Date: 21 October 2026. Reviewer status: An independent, skeptical review simulated by ChatGPT, not an assessment by Stephanie Pierce or any member of her lab. Evidence: Standalone source inspection and Chromium interaction tests (in-memory HTML because local-server navigation was blocked by the execution environment); no biological specimen measurements.
| Evaluated aspect | Bones → missing muscles | Soft tissue → joint inference | Regionalized spine → whole-column mechanics |
|---|---|---|---|
| Addresses a concrete evolutionary question | 9.0 | 9.1 | 9.0 |
| Distinguishes established work from proposed novelty | 8.7 | 8.6 | 8.8 |
| Interactive clarity and immediate accessibility | 8.8 | 8.7 | 8.7 |
| Alternative hypotheses and counterfactual controls | 8.5 | 8.6 | 9.0 |
| Explicit uncertainty/assumption labeling | 8.7 | 8.2 | 8.9 |
| Operational reproducibility of the demo | 8.6 | 8.5 | 9.0 |
| Biological/anatomical realism of the synthetic demo | 5.5 | 5.5 | 5.0 |
| Validation on actual extant/fossil specimens | Not performed | Not performed | Not performed |
These illustrative 0–10 ratings are an author’s reviewer rubric; they are not independent or empirical scores. The demos communicate the research questions effectively but cannot substitute for planned comparative zoological or paleontological data. Scores for realistic biological validity cannot honestly exceed 8 before further specimen-derived modeling and evaluation.
Closest-competitor issue. Extant phylogenetic bracketing, published Bishop–Pierce muscle maps, and 3D polygonal muscle reconstruction already exist. The intended contribution is not a new claim of bone-to-muscle reconstruction in general: it is a formal characterization of what is and is not identifiable, with held-out phylogenetic calibration and mechanical constraints that improve independently measured anatomy or moment arms.
Interactive experiments provided. Select synthetic skeletal configurations and preservation levels; compare size-only, bracketing and mechanics-screen approaches; adjust prior breadth and screening strength; choose two schematic muscle groups; reveal a synthetic held-out path. The resulting candidate sets, accepted counts and central 80% path interval update. A fossil-like configuration intentionally has no soft-tissue ground truth. All values are 2D nondimensional proxies.
Actual numerical checks. With the same synthetic default, the mechanics-screen retained 259/320 candidates at threshold 50; 316/320 at threshold 0; and 130/320 at threshold 100. The size-only control retained all 320 unfiltered candidates. Stronger screening thus reduces candidate count as expected. Synthetic coverage of one reference was shown but deliberately not called calibration.
Remaining critical concerns. The bone geometry and paths are schematic, not registered 3D animal anatomy; the filter uses abstract geometry and moment-arm proxies, not validated muscle wrapping, physiological force–length curves or volumetric packing. Bayesian interval calibration is not established; intervals come from synthetic input distributions. Before presenting a paper, test leave-one-lineage-out coverage and uncertainty width on registered extant CT/diceCT and dissection data.
Closest-competitor issue. APSE and experimentally documented tissue-removal effects precede this project. This demo’s claim is not that soft tissues affect movement; it is that multiple anatomically and constitutively compatible tissues could change the robustness of a prespecified between-taxon mechanical comparison.
Interactive experiments provided. Four retained tissue-response selections; PCA response atlas; animated synthetic hip; reduced-order feasible-motion envelope; counterfactual that changes geometry and material response separately and together. It maintains a separate under-demo research proposal stating the 6-DOF, load-achievable and phylogenetic validation plan.
Actual control check. Changing the simulated geometry and material selections switched the displayed sign of the comparative functional contrast. These are toy calculations, not observed evolutionary reversals. No JavaScript runtime errors were detected in Chromium.
Remaining critical concerns. A reduced-order 2-DOF toy hip is not a 6-DOF biomechanical restoration. Tissue response library/prior maps are unvalidated for fossil taxa, and real active locomotor functions require more than passive pose viability. No direct claims about prehistoric locomotion are supported.
Closest-competitor issue. The Pierce–Jones vertebral studies already studied anatomy and jointwise mobility. The proposed additional experiment is whether whole-column mechanical coupling changes held-out deformation predictions and inferential conclusions in a way that isolated-joint models miss.
Interactive experiments provided. Regionally clustered versus shuffled segments with the same stiffness inventory; lateral, sagittal and torsional loading; adjustable neighbor-joint coupling; a segment-level response histogram and whole-spine comparison. This is a dimensionless 10-joint thought experiment.
Analytical null test. At coupling 0, the simulated observed and shuffled columns both generated 37.9° total rotation, with exactly +0.0% difference. At nonzero coupling (default 0.7), they differed by +5.6% for the default scenario. These values are model outputs, not evolutionary measurements or calibrated taxon parameters.
Remaining critical concerns. Joint stiffness values, species configurations and couplings are hypothetical; no rib, muscle, ligament or contact calibration has occurred. Some qualitative regional-order effects can emerge from assumptions about the coupling law; a successful biological study must outperform an analytic independent-joint baseline on held-out extant whole-column load–deformation measurements and compare fossil configurations under identical scaled loading and phylogenetic controls.
No. They are strong illustrations of falsifiable research designs, not scientific validation. What would change the verdict is access to real paired bone–muscle data for A, layered XROMM/soft-tissue data for B, and independently loaded extant whole-column data for C. Evidence of generalization must precede fossil claims. These empirical dependencies, rather than visual polish, determine whether the projects produce major evolutionary findings.