The brain treats loneliness like hunger, and a brutal cancer finally yields a cure

She follows her sister around the cage, crawls underneath her as if trying to get a hug. She looks exactly like someone reuniting with family after a long time apart [3]. Researchers studying this behavior, across species from birds to monkeys to cockroaches, have converged on a striking conclusion: social connection isn't a preference or a personality trait. It's a homeostatic drive, like hunger or thirst, with a neural setpoint for how much company each creature needs. When the need isn't met, the brain pushes the animal to fix it, just the way it pushes a hungry one toward food [3]. Neuroscientists at the Salk Institute and elsewhere are now hunting for what they call the "cellular substrate of loneliness," the specific neurons that register social deprivation and fire the drive to reconnect [43]. The amount that's enough differs by species and by individual. Wild meerkats separated from their group scan constantly and look visibly distressed [3]. In the cultural register, a Psyche essay traces how this same biological need became a source of shame in America, driving some people toward AI companions rather than asking other humans for what they need [311].

That last part is what makes it unusual. Most psychiatric drugs take weeks to begin working and don't specifically target suicidal ideation at all. In a separate case that same week, a woman with Alzheimer's disease showed a striking reversal of cognitive symptoms after using psilocybin, prompting researchers to look more carefully at what the compound does to a brain that has been losing ground [532].

SingularityHub reports that the "undruggable" cancers, tumors driven by proteins so slippery that decades of drug design couldn't get a grip, are now beginning to crack open. AI is finding molecular shapes no human researcher had thought to try, and drugs built around those shapes are entering trials for pancreatic and lung cancers that previously had almost nothing to offer patients [554].

It's the first time any animal's complete nervous system has been mapped end to end. A fruit fly has about 140,000 neurons. The human brain has 86 billion. But the principles governing how circuits wire themselves, and what they compute, were never going to be discovered any other way. In a related paper the same week, researchers identified a single brain hub that coordinates sensory predictions across the entire body, providing the first complete circuit-level map of how a nervous system distinguishes sensations it caused from sensations the world caused [224].

In the same week's coverage, a quieter story out of Orange County, California: a program called Shells for Shorelines is collecting oyster shells from local restaurants and using them to rebuild reef habitat that was lost over the past century [748]. The shells from one night's dinner become the foundation for the next decade's ecosystem. A Noema essay lands alongside both stories with an argument that shifts the frame: nature's heritage, the migration routes held in the bodies of bighorn sheep, the behavioral patterns encoded in a forest, belongs to the animals, not to us, and conservation should start from there [359].

The study found a mechanism that could be activated in isolated brain regions without putting the whole animal under. It's an animal study, but the mechanism is specific enough to suggest a path toward clinical applications.

The site is on Indigenous land that the city could not develop. By exercising their own authority over that land rather than waiting for municipal approval, the Squamish were free of Vancouver's building codes, zoning restrictions, and the regulatory layers that have kept density artificially low across the city for decades. They built denser, faster, and more profitably than any conventional developer has managed. The article argues this isn't just an Indigenous success story but a template for solving the housing shortage in cities that have tried and failed every other way.

It is the first vaccine designed entirely by AI to enter human clinical trials. The model didn't refine an existing approach; it explored molecular territory that human researchers hadn't mapped, and found a candidate structure neither group would have looked for by hand.