By analyzing more than 800,000 single-cell profiles in tissues and blood, researchers assemble a massive biomolecular compendium of #lupus# nephritis that maps out its impact on cell populations and could inform future therapeutics. @ScienceTM

Claude Code can now work like a research scientist. K-Dense open-sourced 177 science skills for coding agents: - Drug discovery: query ChEMBL, screen molecules with RDKit and DiffDock - Single-cell RNA-seq analysis with Scanpy - 100+ scientific databases in one place - Biology, chemistry, medicine, physics, engineering, Earth science - Works with Claude Code, Codex and Cursor 52K stars. free. MIT license.

Tesla has released a new blog post about how they design vehicle battery packs around safety, durability, and longevity. "Tesla takes a layered approach to battery safety. The design objective is that even if an issue occurs in a single cell and results in a thermal reaction, the problem is contained within the battery pack. This principle is called passive propagation resistance, or PPR. In a Tesla battery pack, cells are surrounded by thermal barriers, structurally isolated and cooled so that failure in one cell does not spread to neighboring cells. The goal is simple: one cell failure should never become a catastrophic battery pack failure. We ensure safe performance at three levels: • Cell level: How does a single cell behave when it fails? Does it vent safely, or does it rupture unpredictably? We study many types of cell failures at high speed to understand exactly how they break in various conditions. Then our engineers adjust the cell design so that if it fails, it releases energy in a controlled manner rather than unpredictably. • Module level: If one cell enters a self-sustaining thermal reaction (thermal runaway), can neighboring cells withstand the heat and pressure without joining in? Thermal isolation materials, deliberate spacing and liquid-filled cooling channels are all engineered to absorb and redirect the thermal energy to prevent propagation. • Vehicle level: How does one cell’s thermal runaway interact with the rest of the vehicle? We test PPR under harsh conditions—high temperatures, full state of charge, loss of coolant flow and others—and validate that the system remains self-contained. Car manufacturers are not typically required to perform PPR testing. It has always been a core requirement for our products, and we test voluntarily across many failure modes and environmental conditions before releasing a product. In addition to this passively safe design, we leverage our anonymized vehicle data to learn about rare issues and to develop active diagnostics with over-the-air software updates that make our products safer over time. Our layered approach to safety has demonstrated results: Our latest data shows that vehicle fires are significantly less likely to occur with Tesla vehicles than the average U.S. vehicle. In addition to having a lower rate of vehicle fires from all causes, our vehicles have shown an extremely low risk of battery fires. From over 265 billion miles of operation as of the end of 2025, we have no evidence of even a single case of a spontaneous battery failure leading to a vehicle fire in a Model 3, Model Y, Cybertruck or Semi. Thermal Management for Every Cell Batteries perform best and last longest within a specific temperature range. If they’re too hot, they degrade more quickly. If they’re too cold, they temporarily lose charge capacity. Our thermal management system is designed to keep every cell at the optimal temperature. Liquid cooling and heating channels run through the battery pack, actively managing cell temperatures during charging, discharging and while the car is parked. It is not purely reactive—the system anticipates conditions as well. For example, if you set your destination as a Supercharger stop, the pack will pre-condition itself to the ideal temperature for faster charging before you arrive. Why this matters for longevity: • Cells that spend less time at extreme temperatures degrade more slowly. Our thermal management system directly extends the life of the battery. • Effective temperature control ensures both safety and longevity. A well-managed cell is less likely to enter an unstable state and has less chemical degradation over time. • Consistent temperatures across the pack increase longevity. Since all cells will age at roughly the same rate, this prevents weak links that could limit total pack capacity. Inside Our 4680 Cell In addition to the cells that we purchase from third-party vendors, we also manufacture our own 4680 cells for use in vehicles like Cybertruck and our Berlin-manufactured Model Y. This vertical integration means we control the entire supply chain, from materials selection to cell chemistry, manufacturing process and quality inspection. Materials and Chemistry A battery cell has two major components: the mechanical exterior structure (the can, lid and housing) and the internal electrochemistry. The electrochemistry is what determines how fast the cell charges, how much energy it stores and how long it lasts. Three pieces of this chemistry matter most: • Cathode: This is where the lithium lives when the cell is discharged. We engineer high-nickel layered oxide cathodes (like NMC [nickel, manganese and cobalt]) for high energy density in 4680, and lithium iron phosphate (LFP) cathodes in standard-range vehicles for durability and cost in non-4680 cells. The cathode can be coated in protective layers to help resist cracking during charge cycles, like how a protective clear coat on paint helps prevent chipping. • Anode: This is where the lithium goes during charging. The anode's material properties determine how fast the cell can accept a charge and how well it holds a charge over thousands of charge cycles. • Electrolyte: This is the liquid medium that lithium ions travel through between cathode and anode. We directly formulate our own electrolyte recipes, tuning the mix of lithium salts, solvents and additives to balance charging speed, energy capacity and lifespan. A critical factor in cell longevity is the solid electrolyte interphase, or SEI, which is a thin layer that forms on the anode surface. Think of it as the patina that develops on copper: it is a natural byproduct of use, and its quality determines whether the underlying material is protected or slowly consumed. We engineer the electrolyte chemistry and anode material specifically to enable a stable, resilient SEI, which directly extends the cell's useful life and supports fast charging without excessive degradation. Dry Electrode Manufacturing We are the only battery manufacturer using dry electrode technology at scale. Traditional electrode manufacturing uses liquid solvents that must be evaporated and captured in energy-intensive ovens. Dry electrode processing skips this step, making the process cleaner and less energy intensive. It is a meaningful environmental improvement in how the cell itself is made before it ever goes into a car. Inspecting Every Cell for Quality At the scale we manufacture our vehicles, cell defects are statistically inevitable. The goal is not to pretend defects do not exist—it is to catch every single one before it leaves the factory, and to design the battery pack so that even if a defect slips through, it cannot cause harm. Our quality inspection regimen includes: • 100% inspection: Every cell is tested. • Vision systems: Automated cameras inspect cells for visual defects on the production line. • X-ray and CT scanning: X-ray scans are performed on 100% of cells, and CT scanning is performed on a significant fraction of cells to catch internal defects that are invisible from the outside. • AI-based detection: AI-powered vision and anomaly detection systems are being developed to identify subtle patterns that human inspectors and traditional algorithms could miss. These systems allow for the inspection of much larger sets of data and volumes of cells compared to what would be possible with more traditional methods. There is also a deliberate engineering tradeoff in cell size. Larger cells are more efficient to manufacture and pack, but smaller cells reduce the energy released if a single cell fails. Our 4680-cell format strikes a balance—it’s large enough for manufacturing and energy efficiency, but small enough that the safety systems can comfortably handle a single-cell event. Rigorous, Fundamental and Continuous Testing Our testing program for battery materials and cells is built on two principles: • Vertical integration drives testing design. Because we build our own cells for our own packs, we design testing protocols that reflect real-world usage in Tesla vehicles—not generic industry benchmarks. The test regimen is tailored to the actual product. • Testing is rooted in fundamentals. Rather than simply cycling cells until they fail and noting the number, our materials scientists investigate why degradation occurs at the component and material level. When a cathode cracks or an SEI layer grows too thick, we trace it to the root cause and have the expertise to modify the chemistry or redesign it accordingly. This means we run thousands of cycling tests under varied protocols—different charge rates, temperatures and depth-of-discharge patterns—and use the results to forecast long-term performance. It also means that every material change, however small (down to trace-level chemical dopants that prevent cathode cracking), is validated against real degradation data before it enters production. Learning From Every Mile Driven We have a unique advantage that no other startup or traditional automaker can easily replicate: vehicle data telemetry from millions of vehicles on the road. Every Tesla vehicle transmits anonymized, real-time data on battery health, temperature, charge cycles and performance. This data feeds directly back into engineering. When a rare cell anomaly occurs in the field, we can study exactly what happened, correlate it with manufacturing data and refine our designs and quality checks. This closed loop between the road and the factory is one of the most significant advantages we have in battery development. Our batteries last because they are engineered to—from the molecular composition of the electrolyte to the structural design of the pack and the software that manages temperature in real time. They are safe because safety is not an afterthought—it is the foundation the entire system is built on, validated layer by layer, cell by cell, and tested against scenarios that go beyond what regulation demands. We also offer a Battery and Drive Unit Warranty on all new Tesla vehicles for 8 years or 100,000 miles (minimum based on vehicle) with a minimum of 70% battery capacity retention over the warranty period."

Massive blood draw today: > 47 tubes > 250 mL What I'm measuring: > single cell sequencing of circulating immune cells > inflammation, oxidative stress, vascular health > metabolic, lipid, blood glucose regulation > brain

🧬 Over the past few years Tencent has gone deep on AI for Life Science. We've built single-cell models, AI-designed antibodies, and now cell-level drug prediction. Today we share one of our latest results: UniPert-G2CP, a two-stage framework (published in Cell) that predicts how a drug molecule affects a specific cell, trained and evaluated on 4,994 genes × 7,860 compounds across 5 cancer cell lines. I'm not a specialist, so here comes the easy version: the same drug can work very differently across genetic backgrounds and cell types. Prior research mostly worked at the molecular level, but our result gives AI the ability to predict how a drug molecule affects a cell. For the real mechanics, the full Cell paper is here👀: We'd love your takes

MSK Employee Spotlight | @dana_peer, Chair of the Computational and Systems Biology Program the Sloan Kettering institute (SKI) at Memorial Sloan Kettering Cancer Center (MSK), was recently elected to the National Academy of Medicine (NAM), one of the most respected distinctions in science and medicine. Dr. Pe’er is a preeminent computational biologist whose work has transformed how scientists use large-scale genomic data to understand complex biological systems. Her lab has pioneered widely used machine-learning and AI approaches to reveal how cells take on certain states, communicate with one another, and organize into tissues. She applies these approaches to fundamental questions in organismal development and in cancer, with a focus on cellular plasticity, metastasis, and how cancer cells shape — and are shaped by — their tissue ecosystems during progression and therapy. Currently, Dr. Pe’er is the Chair of Computational and Systems Biology and Scientific Director of SKI’s Single-cell Analytics Innovation Lab, as well as a Howard Hughes Medical Institute Investigator. She holds the Alan and Sandra Gerry Endowed Chair. Join us in congratulating Dr. Pe'er in the comments below!

An AI model just rediscovered Kepler's law from latent space, with zero knowledge of orbital mechanics. And the same core architecture spans molecular dynamics, cell prediction, and even generating cancer treatment hypotheses. Title: JEPA-Anything: Learning Predictive Models across Different Worlds URL: In one sentence: it decomposes a target's representation into K orthogonal subspaces, each with its own dedicated predictor — "Orthogonal Predictive Factorization" (OPF) — and applies this single mechanism across radically different domains, from vision and biology to clinical data, control, and molecular dynamics. 🔭 Highlight 1: Rediscovering a physical law from latent space Trained only on orbital motion data, its latent frequency modes recovered Kepler's law f=(2π)⁻¹a⁻³/². The fitted slope was -1.4991 against a theoretical -1.5, with R²=0.9999999. 🧬 Highlight 2: Generating and validating a cancer treatment hypothesis Factor analysis on liver cancer data proposed combining IL-18 and CD73 blockade, which then showed the strongest tumor cell killing in actual patient-derived organoids. ⚛️ Highlight 3: Consistently strong across molecular dynamics and cell prediction It achieved the lowest error across 100-step molecular simulations of water, quartz, paracetamol, and benzene, and also beat prior methods on single-cell perturbation prediction. It's striking that one core architecture spans such wildly different scientific domains this well. #WorldModels# #AIforScience#

Grok 4.6 just beat GPT-5.6 Sol, Claude Opus 5 and every other major frontier model tested on an independent biosecurity benchmark And the reason it won is even more impressive than the #1# ranking LatchBio tested whether models can recognize concealed biological risks without becoming useless for legitimate scientific research Grok 4.6 came out as the most performant model tested when balancing: • Refusing disguised biosecurity attacks • Still helping with legitimate dual-use research • Maintaining strong biological intelligence That balance is insanely difficult A weaker safety approach just blocks anything that looks sensitive Grok actually understands the intent behind the request And LatchBio found that much of this behavior comes from the intelligence and reasoning of Grok 4.6 itself rather than heavily relying on external classifiers or system-level blocks At the same time, Grok remains near the absolute frontier across therapeutics, variant discovery, epigenomics, spatial transcriptomics, single-cell transcriptomics and pathogen surveillance