Microsoft researchers introduce MatterGen, a model that can discover... new materials tailored to specific needs—like efficient solar cells or CO2 recycling—advancing progress beyond trial-and-error experiments.show more

Enhanced Model Selection UI We've upgraded the model selection... interface, now displaying key insights like intelligence, speed, and creativity to help you choose the right model for your needs. Whether you're prioritizing fast prototyping, complex problem-solving, or creative flexibility, this update ensures a more informed selection process—making app generation smoother and more tailored than ever.show more

A stroke can destroy brain cells within minutes, often... leaving people with long‑term problems in movement, memory, or speech. But scientists may have found a way to help the brain heal itself. In studies of mice, researchers injected stem cells into areas of the brain damaged by stroke. These stem cells began turning into early‑stage brain cells, and many later developed into active neurons that connected with surrounding tissue. The treatment did more than just grow cells.show more

As life expands, so does what you value. Inside... the all-new electric #GLB, innovation and comfort create a space that moves with you. From the refined materials to the intelligent MBUX interface, every detail is designed to make driving feel effortless and arriving feel exceptional. Discover an interior that welcomes you with light, space, and intuition. Because true progress begins on the inside. #MercedesBenzshow more

Excited to introduce Diffusion Augmented Agents (DAAGs)✨. We give... an agent control of a diffusion model, so it can create its own *synthetic experience*.🪄 The result is a lifelong agent that can learn new reward detectors and policies, much more efficiently. Here's how. 👇show more

Meet #DBRX: a general-purpose LLM that sets a new... standard for efficient open source models. Use the DBRX model in your RAG apps or use the DBRX design to build your own custom LLMs and improve the quality of your GenAI applications.show more

Fine-tune DeepSeek-OCR on your own language! (100% local) DeepSeek-OCR... is a 3B-parameter vision model that achieves 97% precision while using 10× fewer vision tokens than text-based LLMs. It handles tables, papers, and handwriting without killing your GPU or budget. Why it matters: Most vision models treat documents as massive sequences of tokens, making long-context processing expensive and slow. DeepSeek-OCR uses context optical compression to convert 2D layouts into vision tokens, enabling efficient processing of complex documents. The best part? You can easily fine-tune it for your specific use case on a single GPU. I used Unsloth to run this experiment on Persian text and saw an 88.26% improvement in character error rate. ↳ Base model: 149% character error rate (CER) ↳ Fine-tuned model: 60% CER (57% more accurate) ↳ Training time: 60 steps on a single GPU Persian was just the test case. You can swap in your own dataset for any language, document type, or specific domain you're working with. I've shared the complete guide in the next tweet - all the code, notebooks, and environment setup ready to run with a single click. Everything is 100% open-source!show more

🚨 SCIENTISTS JUST REVERSED AGING IN BLOOD STEM CELLS... Researchers at Mount Sinai restored old blood stem cells to a youthful state by repairing tiny cellular recycling systems called lysosomes. In mice, aged stem cells regained the ability to: • regenerate healthy blood • rebuild immune cells • reduce inflammation • behave like young stem cells again One treatment increased blood-forming capacity by more than 8x. Why this matters: Your blood stem cells are responsible for constantly rebuilding your immune system and blood supply throughout life. As they age: • immunity weakens • inflammation rises • cancer risk increases • regeneration declines But this study suggests aging inside these cells may not be permanent. Scientists may have found a biological “reset switch.” The future of anti-aging may not come from replacing organs… …but from restoring the cells that rebuild the body itself. Follow for more future physics and biotech breakthroughs.show more

We’re excited to introduce Text-to-LoRA: a Hypernetwork that generates... task-specific LLM adapters (LoRAs) based on a text description of the task. Catch our presentation at #ICML2025! Paper: Code: Biological systems are capable of rapid adaptation, given limited sensory cues. For example, our human visual system can quickly adapt and tune its light sensitivity to our surroundings. While modern LLMs exhibit a wide variety of capabilities and knowledge, they remain rigid when adding task-specific capabilities. Traditionally, customizing these models requires gathering large datasets and performing often expensive, time-consuming fine-tuning for specific applications. To bypass these limitations, Text-to-LoRA (T2L) meta-learns a “hypernetwork” that takes in a text description of a desired task, as a prompt, and generates a task-specific LoRA that performs well on the task. In our experiments, we show that T2L can encode hundreds of existing LoRA adapters. While the compression is lossy, T2L maintains the performance of task-specifically tuned LoRA adapters. We also show that T2L can even generalize to unseen tasks given a natural language description of the tasks. Importantly, Text-to-LoRA is parameter-efficient. It generates LoRAs in a single, inexpensive step, based solely on a simple text description of the task. This approach is a step towards dramatically lowering the technical and computational barriers, allowing non-technical users to specialize foundation models using plain language, rather than needing deep technical expertise or large compute resources.show more

🚨 SCIENTISTS JUST SHOWED THAT LIGHT CAN TWIST MATTER... USING ITS MAGNETIC FIELD. For a long time, physicists assumed the magnetic component of light was far too weak to have any meaningful effect on matter compared to its electric field. New experiments are challenging that view. Researchers have demonstrated that intense, properly structured light can induce magnetization and even mechanical twisting in materials through its magnetic field alone. This is an extension of the inverse Faraday effect, but observed with greater strength and control than many expected. Why this matters: • It opens new ways to control magnetism and material properties using only light • Could lead to faster, more energy-efficient magnetic memory and spintronic devices • Offers a new tool for manipulating matter at the nanoscale without physical contact • Bridges optics and magnetism in ways that were previously difficult to achieve The deeper implication: Light is not just an information carrier under the right conditions, its magnetic field can directly reshape the magnetic and mechanical state of matter. This blurs the line between electromagnetic waves and material control. If these effects can be scaled and made practical, we may eventually use light itself as a precise tool to write magnetic information or mechanically actuate tiny structures, rather than relying solely on electric currents or physical forces. We’re discovering that light still has hidden capabilities we haven’t fully exploited. How do you think being able to control matter with light’s magnetic field could change technology in the next decade? Follow for more frontier optics, quantum materials, and light-matter physics.show more

🚨 AI JUST DISCOVERED QUANTUM EFFECTS THAT SCIENTISTS DIDN'T... KNOW EXIST. Researchers at the University of Washington used artificial intelligence to simulate dozens of atomically thin sheets of molybdenum ditelluride stacked in precise twisted patterns. At small scales, these materials look relatively ordinary. But when the AI modeled much larger stacks, completely new quantum behaviors emerged phenomena that only exist because of the complex, repeating moiré patterns formed across many layers. Why this matters: • Many of the most interesting quantum effects only appear at scales that are too large for traditional supercomputers to simulate • AI can act as a fast “surrogate” that learns from smaller calculations and predicts behavior at much bigger scales • These large-scale moiré systems can host exotic quantum states useful for quantum computing and new types of electronics • The same approach could be used to discover many other hidden quantum materials The deeper implication: We are entering an era where AI doesn’t just help us analyze data it helps us discover entirely new quantum phenomena that were previously invisible because they only exist in systems too complex for conventional modeling. This could dramatically speed up the search for materials that power future quantum technologies. What do you find more exciting using AI to uncover hidden quantum effects in materials, or the possibility that these stacked atomic sheets could become building blocks for future quantum computers? Follow for more frontier quantum materials and AI-driven discovery.show more

Business notebooks in Gemini now allow you to manage... your business in one focused space. Simply connect your Google Business Profile, add sources or past chats, and get tailored insights and actionable steps to manage your business. Instead of having to start a conversation from scratch every time, Gemini is able to reference your business’ info so you can pick up right where you left off: ✅ Proactive recommendations: See action items like unanswered customer questions or missing holiday hours ✅ Tailored insights: Analyze business trends, discover new market opportunities, and get suggestions on how to optimize your daily operations, such as product pricing or positioning ✅ Streamlined execution: Brainstorm creative concepts, develop personalized assets for upcoming promotional campaigns, and analyze performance metricsshow more

🚨 SCIENTISTS DISCOVERED HUMAN CELLS CAN RESPOND TO SOUND... BY SWITCHING GENES ON AND OFF. Researchers found certain human cells especially fat cells react to mechanical sound vibrations by changing genetic activity. That means sound may influence biology far deeper than we realized. Why this matters: Your cells are not passive. They constantly sense: • pressure • vibration • movement • mechanical stress And now scientists are discovering those signals can directly affect gene behavior itself. In simple terms: Sound may act almost like a biological instruction signal. The implications are enormous: • regenerative medicine • targeted therapies • metabolic control • tissue engineering • non-invasive treatments • future bioelectric technologies The deeper science goes… …the more the human body starts looking less like a machine… …and more like a responsive living frequency system. What happens if biology is listening more than we ever imagined?show more

Think Solar panels are "Green"? think again. Making solar... panels requires toxic chemicals, huge amounts of energy and water, and creates toxic waste. The common types (mono- and poly-crystalline panels) need extremely high heat at every stage of production. For example, temperatures of 2,700°F to 3,600°F are required just to turn sand-like silicon dioxide into usable silicon. Up to half of that silicon is then wasted when the thin wafers are cut. For every 1 megawatt of solar panels produced: About 1.4 tons of toxic chemicals are used (including hydrochloric acid, sodium hydroxide, sulfuric acid, nitric acid, and hydrogen fluoride). Nearly 2,900 tons of water are needed. 8.6 tons of pollution are released—most of it from powerful greenhouse gases (SF₆, NF₃, and C₂F₆) that trap heat thousands of times more effectively than carbon dioxide. Other dangerous waste includes toxic gases like trichlorosilane and silicon tetrachloride, plus harmful dust from the cutting process. The thinner “amorphous” or thin-film solar panels are made with cadmium, a chemical known to cause cancer and damage DNA. By the end of 2016 the world already had about 250,000 tonnes of discarded solar panels. That was just 0.5 % of all electronic waste produced that year. The International Renewable Energy Agency estimates that by 2050 we could be throwing away 6 million tonnes of solar panels every single year, adding up to a total of 78 million tonnes. At that point dead solar panels could make up 10 % of all electronic waste, and the total amount of worn-out panels could be larger than all the electronic waste the world produced in 2018. Recycling is often presented as the easy fix, but it isn’t. It still uses large amounts of energy, water and other resources, and workers have to handle toxic materials that eventually need to be disposed of. Right now only two commercial recycling methods exist for solar panels, and there are just a handful of facilities that can do the job anywhere in the world. #GreenEnergyshow more

🚨 BREAKING: MIT researchers just built a quantum sensor... that can measure multiple properties at once. That matters more than it sounds. Most solid-state quantum sensors have to measure things one by one: magnetic field, temperature, strain, frequency, phase. But reality doesn’t wait its turn. MIT used entangled qubits inside a diamond defect to measure multiple signal properties in a single shot. Read that again. This means Faster measurements Less error from repeating experiments Better sensing inside complex systems like materials and living cells The wild part? They did it at room temperature. Not in some ultra-cold, impractical lab-only setup. In a platform that could actually matter for real-world sensing. Inside a tiny defect in diamond, quantum correlations were used to pull out: amplitude frequency detuning phase all from the same measurement. That’s a big shift. Because the future of quantum tech isn’t just quantum computers. It’s quantum devices that can see more of reality at once. So the real question is When sensors stop measuring one thing at a time… how much of the hidden structure of matter becomes visible? Follow me for more physics breakthroughs that actually matter.show more

Russian scientists breakthrough in safer rocket design with model... that predicts when space fuel ignites The new mathematical model calculates ignition delay time with high accuracy (within 10–15% error) for hypergolic fuels — propellants that ignite on contact without a spark or heat 🔸The model accounts for heat exchange, evaporation, vapor mixing, and chemical reactions — both in liquid and gas phases 🔸Ignition timing depends on droplet size and temperature; the model accurately predicts both 🔸Larger droplets evaporate faster, but beyond a certain size (about 3 mm), ignition delay time stops changing 🔸The technology allows real-time monitoring of temperature and mixture composition during ignitionshow more

Stop relying on generic AI output Train a model... on your world, your character, your style, and reuse it forever That is the idea behind LTX Trainer, an open-source training framework that lets you train reusable IC-LoRAs for specific video transformations like colorization, deblur, decompression, reference sheet control, and a lot more Each LoRA gives the model a specific behavior it can repeat across clips, so your videos become more consistent, more controlled, and more connected to your own creative direction Check these amazing examples 👇🏻show more

🚨 SOUTH KOREAN SCIENTISTS JUST CREATED HOLLOW SILICON NANOTUBES... THAT TRAP HEAT AND TURN WASTE ENERGY INTO ELECTRICITY. Researchers at POSTECH have developed a new hollow silicon nanotube structure that dramatically reduces thermal conductivity. By turning solid nanowires into microscopic pipes, they trapped heat-carrying particles (phonons) inside the tubes, cutting thermal conductivity by 70% compared to solid wires. Even when both structures had the same surface area, the hollow nanotubes still ran 33% cooler. This phonon localization effect previously thought to require extreme cold or exotic materials was achieved at near-room temperature using simple silicon nanotubes. Why this matters: • Waste heat from data centers, EV batteries, factories, and electronics is currently lost this could capture and convert it into usable electricity • The technology uses abundant, cheap silicon instead of rare and expensive materials like bismuth and tellurium • It’s highly compatible with existing semiconductor manufacturing, making large-scale production more realistic • It solves a long-standing problem: silicon is great for chips but terrible for thermoelectric energy conversion The deeper implication: This breakthrough shows that clever nanoscale engineering can unlock new capabilities from ordinary materials. By controlling how heat moves at the atomic level, researchers are opening a path to more efficient energy recovery systems without relying on scarce resources. As AI and computing power keep growing, finding ways to recycle the massive amounts of waste heat they generate will become increasingly important. How significant do you think waste-heat recovery technologies like this could become in the next decade? Follow for more frontier materials science and energy innovation.show more

PaLM-E or GPT-4 can speak in many languages and... understand images. What if they could speak robot actions? Introducing RT-2: our new model that uses a VLM (up to 55B params) backbone and fine-tunes it to directly output robot actions!show more

NEW SECRET MYSTERY AI VIDEO MODEL ARENA CODE NAME:... BOUNCYBOHR Unlike Artificial Analysis, on Arena. AI you can submit your own prompts and this prompt is designed to try and unmask who created the model. Models powered by an LLM (like Seedance 2.0 and Omni) are smart and can see what you are doing. As far as thoughts? Eh? While another screenshot looked like Omni, this does look pretty Seedance to me. Judging by the result, this looks like the rumored new "improved" FAST model and not upgraded full 2.1 model that is supposed to be 25% better. Text To Video close-up cinematic scene from a film, professionally directed, focused, ultrafine details a woman laughs as she whiteboard writes out what model made this videoshow more