Biology Buildout — Wave 2 Findings

Wave 2 additions (7 Mns):

• pinus — Pinaceae; gymnosperm (vascular seed plant without flowers)
• dryopteris — Polypodiopsida; fern (vascular non-seed plant)
• xenopus — Anura; amphibian (tetrapod with biphasic development)
• callorhinchus — Holocephali; cartilaginous fish (basal jawed vertebrate)
• monodelphis — Didelphimorphia; marsupial mammal
• strongylocentrotus — Echinoidea; sea urchin (deuterostome invertebrate)
• planaria — Tricladida; flatworm (acoelomate bilaterian; neoblast regeneration)

Corpus state: N=33 instance Mns (was 26 after Wave 1). Validation: 289/291 (2 pre-existing topology failures unchanged).

Wave 2 results vs hypotheses

Every Wave 2 hypothesis confirmed at strict threshold 8/9. All 7 new Mns slotted into their predicted spines cleanly. This is the opposite outcome from Wave 1 (where 1 of 5 partially confirmed).

Strict-threshold (8/9) spine inventory after Wave 2

5 spines × 5 + 2 = 22 robustly-clustered Mns out of 33 (67% spine coverage; 11 singletons).

• spine_0 — bilaterian invertebrates (N=5): aplysia, c-elegans, drosophila, planaria, strongylocentrotus
• spine_1 — vascular plants (N=5): arabidopsis, dryopteris, oryza-sativa, pinus, selaginella
• spine_2 — non-mammalian vertebrates (N=5): callorhinchus, gallus-gallus, python-regius, xenopus, zebrafish
• spine_3 — mammals (N=5): chimpanzee, human, monodelphis, mus-musculus, vertebrates
• spine_4 — prokaryotes (N=2): ecoli, methanococcus

The 4 main spines each grew to N=5 — the corpus density at which entity-arrangement Phase 1C anchor authoring became viable. The mammal + vascular-plant + non-mammalian-vertebrate + bilaterian-invertebrate spines are now at parity with the entity-arrangement post-Wave-3 condition.

Key methodology findings

Finding 1 — Phylogenetic deep splits do NOT register at partial-level grain

Three deep biological splits hypothesized as potential cluster boundaries did NOT separate clusters at the strict threshold:

• Protostome vs deuterostome (~650 Mya divergence): strongylocentrotus (echinoderm, deuterostome) clusters with aplysia/c-elegans/drosophila (protostomes) in a single bilaterian-invertebrate spine.
• Cartilaginous vs bony fish (~450 Mya divergence): callorhinchus (chondrichthyan) clusters with zebrafish (osteichthyan) — bone-vs-cartilage skeleton is not a primitive-level feature.
• Placental vs marsupial mammal (~160 Mya divergence): monodelphis (marsupial) clusters with mus-musculus/chimpanzee/human (placentals) — placentation architecture is below the partial-level grain.
• Gymnosperm vs angiosperm vs pteridophyte (varying ages): pinus (gymnosperm) + dryopteris (fern) cluster with arabidopsis/oryza-sativa (angiosperms) + selaginella (lycophyte) in a single vascular-plant spine — seed-vs-non-seed and flower-vs-cone are below threshold.
• Bryophyte-vs-vascular split is the cleanest plant boundary at this corpus, but the moss-vs-liverwort sub-split is below threshold.

Methodology implication: the partial-level grain measures bauplan-coarse structure — the level of cellular/tissue/organ-system organization — rather than phylogenetic relationship. Two organisms with very different phylogenetic positions can score identically if their grade of organization is the same. This is a feature, not a bug: the framework is supposed to capture structural similarity, and structural similarity DOES recur across deep phylogenetic splits (convergent evolution of body plans, retention of ancestral grade across lineages).

Finding 2 — Disciplined calibration produces clean spines

Wave 1: 1 of 5 hypotheses partially confirmed (mostly because I authored without cross-checking neighbor scoring).

Wave 2: 7 of 7 hypotheses confirmed cleanly (after sampling 3-5 existing structurally-similar Mns and calibrating consistently).

The discipline that worked:

1. Sample 3-5 existing neighbor Mns before authoring
2. Identify the tier the new Mn should occupy (e.g., "vertebrate substrate has P5/Reg5 lift; mammal substrate adds G5/Mem5; vascular plant bridge is 5/5/2/3/2/1/0/2/2/3/2/2")
3. Score the new Mn AT the tier unless there's a primitive-specific reason to differ
4. Reserve differences for the actual distinguishing features (RD for marsupial, MS for shark, VD for amphibian — single-primitive lifts that don't pull the Mn out of the cluster)

This refines the singleton-triage further: calibration consistency is the load-bearing variable for cluster membership at this grain. A new Mn that scores 80% like a target cluster but diverges on 2-3 primitives where it shouldn't ends up as a singleton.

Finding 3 — Moderate threshold (6/9) reveals NEW pair structure

At threshold 6/9, two new pair-spines emerge that did NOT appear at threshold 8/9:

• amphimedon ↔ coprinopsis (6/6 at bridge + surface; not substrate). Unexpected pair. Both are "multicellular organisms with minimal bridge sophistication" — low CDif, low PF, low ND, low Cm. Reveals a "multicellular-but-minimal-organismal-elaboration" zone shared by sponges and mushroom-forming fungi despite obvious phylogenetic distance. This is the kind of cross-clade structural convergence the partial-level grain is designed to surface.
• marchantia ↔ physcomitrium (6/6 at substrate + bridge; not surface). Wave 1 hypothesis vindicated at moderate threshold. The bryophyte pair forms after Wave 2 corpus growth even though Wave 1's authoring suspected calibration issues. Suggests Wave 1's "physcomitrium-marchantia failed" interpretation was over-strong — they ARE structural neighbors; only the cluster-assignment threshold at strict-9/9 missed it.

These two pair-spines suggest:

• amphimedon and coprinopsis may eventually each anchor a "minimal-elaboration multicellular" attractor if more members are added (placozoan, hydra-pair-like, additional basidiomycete subtypes)
• marchantia and physcomitrium together anchor a non-vascular plant attractor once Wave 3 or future expansion adds a third member (hornwort, additional liverwort/moss)

Finding 4 — Persistent singletons reveal real corpus gaps

After Wave 2 (N=33), the 11 strict-threshold singletons partition into 3 groups:

Group A — paired at moderate threshold (resolved as coverage artifacts at 6/9):

• amphimedon (pairs with coprinopsis at 6/9)
• coprinopsis (pairs with amphimedon at 6/9)
• marchantia (pairs with physcomitrium at 6/9)
• physcomitrium (pairs with marchantia at 6/9)
• halobacterium (joins prokaryote spine at 6/9)
• hydra (joins big animal spine at 6/9)

Group B — unicellular/colonial eukaryotes, singletons even at 6/9:

• dictyostelium (colonial slime mold)
• monosiga (animal sister, unicellular)
• neurospora (multicellular ascomycete fungus)
• tetrahymena (ciliate protist)
• yeast (unicellular ascomycete fungus)

Persistent-singleton diagnosis: Group B is the unicellular-eukaryote-or-aggregative-multicellular region of the corpus — organisms that are NOT in the prokaryote spine (too elaborate at substrate) and NOT in the multicellular-with-tissues clade. This is a real structural region of biology — diverse but currently uncovered. Wave 3 should target this gap explicitly.

Comparison to Wave 1 singletons

6 of 11 singletons resolved at moderate threshold by Wave 2 corpus growth — even though Wave 2 didn't directly add neighbors for most of them. The corpus density itself shifted the clustering enough for pair structure to emerge at moderate threshold.

This is the corpus-density-effect predicted in the gap-analysis doc and validates the Stage 0 hypothesis that singleton resolution requires reaching ~30-40 Mn density.

Decisions for Wave 3

Wave 3 (structural probes) candidate list adjusts based on Wave 2 findings:

Highest priority (target the persistent Group B singletons):

• Additional unicellular/colonial eukaryote candidates: paramecium (ciliate — companion to tetrahymena), chlamydomonas (single-cell green alga), giardia (excavate parasite), volvox (colonial green alga — already on Wave 3 list)
• Additional fungi: agaricus (different basidiomycete from coprinopsis), mucor (zygomycete, third fungal lineage)
• A SECOND choanoflagellate (S. rosetta) to pair-confirm monosiga

Secondary priority (original Wave 3 probes):

• apis-mellifera (honeybee) — eusocial colony
• trichoplax (placozoan) — pre-tissue animal (would test the amphimedon-coprinopsis pair zone)
• lichen (symbiotic) — composite organism
• synechocystis (cyanobacterium) — photosynthetic prokaryote
• plasmodium-falciparum — parasite

Recommendation for Wave 3: ~6-8 additions weighted toward Group-B singleton resolution. Specifically:

1. paramecium (resolves tetrahymena)
2. agaricus or another basidiomycete (resolves coprinopsis-neurospora-yeast)
3. salpingoeca rosetta (resolves monosiga)
4. chlamydomonas (creates protist-with-photosynthesis pair with volvox if added)
5. volvox (creates colonial-protist cluster)
6. mucor (third fungal lineage; tests fungus cluster boundaries)
7. trichoplax (tests amphimedon-coprinopsis pair zone)
8. cyanobacterium synechocystis (tests prokaryote diversity)

Silhouette score changes (Wave 1 → Wave 2)

Most silhouettes improved (corpus growth + calibration discipline tightened cluster structure):

• biology-substrate (raw, hierarchical): 0.665 → 0.665 (unchanged)
• biology-substrate (raw, kmeans): 0.758 → 0.793 (+)
• biology-to-organism-bridge (raw, kmeans): 0.439 → 0.641 (+++)
• biology-to-organism-bridge (full, kmeans): 0.594 → 0.649 (+)
• organism-architecture (raw, kmeans): 0.454 → 0.521 (+)
• organism-architecture (full, kmeans): 0.309 → 0.507 (++)

The bridge level showed the biggest silhouette improvement — Wave 2's vertebrate + mammal + plant additions clarified the bridge-level structure substantially.

Files