World Labs co-founders Fei-Fei Li, Justin Johnson, and Ben Mildenhall join a16z General Partner Martin Casado to discuss Atlas, their latest world model, and what it reveals about the pursuit of spatial intelligence. At the center of Atlas is what the team calls “new view prediction”: given images or views of a scene, the model predicts what that environment should look like from a different position in space and time. This brings generation and 3D reconstruction into the same model, and raises a broader question about whether predicting views could become a useful primitive for understanding the physical world. They discuss the technical bets behind the model, what it can and can’t yet capture, and the importance of dynamics, editability, and simulation as world models develop. The conversation also explores applications in creative work, architecture, and robotics, where Fei-Fei argues that one of today’s biggest constraints is access to real-world training data. Timestamps: 00:00 — Intro 00:51 — What Atlas Is & Why It Matters 05:15 — Is This a Scaled-Up Video Model or a New Architecture? 08:15 — Spatial Intelligence & Why New View Prediction Matters 21:27 — Did You Know It Was Going to Work? 24:42 — Use Cases: Creatives, Games & Robotics 35:21 — The Elephant in the Room: Video Models vs World Models 37:55 — Will We Get 4D Video You Can Walk Around In? 42:22 — Why New View Prediction Is the Next Token Prediction Resources: Follow Fei-Fei Li on X: https://twitter.com/drfeifei Follow Justin Johnson on X: https://twitter.com/jcjohnss Follow Ben Mildenhall on X: https://twitter.com/BenMildenhall Follow Martin Casado on X: https://twitter.com/martin_casado Learn more about Atlas: https://www.worldlabs.ai/blog/atlas Stay Updated: If you enjoyed this episode, be sure to like, subscribe, and share with your friends! Find a16z on X: / a16z Find a16z on LinkedIn:
/ a16z Listen to the a16z Show on Spotify: https://open.spotify.com/show/5bC65RD… to the a16z Show on Apple Podcasts: https://podcasts.apple.com/us/podcast… Follow our host: https://twitter.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see http://a16z.com/disclosures.
Get the latest international news and world events from around the world.
Researchers tune into Arctic under-ice sounds and test through-ice communication
Beneath the Arctic Ocean is an orchestra featuring natural and human composers, from cracking sea ice and whistling beluga whales to humming shipping-vessel engines. Researchers from MIT Lincoln Laboratory heard some of this cacophony when analyzing data from commercial off-the-shelf sensors that they integrated and deployed in 2024 during the U.S. Navy’s Operation Ice Camp (OIC). This past March, during OIC 2026, the researchers returned to the Arctic with a higher-fidelity version of one of the sensors, a geophone that detects vibrations in the sea ice.
“We’re interested in things that make sound underneath the ice,” says Ben Evans, a researcher in the laboratory’s Advanced Undersea Systems and Technology Group. “For example, our OIC 2024 data contained marine mammal songs. We need a better understanding of how such signals propagate through ice and how to distinguish these signals from other sources.”
This underwater soundscape is shifting as sheets of Arctic sea ice rapidly break and melt, opening previously impassable maritime routes for military and commercial activity. Determining the unique sound profiles, or acoustic signatures, produced by fracturing ice will enable researchers to develop predictive capabilities that can build coastal community resilience, inform geopolitical strategy and surveil adversary Arctic activity.
Targeting Tumor ‘Softness’ Enhances Immunotherapy
Researchers at the University of Southern California (USC) developed a novel tool in which they can detect ‘softness’ of a tumor and help inform therapeutic outcomes. The physical attributes of tumors, particularly hardness, is a persistent obstacle for cancer immunotherapy. Historically, firmness of most solid tumors correlates with the penetration of therapy and could lead to drug resistance. Scientists that developed this technology recently published their findings in Nature Biomedical Engineering. The team, led by Dr. Yingxiao Wang, details how the softness of tumors can allow them to adapt and evade the immune system and treatment. This finding is contradictory from what previous articles in the field have stated.
Wang is the Dwight C. and Hildagarde E. Baum Chair in Biomedical Engineering and Professor of Biomedical Engineering and Molecular Microbiology & Immunology in the USC Viterbi School of Engineering and associated with the Keck School of Medicine. Collaborations with the Wang Lab developed a way to target soft stem-like cancer cells, which are extremely hard to treat. Cancer stem-cells are key drivers of tumor growth, drug resistance, and immune evasion. These cells are a major topic of interest in the Wang Lab. Wang and his team also focus their research on techniques to detect biomarkers and visualize molecular events in cells. These investigations could lead to optimal treatment delivery to the tumor site and enhancement of immunotherapy.
Researchers are using an immunotherapy known as chimeric antigen receptor (CAR)- T cells, which programs T cells to specifically target the tumor. T cells are specialized immune cells tasked with identifying and eliminating disease. They are a critical component of the immune system and correlate to survival. However, in the context of cancer, they become inert and less active due to tumor-secreting molecules and proteins that dysregulate their function. As a result, scientists in the field of immuno-oncology (IO) have focused on these cells to overcome therapeutic resistance. To generate CAR-T cells, scientists take T cells from a patient and engineer them to redirect the immune response toward the tumor. The CAR-T cells are then able to identify specific proteins on the tumor, which reduce off-target cell death and limit toxicity. Unfortunately, CAR-T cells are less effective against solid tumors.
New AI tool maps the hidden universe of small molecules
The human body and its gut microbiome produce thousands of small molecules that shape how the body functions—influencing immunity, metabolism and more. Identifying what those molecules actually are has been one of the biomedical sciences’ most persistent bottlenecks. More than 80% of compounds detected in a typical biological sample cannot be matched to any known structure using current methods.
Researchers at the Boyce Thompson Institute (BTI) and Cornell University have developed a tool that begins to change that. AIMe, short for AI Molecule Explorer, uses a form of artificial intelligence called neuro-symbolic AI to predict, organize and search the mass spectra of more than 100 million known small organic molecules—effectively building a vast searchable map of chemical space that can accelerate hypothesis generation and compound identification.
The work is a collaboration between Frank Schroeder, professor at BTI and in Cornell’s Department of Chemistry and Chemical Biology, and Carla Gomes, professor of computing and information science and director of Cornell’s AI for Science Institute.
How Mercury formed its graphite crust and core
As the BepiColombo mission prepares to enter the final phase of its journey to Mercury, a series of studies conducted by researchers at the University of Liège and KU Leuven sheds new light on the early stages of the evolution of the planet closest to the sun. Using experimental petrology, the researchers are reconstructing in the laboratory the formation of Mercury’s core, the crystallization of its magma ocean and the formation of its mantle. The studies are published in Earth and Planetary Science Letters, Nature Communications and Advances in Geochemistry and Cosmochemistry.
The terrestrial planets (Mercury, Venus, Earth and Mars) are the result of more than four billion years of evolution, which began with accretion from the disk surrounding the young sun. During the early stages of evolution, the heat released caused these planets to melt, creating what is known as a magma ocean.
This key stage determines the distribution of elements between the metallic core and the mantle. As it crystallizes, this ocean structures the solid mantle, the partial melting of which will subsequently generate the magmas that form the crust. It is also at this stage that an initial atmosphere may form.
AI digital twins struggle to predict human behavior, creating ‘funhouse mirror’ distortions
While many fear artificial intelligence will replace humans, using AI to take over some human roles has benefits. Companies can use the technology to conduct surveys and polls, while behavioral scientists can run experiments on digital twins to gather faster insights without risking harm or distress to real participants.
But there’s a huge snag.
Today’s digital twins are just not up to it. New research published in the journal Science Advances found that digital twins seriously misrepresent human behavior.
‘Hidden order in disorder’ makes nanodevices easier to design
Augmented reality (AR) glasses, lenses thinner than a human hair, and holograms floating above your fingertips may sound like technologies from science fiction. At the heart of these emerging technologies, however, lies a nanoscale optical device known as a “metasurface.”
Made of nanostructures smaller than the wavelength of light, metasurfaces can precisely control the direction, color and other properties of light. Despite their remarkable capabilities, their complex structures have long posed a major challenge to researchers.
A research team led by Professor Junsuk Rho and Dr. Seokwoo Kim of POSTECH discovered hidden repeating patterns within seemingly disordered nanostructures, opening a new route toward faster and more accurate analysis and design of metasurfaces. The study was published in Nature Communications.
Ultrafast electrons and lasers reveal unexpectedly strong radiation signals in common semiconductors
Detecting radiation is key to technologies ranging from particle accelerators and scientific instruments to medical imaging and security screening. But current detectors must often make trade-offs, providing signals that are strong but slow or fast but weak. The trade-off between signal strength and speed can limit precision detection.
In new research, Stanford University researchers worked with the Department of Energy’s SLAC National Accelerator Laboratory on a unique experimental setup to detect radiation across a range of materials.
What they found surprised them. Not only did the researchers observe an unexpectedly strong ultrafast radiation signal, they also found that the charge within the materials acted differently than expected.
‘Soft crosslinking’ strategy makes brittle, glassy plastics tougher
Glassy polymers are those whose chains become immobilized below their glass transition temperature. The immobilized chains make them hard and stiff but also brittle, causing them to fracture when stretched. One promising strategy for overcoming this trade-off is to incorporate ionic groups whose reversible electrostatic attractions form physical crosslinks that improve toughness while maintaining stiffness.
Although ionic liquid-based materials have demonstrated that uniformly distributed ionic interactions can improve toughness, this strategy has been difficult to apply to conventional glassy polymers. A more general molecular design strategy is therefore needed to create homogeneous ionic interactions in a wider range of glassy polymers.
Now, researchers from Tokyo University of Science (TUS), Japan, in collaboration with the Japan Science and Technology Agency (JST), Japan, have developed an ionic comb polymer that combines a comb-shaped architecture with bulky 4-dimethylaminopyridine (DMAP) counterions to maintain a homogeneous nanostructure with uniformly distributed ionic interactions, enabling glassy polymers to become both stiff and tough.
A new type of LED light could bring significant efficiency gains
Researchers at Lund University have developed a new type of light-emitting diode based on thin, branched nanowires that could offer significantly higher efficiency and lower production costs than current technology. By controlling where in the structure the light is generated, the researchers have reduced the losses that would otherwise limit the amount of light that can be used. Their study is published in the journal Nano Research.
In materials used for conventional light-emitting diodes—such as the LED bulbs found in most households—a large proportion of the light is trapped inside the material because of what is known as total internal reflection. This phenomenon means that only a small proportion of the light comes out, even though it is generated inside the material.
The new design aims to overcome this problem. The method is based on the fact that light is emitted from very thin side branches that extend from a central nanowire. If the structures are made thin enough—thinner than the wavelength of light—the light cannot be trapped inside the material in the same way.