Case studies
Pre-registered structural analyses. Each study defines layers, commits a hypothesis hash, runs extraction, and reports the result — including denials with named mechanisms.
Published
Softwarehub_shadow
WordPress wp-includes
post.php as a hub shadow
r = 0.5156, p = 0.012
n = 25 modules · 23,000+ commits · 3 layers · hub shadow rank gap = 15
AI / ML Researchuniversal_hub
AI Model Architecture Graph
transformer confirmed as #1 structural hub from topology alone
r = 0.9147, p = 0.002
n = 20 models · 83 relations · 3 layers · 4/5 confirmed · 1 denied with named mechanism
Softwarechameleon
Next.js App Router Transition
bundles as infrastructure chameleon
r = 0.8995, p = 0.002
n = 25 modules · 23,444 commits · 3 layers · gradient Δr = 0.35
Neurosciencechameleon
Drosophila Larval Connectome
FPL confirmed across 600 million years — PN vs MBIN hub identity divergence
Spearman ρ = 0.663, p = 0.002
n = 2,952 neurons · 2 layers · Winding et al. 2023 · cross-species replication from C. elegans · F13
Neuroscienceuniversal_hub
C. elegans Connectome
PVCL/PVCR confirmed as primary command interneurons
r = 0.7774, p = 0.004
n = 15 neurons · 2 layers · true external validation · White et al. 1986
Financehub_shadow
2008 Financial Crisis
AIG as cross-layer hub — law DENIED with named mechanism
r = 0.042 (DENIED)
n = 16 institutions · 3 layers · institutional operating model boundary · 10 hub shadows
Project Managementuniversal_hub
Mars Mission Topology
Launch Window as universal hub — critical path confirmed structurally
r (Gantt ↔ Risk)
Gantt + Risk layers · hub shadows expose schedule vs. risk divergence · try it live at /project-risk
What each case study documents
·
Data source — Where nodes and edges come from. Always public or cited.
·
Layer definitions — Exactly how d1, d2, and d3 are constructed. No undisclosed approximations.
·
Pre-registration — SHA-256 hash of the hypothesis file committed before extraction.
·
Cross-layer r — Pearson r per layer pair, p-value, 95% CI, effect size.
·
Hub ranking table — All nodes ranked per layer with archetype classification.
·
Denied hypotheses — Every DENIED result includes a named mechanism explaining why.