#32 — At the Intersections — October 24, 2026

Computational models and neural network mapping are altering how psychiatric conditions are distinguished and categorized. That junction is our starting point on ComplexityPod, examining how systems concepts inform research across disciplines. Here is this week's issue. We begin with an intersection of medical imaging and archaeology, looking at an assessment of the ancient Egyptian mummy known as the Braided Lady. This body had been imaged before with standard radiography, but this study with Andrew Nelson moved to multidetector computed tomography with slice thicknesses of point-six-two-five millimeters. That resolution allows for three-dimensional reconstructions of the tissues without disturbing the physical wrapping. And those reconstructions showed incomplete evisceration alongside preserved soft tissue, but no excerebration—meaning the brain was left in place rather than cleared through the nasal cavity. Which matches known funerary protocols for nonroyal elites during the New Kingdom period. It places the individual socially. It also showed up in the dentition. There was wear consistent with the standard abrasive diet, but also damage aligning with the ritual Opening of the Mouth. Yet as far as cause of death, the scans showed no acute trauma or destructive bone pathology. There were some calcifications that might indicate chronic parasitic infection, but nothing definitive. Tissue dehydration limits what soft tissue analysis can confirm. The scan provides physical structure, not a medical chart. From physical remains, we move to physical infrastructure and resource planning. A study with Slobodan Simonovic examines the gap between empirical hydrology and water policy. They analyzed ten case studies aligned with the International Association of Hydrological Sciences HELPING decade. The common friction is that local governance rarely reflects field data. Because water managers are balancing municipal supply, agriculture, and flood protection against climate variability and population growth simultaneously. Right, and the study categorizes the interventions—surface and subsurface storage, restoration of aquatic ecosystems, transboundary compacts, and computational modeling tools. Simonovic also applied system dynamics to another utility sector: renewable energy resilience during tornado strikes. Specifically calculating the loss of resilience as the total performance deficit integrated across the entire duration of repairs. In simulations, an EF1 tornado caused an average recovery loss of sixty-two days, while an EF2 strike pushed that to one hundred sixteen days. With the EF1 events, recovery time split into two distinct modes. The difference came down to whether replacement parts were already in local inventory or had to be manufactured and shipped. There is also the problem of undetected structural damage. If a turbine keeps spinning with hidden stress, capacity drops over time until detection occurs. The model showed that comprehensive post-event inspections reduced recovery by seven days and cut recovery variance in half compared to expedited checks. And repowering damaged sites with turbines carrying one-point-three times the capacity reduced total resilience loss by seven percent. When applied to the Roscoe Wind Farm in Texas, historical data yielded return intervals of forty-five years for EF1 and fifty-seven years for EF2 events, totaling a hundred-year loss of five hundred twenty-four turbine-days. Turning to clinical technology, a report with Lena Palaniyappan covers the twenty-twenty-six Schizophrenia International Research Society meeting on artificial intelligence. The sessions covered tools like automated speech analysis, retinal imaging, and computational neural modeling to identify psychosis markers. Though a central caution was validating models across external cohorts rather than assuming initial diagnostic boundaries hold up everywhere. And treating lived experience as structured evidence alongside algorithmic predictions, rather than relying solely on sensor data. Palaniyappan's team followed that dimensional approach in a neuroimaging study testing whether affective and non-affective psychoses exist on a single continuum or follow separate biological paths. They examined resting-state functional MRI across four groups: affective psychosis, non-affective psychosis, non-psychotic affective disorders, and healthy controls. The non-affective psychosis cohort showed reduced whole-brain efficiency along with lower efficiency in sensory and somatomotor networks. Whereas whole-brain efficiency remained intact in the affective groups. Instead, those cohorts displayed elevated efficiency in control networks like the dorsal attention and salience networks. That contrast held up in sensitivity analyses. Sensory and motor drops stayed tied to schizophrenia regardless of depressive symptoms, while dorsal attention elevation appeared across both psychotic depression and psychotic bipolar disorder. Which points to two distinct neural axes: an affective axis marked by elevated control-network efficiency, and a psychosis axis defined by sensory-motor and whole-brain deficits. We will return with more research at these crossroads next week. From ComplexityPod, thanks for listening.