MetaGraph: Searching the world’s DNA at petabase scale Modern biology is drowning in data. Public sequencing archives now hold more than 67 petabases of DNA and RNA—yet, paradoxically, most of it is beyond search. Finding a gene, variant, or viral fragment across this ocean

Statistical physics of large-scale neural activity—with loops As neural datasets cross 10,000+ simultaneously recorded cells, we need models that capture collective structure without giving up tractability. David P. Carcamo and Christopher W. Lynn take a major step: they solve

Attention to quantum complexity: learning entanglement and dynamics from noisy measurements Large language models taught us that “attention” can extract structure from messy sequences. Hyejin Kim and coauthors ask a bold question: can attention do the same for quantum

Concise networks: Learning how systems remember In many real-world systems—air travel, web navigation, even information flow in social networks—what happens next depends not just on where you are, but on where you’ve been. Yet most network models ignore this “memory,” treating

Digital twins for X-ray imaging: Seeing the invisible without destroying it High-brilliance X-ray imaging lets scientists map the chemical makeup of materials in stunning detail—but there’s a catch. The same radiation that reveals molecular structure can also destroy the very

Implicit neural fields: compressing biological microscopy data with AI Modern biological microscopes can capture breathtakingly detailed 3D, 4D, even 5D images — but this richness comes at a cost. The data are enormous, often terabytes per experiment, overwhelming storage and

Do co-folding models learn physics—or just memorize pockets? Deep models like AlphaFold3 and RoseTTAFold All-Atom can place small molecules into protein pockets with eye-popping accuracy. But when chemistry changes, do they still obey sterics, electrostatics, and chemistry—or

Calibrated uncertainty for polygenic scores: a step toward trustworthy predictive AI in genomics Polygenic scores—genetic models that estimate disease risk from millions of variants—are rapidly entering the clinic. But like many predictive models, they often report a single

The physics of quantum advantage: Coherence as computational fuel What really gives quantum computers their edge over classical ones? For years, physicists have explored key quantum resources such as entanglement, non-stabilizerness (or “magic”), and coherence—each offering a

Modeling disorder with ML: from single snapshots to ensemble properties In real materials, atoms don’t line up perfectly. Surface terminations shuffle, vacancies appear, and local arrangements fluctuate with temperature. Those “messy” details drive transport and optical

Evaluating LLM agents for automation of atomic force microscopy Large language models are beginning to enter the lab. Beyond generating text, they can now plan, execute, and interpret real experiments—pushing the boundary of what “autonomous science” can mean. In this new work,

A future where neural networks don’t just simulate physics—they are physics. Artificial neural networks usually run on chips filled with billions of transistors, consuming vast amounts of energy to move electrical signals between processors and memory. But what if computation

Machine learning meets cohort-scale single-cell biology Modern single-cell technologies can now capture millions of cells from hundreds of people, tissues, and treatments. Yet most analyses still rely on averages—collapsing data into per-sample summaries or predefined clusters

Turning reaction figures into machine-readable chemistry with a multimodal LLM Much of organic chemistry still lives as pictures: arrows, rings, branches, and tiny annotations packed into reaction schemes. Great for humans, tough for machines. That gap bottlenecks AI in

When AI, information theory, and quantum gravity meet The dialogue between theoretical physics and artificial intelligence has been unfolding for years — from statistical physics inspiring neural networks, to geometry and topology guiding new architectures, and even quantum

A free, beginner-friendly course from MIT on computing fundamentals. These classic lessons break down concepts such as programming language design, abstraction, & recursion: bit.ly/46zVDEy v/MIT OpenCourseWare

Frugal spike intelligence: Unsupervised pattern discovery and spike sorting with minimal hardware Neural interfaces stream massive, multichannel time series—but today’s pipelines still lean on heavy detection, feature extraction, and clustering, often offloaded to power-hungry

Molecular intelligence: enzymes that compute Living matter already processes information — it senses, reacts, and adapts — but doing so with the precision of a computer has long seemed out of reach. What if molecules themselves could perform computation, without any electronics?

gReLU: Unifying deep learning for DNA sequence modeling and design Deep learning has transformed genomics — from predicting regulatory activity to designing synthetic DNA. But progress has been fragmented: every new model comes with custom code, incompatible data pipelines, and

ECloudGen: Scaling structure-based drug design with electron clouds Structure-based generative models try to design ligands that fit a specific protein pocket. The catch: we have orders of magnitude more ligand-only data than protein–ligand complexes, so pocket-aware models are