Xiangyi Meng, Benjamin Piazza, Csaba Both, Baruch Barzel, Albert-László Barabási (2026), Nature 649, 315–322 (2026) — DOI-10.1038/s41586-025-09784-4
Key takeaways
- The local branching of real physical networks (neurons, vasculature, trees, corals, plants) is predicted by minimizing surface area under a minimum-thickness constraint, not wire length. That explains the observed trifurcations, non-planar and asymmetric branch angles, and orthogonal sprouts, which Steiner and volume minimization forbid.
Seeded from the abstract and a skim, before the reading in NOTE-075. What follows is what the work says about itself.
Physical networks such as connectomes and vasculature are built from material, and “wiring economy” has long predicted their structure. The authors measure local branching geometry across six classes of 3D physical network and find systematic violations of wire-length minimization. They argue that the true material cost depends on full 3D surface geometry, which makes optimization hard. They then find an exact mapping from the surface-minimization problem to string-theory Feynman-diagram worldsheets (Nambu–Goto action). This predicts that as links thicken, tree-like motifs pass from Steiner bifurcations to stable trifurcations, and that the branching-angle distributions match data. They also predict stable orthogonal sprouts, which turn out to be common and functionally important, for example ending on synapses.
Standing in the record
Filed from the survey of 2026-09-25 of work the anthology set aside as out of scope (tier C): 500 seconds of active reading over 1 session in the papers-feed tracker. It was filed Deferred, unread. NOTE-075 is the close reading of 2026-09-25, and it placed the work: Proposed — a promising and cleanly falsifiable hypothesis, but the evidence does not yet separate it from simpler accounts. Three things would settle it: recomputing the empirical P(λ) without the SI’s trifurcation-merging rule, running the actual surface minimiser (rather than a sigmoid fitted to one tetrahedron) on random terminal sets, and testing the branching-to-sprouting breakpoint against a no-breakpoint null.
When seeded, before the reading in NOTE-075, the entry gave a priority for a deeper reading: medium — High-profile and recent (Nature 2026). The owner’s time is the highest in this batch (t=500 s, one session). The main text is captured, but the SI holds the substance of the mapping.
What the seeding asked a deeper reading to check (NOTE-075 is that reading):
- A strong, falsifiable claim about physical network design, with neuroscience relevance (synapse formation). The mapping to string worldsheets is striking and should be checked in the Supplementary Information (§§3–7).
- A deeper reading should check how sensitive the skeletonization and high-degree-node detection are (SI §1), and whether “excellent agreement” holds per dataset or only in aggregate.
Access when seeded: I read the Nature article page, which is open access, in full: abstract, main text sections (Steiner graphs; physical networks as manifolds; degree distribution; angle asymmetry), figure captions 1–4 and the Discussion. I did not read the Supplementary Information. Crossref gives online 2026-01-07 and print 2026-01-08. I found no arXiv preprint; the one arXiv link on the page is a cited reference (2401.02579).