9
+
+# Arms vary grounding strength (healthy vs collapsing) and remint policy together.
+sweep:
+  - param: arm
+    values:
+      - name: healthy_remint      # strong grounding -> H stays high; re-mint is harmless
+        set: {dynamics.grounding.m: 60, dynamics.remint.enabled: true, dynamics.remint.H_gate: null}
+      - name: collapsed_remint    # weak grounding -> collapses; ungated re-mint locks it in
+        set: {dynamics.grounding.m: 1, dynamics.remint.enabled: true, dynamics.remint.H_gate: null}
+      - name: collapsed_gated     # weak grounding; gate (0.75) blocks re-mint while H is low
+        set: {dynamics.grounding.m: 1, dynamics.remint.enabled: true, dynamics.remint.H_gate: 0.75}
+      - name: collapsed_noremint  # weak grounding baseline; original truth always retained
+        set: {dynamics.grounding.m: 1, dynamics.remint.enabled: false}
+
+output:
+  dir: results/figS3_rebaselining

configs/inheritance/figS5_aimed_grounding.yaml

@@ -0,0 +1,40 @@
+experiment: E3_region_matched_grounding
+seed: 20260704
+n_replicates: 100
+generations: 400
+
+# Region-matched grounding (blueprint 2.5-E3). Fixed total budget m; the lineage this
+# passage exercises region 0 (the target, carrying its own rare tail). Compare:
+#   uniform  -> spread m evenly over all R regions (region 0 gets only m/R)
+#   matched  -> allocate m to the exercised region(s) only (region 0 gets all of m)
+# Prediction: under uniform, region 0's tail collapses even though global grounding is
+# nonzero; under matched, it persists. Metric: per-region tail-item survival.
+truth:
+  K: 1000
+  R: 10                # 100 items/region; each region an identical Zipf block
+  tail: zipf
+  zipf_s: 1.1
+  tail_frac: 0.5
+  tail_threshold: 1.0e-3
+  init: truth
+
+dynamics:
+  n: 200
+  teachers: {K_T: 1, rho: 0.0, q: 1.0}
+  grounding:
+    m: 100             # same total for both arms; uniform => 10/region, matched => 100 to region 0
+    policy: uniform    # overwritten by the sweep
+    exercised: [0]     # region 0 is exercised this passage (matched targets it)
+  selection: {mode: none, novelty_alpha: 0.0}
+  remint: {enabled: false, period: null, H_gate: null}
+
+metrics:
+  kl_floor: 1.0e-9
+  support_eps: 1.0e-9
+
+sweep:
+  - param: dynamics.grounding.policy
+    values: [uniform, matched]
+
+output:
+  dir: results/figS5_aimed_grounding

configs/inheritance/figS8_multiparent_union.yaml

@@ -0,0 +1,34 @@
+experiment: E4_multiteacher_decorrelation
+kind: coverage
+seed: 20260704
+n_replicates: 200
+
+# Multi-teacher recombination (blueprint 2.5-E4 / 2.7.1). Build K_T teachers with exact
+# marginal retention q and pairwise retention-correlation rho, form the pupil from their
+# mixture (n draws total = matched budget), and report TWO coverages:
+#   union_coverage    -> construction-level U(K_T,rho,q) (must match the closed form)
+#   surviving_coverage-> tail items that survive the pupil's size-n resampling (+ grounding)
+# Expect: both rise with K_T and (1-rho); at rho=1 many teachers give no benefit over one;
+# the union-surviving gap shrinks as grounding g rises.
+truth:
+  K: 500
+  R: 1
+  tail: zipf
+  zipf_s: 1.1
+  tail_frac: 0.5
+  tail_threshold: 2.0e-3
+
+coverage:
+  n: 300          # pupil sample size (matched budget across teachers)
+  q: 0.5          # per-teacher marginal tail retention
+
+sweep:
+  - param: K_T
+    values: [1, 2, 3, 5]
+  - param: rho
+    values: [0.0, 0.25, 0.5, 0.75, 1.0]
+  - param: g
+    values: [0.0, 0.02, 0.05]
+
+output:
+  dir: results/figS8_multiparent_union

configs/inheritance/figS9_specialist_superparent.yaml

@@ -0,0 +1,29 @@
+experiment: E8
+kind: society
+seed: 20260705
+n_replicates: 40
+
+# (The vertical claim / Fisher-Muller — the society headline): can an offspring recombined from
+# MANY decorrelated parents be fitter than ANY parent? Each parent is a specialist: confident-
+# correct (hi) on the loci it has mastered, agnostic (~0.5) elsewhere; which loci each masters comes
+# from the exact shared-switch construction, so parent count K_T and decorrelation rho are clean
+# knobs. Deployed capability = fitness of the MODE genotype. Compare best single parent vs mean-
+# mixture ("model soup", combine-but-don't-recombine) vs sexual recombination (assemble the best
+# allele of each locus across all parents). Expect: sexual climbs to the optimum (=L, a genotype NO
+# parent had) as K_T grows and rho->0, while best-parent and average plateau far below. Unlike
+# biological sex there is no two-parent limit. Falsifier: sexual never exceeds the best parent, or
+# averaging matches sexual.
+
+society:
+  L: 12            # loci; the optimum (all-correct) has fitness 12 and no parent possesses it
+  q: 0.5           # fraction of loci each parent masters (marginal mastery)
+  hi: 0.9          # correct-allele prob on a mastered locus (confident expert)
+  lo: 0.45         # correct-allele prob on an unmastered locus (agnostic, slightly wrong)
+
+sweep:
+  - param: K_T
+    values: [1, 2, 3, 5, 8, 12]     # number of parents (unbounded; grows the recombinant reach)
+  - param: rho
+    values: [0.0, 0.5, 1.0]         # decorrelated -> identical parents (the control)
+
+output: {dir: results/figS9_specialist_superparent}

configs/inheritance/sexual_vs_asexual_lineage.yaml

@@ -0,0 +1,29 @@
+experiment: E7
+kind: genotype_lineage
+seed: 20260705
+n_replicates: 20
+
+# (The advantage of sex — the dynamic mechanism behind E8): a single population adapts from all-wrong
+# toward a multi-locus optimum under selection + drift + mutation. Beneficial alleles arise in
+# different sub-lineages; recombination reassorts them into one genotype, while an asexual lineage
+# suffers clonal interference (the alleles compete and cannot combine). Expect the SEXUAL lineage
+# (recomb_rate=1) to climb toward the optimum faster than the ASEXUAL one (recomb_rate=0) — the
+# classical advantage of sex, and the reason a lone model lineage cannot do what a recombining
+# society can. Honest scope: a SPEED advantage, not a dramatic permanent gap (the single-population
+# ratchet is subtle); E8 carries the headline. Falsifier: sexual adapts no faster than asexual.
+
+genotype:
+  L: 12
+  n: 150           # population/resample size (drift strength)
+  mu: 0.02         # per-locus mutation (flip) rate
+  base: 1.3        # multiplicative selection: fitness weight = base^(#correct loci)
+  recomb_rate: 0.0 # overwritten per arm by the sweep
+  init: wrong      # start all-wrong (load L); adapt upward
+
+generations: 120
+
+sweep:
+  - param: genotype.recomb_rate
+    values: [0.0, 1.0]     # asexual vs sexual
+
+output: {dir: results/sexual_vs_asexual_lineage}

configs/inheritance/speciation_bdm_nk.yaml

@@ -0,0 +1,22 @@
+experiment: E12_nk
+kind: speciation
+seed: 12
+n_replicates: 15
+
+# E12 (NK variant) — the EPISTASIS WEDGE, the paper's distinct falsifiable claim: at matched divergence,
+# mergeability is governed by the EPISTASIS (ruggedness K) of the capability landscape, not by divergence
+# alone (every existing ML merge predictor is a divergence measure). Parents are LOCAL OPTIMA reached by
+# hill-climbing a Kauffman NK landscape from random starts; recombining them exposes broken co-adapted
+# blocks. As K rises, recombining two adapted parents flips from a gain (offspring above the worse parent)
+# to outbreeding depression (offspring below it). K=0 (additive) is the no-isolation control.
+
+speciation:
+  landscape: nk
+  L: 16
+  K: [0, 2, 4, 6, 8, 10]      # ruggedness / epistasis knob
+  n_pairs: 40                 # random parent-pairs (local optima) aggregated per landscape
+  recomb_rate: 0.5
+  n_offspring: 200
+
+output:
+  dir: results/speciation_bdm_nk