Particle & NuclearTwenty-one abstract states claim to generate 27 particle-physics observables, including a fine-structure constant within 22 parts per million of the reference value.A six-bit hypercube contains 64 states, and this preprint filters that space down to a 21-state operator whose transition graph is a dodecahedron. Sharad H. Bachani identifies 13 graph channels with the dimensions of the Standard Model gauge groups and derives 27 reported observables without fitted parameters. The table spans 13 orders of magnitude with a stated median error of 0.33 percent. Two new masses, 180.5 GeV and 5.17 TeV, provide prospective targets, while the other matches demand a complete audit of every algebraic choice that produced them. The idea lives or dies where a compact mathematical object meets independent calculation and particle data.
Microbiology & ImmunologySixteen synthetic viruses grew from AI-written genomes, and several outcompeted the natural phage that supplied their evolutionary starting point.Bacteriophages kill bacteria, yet their genomes are tightly coordinated machines that resist casual redesign. A genome language model generated hundreds of Microviridae candidates, 285 assembled successfully, and 16 produced viable phages in the reported assay. The living designs remained 93.0 to 98.8 percent identical to their nearest training genomes, while some beat the natural ΦX174 phage in direct competition. A generated cocktail also evolved activity against three resistant bacterial strains after one, two, and five passages. The experiment turns genome generation into a physical test of whether an AI sequence can reproduce, compete, and keep adapting.
Cell BiologyA human cell can swell roughly 250-fold while keeping enough internal structure for an ordinary microscope to resolve molecular architecture.Mega-expansion microscopy repeatedly embeds, softens, and enlarges fixed biological samples until nanometre-scale structures become separated by distances that a light microscope can see. HeLa cells reached about 228-fold macroscopic expansion after three rounds, while molecular measurements on synaptonemal complexes yielded 253.98-fold. The team reconstructed a neuronal nuclear pore from 20 particles with a reported structural resolution of 34.62 angstroms after scale correction. A fourth round pushed B cells beyond 1,500-fold, where centimetre-scale specimens and diluted fluorescence became practical limits. The method moves the resolution contest from expensive optics toward the chemistry that anchors and labels molecules before the sample expands.
Cell BiologyMitochondria count haem before cells make proteins, using an ancient stress relay that reaches from an inner-membrane protease to the ribosome.Haem carries oxygen and drives essential enzymes, yet excess haem damages cells, so protein production has to follow the available supply. This Nature study identifies a mitochondrial sensor relay in which haem scarcity activates OMA1, releases short DELE1, and recruits the HRI kinase. HRI then phosphorylates eIF2alpha and slows translation while shifting the cell into a stress program. The pathway operated in human cell systems, erythroid K562 cells, and a reconstructed hydra version, suggesting that it predates haemoglobin-based blood. The finding connects mitochondrial chemistry directly to the amount and type of protein a cell attempts to build.
Cell BiologyOne water molecule appears for billionths of a second inside a light sensor, cutting a proton-transfer barrier by 20 kilocalories per mole.LOV photoreceptors react to blue light by forming a temporary covalent bond between a cysteine and the flavin cofactor FMN. Ten crystallographic snapshots from 10 picoseconds to 100 microseconds catch a water molecule moving into the active site before that bond forms. Quantum-mechanical calculations reduce the proton-transfer barrier from 35.3 to 15.2 kilocalories per mole when the water is present, corresponding to a calculated acceleration of roughly 10^14-fold. The covalent adduct reached 45 percent occupancy at 1 microsecond and was complete by 10 microseconds in the experiment. The protein appears to hydrate a dry pocket only when chemistry needs a proton bridge, a moving mechanism that static structures had hidden.
ML MethodsA mass spectrometer loses names for most molecules it sees, so AIMe built a searchable atlas from 105.3 million possible structures.Experimental spectral libraries cover fewer than 1 percent of known compounds, leaving more than 80 percent of detected metabolites without a usable identity in many complex samples. AIMe predicts how molecules fragment, stores more than 800 million spectra, and explains candidate matches as chemical breakage paths. Its forward model reached 0.83 mean cosine similarity on a held-out NIST20 test and retrieved the correct structure first in 35.6 percent of a demanding PubChem-isomer test. At repository scale, it assigned 1,276,371 putative annotations above a 0.8 cosine threshold, and a failed first guess helped direct the synthesis of an unusual cyclic polyamine. The system turns an unknown peak into a navigable chemical neighborhood while leaving final identification to physical standards and experiments.
MetascienceEleven million US patents are drifting farther apart in idea space, and the collisions that once marked crowded invention have almost disappeared.Patent claims from 1836 through 2023 become increasingly dissimilar when measured with a validated modern text embedding. The average patent-interference rate fell from 2.71 percent in 1864 to 1901 to 0.05 percent in 1998 to 2014, a decline above 98 percent. Inventors appear to spread across a growing space of ideas, which reduces duplication while also weakening the knowledge spillovers created by nearby work. A calibrated model assigns 42 percent of the long-run decline in research productivity to these spatial forces, with sensitivity estimates from 33 to 58 percent. The paper reframes scientific crowding as a problem of distance between researchers as well as the number of researchers.
NeuroscienceA half-millimeter probe listens to more than ten brain chemicals at once by turning living sensor cells into a microscopic optical array.Neuromodulators rise and fall together, while most brain probes measure one molecule at a time. MORSE packs engineered living cells around a micro-endoscope and assigns each cell type a fluorescent sensor for a different chemical signal. Representative sixteen-sensor probes showed a median of eight active sensor types, and recordings tracked roughly 70 to 130 cells for as long as eight hours in aggregate. The system estimated norepinephrine and serotonin release in brain slices and detected spontaneous cerebrospinal-fluid serotonin elevations in three of seven mouse recordings. Its living hydrogel, intensity signals, and ventricular location set the current measurement boundary, while the central idea opens a path to following brain chemistry as a moving multichannel state.