🪐 Astronomers Found Two Giant Planets Less Dense Than Cotton Candy
Astronomers have confirmed the existence of two of the puffiest planets ever discovered — gas giants roughly the size of Jupiter, but with densities so low they are less dense than cotton candy.
The pair, named TOI-791 b and TOI-791 c, orbit an F7-type star about 1,110 light-years from Earth in the southern constellation Volans. Their numbers are almost hard to believe: TOI-791 b has an average density of just 0.038 g/cm³, while TOI-791 c comes in at 0.047 g/cm³.
For comparison, Jupiter’s average density is about 1.33 g/cm³. Cotton candy is roughly 0.05 g/cm³. Earth is around 5.5 g/cm³.
That makes these planets not just “fluffy” by astronomical standards — they are among the lowest-density giant planets ever detected.
The discovery, published in Monthly Notices of the Royal Astronomical Society, is especially valuable because the two planets appear to be locked in a rare 5:3 orbital resonance: for every five orbits of the inner planet, the outer one completes almost exactly three. This gravitational interaction slightly shifts the timing of their transits across the star, allowing astronomers to estimate their masses.
🔹 The planets were first spotted by volunteers in the Planet Hunters TESS citizen-science project
🔹 Confirmation required eight years of observations
🔹 Data from the ASTEP telescope at Antarctica’s Concordia Station were crucial
🔹 Each transit lasts more than 11 hours — unusually long for ground-based observations
🔹 Only a handful of systems are known to contain multiple super-puff planets
The leading idea is that these worlds may have relatively small cores surrounded by enormous hydrogen- and helium-rich atmospheres. But exactly how such diffuse planets form — and how they keep their atmospheres for so long — remains an open question.
Important caveat: these measurements come from transits and orbital timing effects, not from direct imaging. The densities are robust within the current model, but the planets’ true atmospheric composition will require follow-up observations — potentially with the James Webb Space Telescope.
Super-puff planets are strange because they sit at the edge of what our planet-formation models can comfortably explain.
If a giant planet can be less dense than cotton candy and still hold itself together, what else is out there that our theories have not yet learned to expect?
📄 Source: https://academic.oup.com/mnras/article-lookup/doi/10.1093/mnras/stag864
#exoplanets #astronomy #space #TESS #superpuffs
Astronomers have confirmed the existence of two of the puffiest planets ever discovered — gas giants roughly the size of Jupiter, but with densities so low they are less dense than cotton candy.
The pair, named TOI-791 b and TOI-791 c, orbit an F7-type star about 1,110 light-years from Earth in the southern constellation Volans. Their numbers are almost hard to believe: TOI-791 b has an average density of just 0.038 g/cm³, while TOI-791 c comes in at 0.047 g/cm³.
For comparison, Jupiter’s average density is about 1.33 g/cm³. Cotton candy is roughly 0.05 g/cm³. Earth is around 5.5 g/cm³.
That makes these planets not just “fluffy” by astronomical standards — they are among the lowest-density giant planets ever detected.
The discovery, published in Monthly Notices of the Royal Astronomical Society, is especially valuable because the two planets appear to be locked in a rare 5:3 orbital resonance: for every five orbits of the inner planet, the outer one completes almost exactly three. This gravitational interaction slightly shifts the timing of their transits across the star, allowing astronomers to estimate their masses.
🔹 The planets were first spotted by volunteers in the Planet Hunters TESS citizen-science project
🔹 Confirmation required eight years of observations
🔹 Data from the ASTEP telescope at Antarctica’s Concordia Station were crucial
🔹 Each transit lasts more than 11 hours — unusually long for ground-based observations
🔹 Only a handful of systems are known to contain multiple super-puff planets
The leading idea is that these worlds may have relatively small cores surrounded by enormous hydrogen- and helium-rich atmospheres. But exactly how such diffuse planets form — and how they keep their atmospheres for so long — remains an open question.
Important caveat: these measurements come from transits and orbital timing effects, not from direct imaging. The densities are robust within the current model, but the planets’ true atmospheric composition will require follow-up observations — potentially with the James Webb Space Telescope.
Super-puff planets are strange because they sit at the edge of what our planet-formation models can comfortably explain.
If a giant planet can be less dense than cotton candy and still hold itself together, what else is out there that our theories have not yet learned to expect?
📄 Source: https://academic.oup.com/mnras/article-lookup/doi/10.1093/mnras/stag864
#exoplanets #astronomy #space #TESS #superpuffs
OUP Academic
ASTEP confirmation of a pair of long-period Jupiter-sized planets with extremely low densities transiting TOI-791
ABSTRACT. Gas giant planets with periods $20~\lt ~P~\lt ~300~\rm d$ orbiting Sun-like stars are a relatively uncommon outcome of planetary formation, and k
👍9👀5🔥2
🌋 Yellowstone May Not Be Powered by a Deep Mantle Plume After All
Yellowstone is one of Earth’s most famous supervolcanoes — and for decades, many geologists explained it with a familiar image: a deep mantle plume, a vertical column of hot rock rising from near Earth’s core, similar to the plume that built Hawaii.
A new study in Science suggests a very different mechanism.
Researchers built a high-resolution 3D geodynamic model of western North America and found that Yellowstone’s magma may be supplied not by a deep plume, but by the shallow asthenosphere — the hot, slowly flowing layer of mantle just beneath the rigid lithosphere.
The driver is what the authors call an eastward “mantle wind”: a broad horizontal flow of hot rock moving beneath North America at geologic speeds.
This flow appears to be linked to the ancient Farallon Plate, which began sliding beneath North America tens of millions of years ago. Remnants of that plate still sit deep under the continent. As they continue to sink, they help generate a large-scale mantle flow that pushes hot asthenospheric material toward Yellowstone.
Then comes the key part: as this buoyant material is forced beneath the thick continental lithosphere, the stretching and pressure changes trigger decompression melting — producing magma without requiring a deep plume rising from the core-mantle boundary.
The model also helps explain Yellowstone’s unusual underground plumbing. Competing tectonic forces appear to tear the lithosphere beneath the region, creating a southwest-dipping channel. This channel acts like a pathway for magma to rise, spread and evolve into a vast “magma mush” system rather than a simple, long-lived liquid magma chamber.
Why it matters: supereruptions can eject more than 1,000 cubic kilometers of material, blanket huge regions in ash and affect climate for years. Understanding what actually sustains systems like Yellowstone is crucial for long-term volcanic hazard models.
The big takeaway: Yellowstone may be less like a blowtorch from Earth’s deep interior — and more like a tectonic wound kept active by the slow, hidden motion of an ancient plate.
Source:
https://www.science.org/doi/10.1126/science.ady2027
Readable summary:
https://www.sciencedaily.com/releases/2026/06/260622014317.htm
#Yellowstone #Supervolcano #Geology #EarthScience #Science
Yellowstone is one of Earth’s most famous supervolcanoes — and for decades, many geologists explained it with a familiar image: a deep mantle plume, a vertical column of hot rock rising from near Earth’s core, similar to the plume that built Hawaii.
A new study in Science suggests a very different mechanism.
Researchers built a high-resolution 3D geodynamic model of western North America and found that Yellowstone’s magma may be supplied not by a deep plume, but by the shallow asthenosphere — the hot, slowly flowing layer of mantle just beneath the rigid lithosphere.
The driver is what the authors call an eastward “mantle wind”: a broad horizontal flow of hot rock moving beneath North America at geologic speeds.
This flow appears to be linked to the ancient Farallon Plate, which began sliding beneath North America tens of millions of years ago. Remnants of that plate still sit deep under the continent. As they continue to sink, they help generate a large-scale mantle flow that pushes hot asthenospheric material toward Yellowstone.
Then comes the key part: as this buoyant material is forced beneath the thick continental lithosphere, the stretching and pressure changes trigger decompression melting — producing magma without requiring a deep plume rising from the core-mantle boundary.
The model also helps explain Yellowstone’s unusual underground plumbing. Competing tectonic forces appear to tear the lithosphere beneath the region, creating a southwest-dipping channel. This channel acts like a pathway for magma to rise, spread and evolve into a vast “magma mush” system rather than a simple, long-lived liquid magma chamber.
Why it matters: supereruptions can eject more than 1,000 cubic kilometers of material, blanket huge regions in ash and affect climate for years. Understanding what actually sustains systems like Yellowstone is crucial for long-term volcanic hazard models.
The big takeaway: Yellowstone may be less like a blowtorch from Earth’s deep interior — and more like a tectonic wound kept active by the slow, hidden motion of an ancient plate.
Source:
https://www.science.org/doi/10.1126/science.ady2027
Readable summary:
https://www.sciencedaily.com/releases/2026/06/260622014317.htm
#Yellowstone #Supervolcano #Geology #EarthScience #Science
Science
Tectonic origin of Yellowstone’s translithospheric magma plumbing system
Yellowstone is widely recognized for its crustal magma reservoirs replenished by asthenospheric melts. However, how primary melts traverse the rigid lithosphere and evolve into bimodal volcanism remains unclear. By leveraging multidisciplinary ...
👍11⚡10🕊4😁1
⚛️ Physicists Create a Strange New Quantum State — the “Fractional Fermi Sea”
Quantum simulators are usually built to recreate known physics in a clean, controllable setting. But a team at the University of Innsbruck has pushed the idea further: they engineered a highly unusual quantum state that appears to go beyond one of the standard frameworks for one-dimensional matter.
The researchers used ultracold cesium atoms confined in one-dimensional tubes and drove them far from equilibrium by cycling the interactions between strongly repulsive and strongly attractive regimes. Normally, this kind of forcing might be expected to heat the system and wash out any structure.
Instead, the atoms reorganized into something unexpectedly ordered.
The state is called a “fractional Fermi sea” — a highly excited, yet stable configuration where particles behave as if the usual occupancy rules have been replaced by a reduced, fractional version. It does not literally rewrite the Pauli exclusion principle, but it realizes behavior long associated with Haldane’s generalized exclusion statistics: particles filling available states in a fractional way.
What makes this especially interesting is that the correlations do not fit neatly into the familiar Tomonaga–Luttinger liquid picture, the classic theory used to describe many one-dimensional quantum systems. The particles show distinctive Friedel oscillations — ripples in density correlations — and decay patterns that point to a new kind of critical quantum phase.
In simple terms: the system is not cold, calm, and sitting in its lowest-energy state. It is highly excited — but not chaotic. Hidden order emerges from the drive.
The theoretical work has now been published in Physical Review Letters, while the companion experimental realization is available as a preprint.
Why it matters: quantum simulators are no longer just “physics replay machines.” They can create and probe states of matter that may be extremely hard — or impossible — to find naturally, opening new ways to study strongly correlated systems, exotic statistics, and future quantum technologies.
The big takeaway: sometimes the deepest order in quantum matter appears not when everything is perfectly still, but when a system is pushed far from equilibrium and refuses to fall apart.
📄 Theory paper: Physical Review Letters 136, 230402 (2026)
https://doi.org/10.1103/j3s5-gjpf
📄 Experimental preprint:
https://arxiv.org/abs/2602.17657
📖 Summary:
https://www.uibk.ac.at/en/newsroom/2026/a-novel-critical-quantum-phase/
#QuantumPhysics #CondensedMatter #QuantumSimulation #Physics
Quantum simulators are usually built to recreate known physics in a clean, controllable setting. But a team at the University of Innsbruck has pushed the idea further: they engineered a highly unusual quantum state that appears to go beyond one of the standard frameworks for one-dimensional matter.
The researchers used ultracold cesium atoms confined in one-dimensional tubes and drove them far from equilibrium by cycling the interactions between strongly repulsive and strongly attractive regimes. Normally, this kind of forcing might be expected to heat the system and wash out any structure.
Instead, the atoms reorganized into something unexpectedly ordered.
The state is called a “fractional Fermi sea” — a highly excited, yet stable configuration where particles behave as if the usual occupancy rules have been replaced by a reduced, fractional version. It does not literally rewrite the Pauli exclusion principle, but it realizes behavior long associated with Haldane’s generalized exclusion statistics: particles filling available states in a fractional way.
What makes this especially interesting is that the correlations do not fit neatly into the familiar Tomonaga–Luttinger liquid picture, the classic theory used to describe many one-dimensional quantum systems. The particles show distinctive Friedel oscillations — ripples in density correlations — and decay patterns that point to a new kind of critical quantum phase.
In simple terms: the system is not cold, calm, and sitting in its lowest-energy state. It is highly excited — but not chaotic. Hidden order emerges from the drive.
The theoretical work has now been published in Physical Review Letters, while the companion experimental realization is available as a preprint.
Why it matters: quantum simulators are no longer just “physics replay machines.” They can create and probe states of matter that may be extremely hard — or impossible — to find naturally, opening new ways to study strongly correlated systems, exotic statistics, and future quantum technologies.
The big takeaway: sometimes the deepest order in quantum matter appears not when everything is perfectly still, but when a system is pushed far from equilibrium and refuses to fall apart.
📄 Theory paper: Physical Review Letters 136, 230402 (2026)
https://doi.org/10.1103/j3s5-gjpf
📄 Experimental preprint:
https://arxiv.org/abs/2602.17657
📖 Summary:
https://www.uibk.ac.at/en/newsroom/2026/a-novel-critical-quantum-phase/
#QuantumPhysics #CondensedMatter #QuantumSimulation #Physics
Physical Review Letters
Exotic Critical States as Fractional Fermi Seas in the One-Dimensional Bose Gas
Critical quantum field theories occupy a central position in modern theoretical physics for their inherent universality stemming from long-range correlations. As an example, the Tomonaga-Luttinger liquid (TLL) describes a wealth of one-dimensional quantum…
👍12👀3🔥1
🤖 The Transformer Era Is Evolving — NVIDIA’s New Hybrid Model Shows What May Come Next
NVIDIA has released Nemotron 3 Ultra, a 550-billion-parameter open model that matters less for its size than for what sits under the hood.
Instead of being a classic Transformer-only system, Nemotron 3 Ultra uses a hybrid architecture: a Latent Mixture-of-Experts design that interleaves Mamba-2 state-space layers with selected attention layers. In simple terms, NVIDIA is not throwing Transformers away — it is replacing some of the expensive attention machinery with more efficient sequence-processing components, while keeping attention where precision still matters.
The numbers are serious: 550B total parameters, 55B active per token, up to a 1-million-token context window, and Multi-Token Prediction layers for faster generation through native speculative decoding. NVIDIA has released the model weights, data, and recipes under the OpenMDW-1.1 license, making this one of the most ambitious open-weight frontier model releases so far.
The benchmarks are also impressive. NVIDIA reports 71.9% on SWE-Bench Verified, 87.0% on GPQA without tools, 56.4 on Terminal Bench 2.1, and strong long-context performance on RULER at 1M tokens.
But the real story is architectural.
For years, the Transformer has been the default architecture behind modern AI. Its weakness is also well known: standard attention becomes increasingly expensive as context grows. Mamba-style state-space layers offer a different way to process long sequences more efficiently. Nemotron 3 Ultra suggests that the next generation of large models may not be “Transformer vs. Mamba,” but carefully engineered hybrids that combine both.
Nikolas Bush Take
The honest caveat: these are NVIDIA’s own benchmark numbers, and real-world agentic performance is always messier than leaderboard scores. A 71.9% SWE-Bench Verified result is impressive, but it still means the model fails a meaningful share of real software-engineering tasks.
The big takeaway: the Transformer is not dead. But its monopoly may be ending. The future of frontier AI may look less like one dominant architecture — and more like modular systems where attention, state-space layers, MoE routing, long-context memory, and inference-time reasoning are mixed together for efficiency and performance.
Sources:
• NVIDIA Nemotron 3 Ultra Model Card
https://build.nvidia.com/nvidia/nemotron-3-ultra-550b-a55b/modelcard
• NVIDIA Research: Nemotron 3 Ultra
https://research.nvidia.com/labs/nemotron/Nemotron-3-Ultra/
• NVIDIA Technical Blog
https://developer.nvidia.com/blog/nvidia-nemotron-3-ultra-powers-faster-more-efficient-reasoning-for-long-running-agents/
#Nemotron3 #NVIDIA #MambaArchitecture #AI #OpenWeights #StateSpaceModels #Transformers
NVIDIA has released Nemotron 3 Ultra, a 550-billion-parameter open model that matters less for its size than for what sits under the hood.
Instead of being a classic Transformer-only system, Nemotron 3 Ultra uses a hybrid architecture: a Latent Mixture-of-Experts design that interleaves Mamba-2 state-space layers with selected attention layers. In simple terms, NVIDIA is not throwing Transformers away — it is replacing some of the expensive attention machinery with more efficient sequence-processing components, while keeping attention where precision still matters.
The numbers are serious: 550B total parameters, 55B active per token, up to a 1-million-token context window, and Multi-Token Prediction layers for faster generation through native speculative decoding. NVIDIA has released the model weights, data, and recipes under the OpenMDW-1.1 license, making this one of the most ambitious open-weight frontier model releases so far.
The benchmarks are also impressive. NVIDIA reports 71.9% on SWE-Bench Verified, 87.0% on GPQA without tools, 56.4 on Terminal Bench 2.1, and strong long-context performance on RULER at 1M tokens.
But the real story is architectural.
For years, the Transformer has been the default architecture behind modern AI. Its weakness is also well known: standard attention becomes increasingly expensive as context grows. Mamba-style state-space layers offer a different way to process long sequences more efficiently. Nemotron 3 Ultra suggests that the next generation of large models may not be “Transformer vs. Mamba,” but carefully engineered hybrids that combine both.
Nikolas Bush Take
The Mamba moment has arrived — but not as a revolution overnight. NVIDIA did not ship a pure state-space model. It shipped a pragmatic hybrid. That is probably the pattern to watch: keep attention where it creates value, replace it where it becomes too expensive.
Open-weight frontier models are now strategic infrastructure. NVIDIA is not just selling GPUs anymore. By releasing serious open models, datasets, and recipes, it is pulling developers deeper into its full-stack AI ecosystem — hardware, software, inference, agents, and deployment.
The next AI race may be less about raw parameter count and more about architecture, inference efficiency, data quality, and agentic reliability. A 55B-active model with strong benchmark results is a signal that “useful scale” is becoming more nuanced than simply making models bigger.
The honest caveat: these are NVIDIA’s own benchmark numbers, and real-world agentic performance is always messier than leaderboard scores. A 71.9% SWE-Bench Verified result is impressive, but it still means the model fails a meaningful share of real software-engineering tasks.
The big takeaway: the Transformer is not dead. But its monopoly may be ending. The future of frontier AI may look less like one dominant architecture — and more like modular systems where attention, state-space layers, MoE routing, long-context memory, and inference-time reasoning are mixed together for efficiency and performance.
Sources:
• NVIDIA Nemotron 3 Ultra Model Card
https://build.nvidia.com/nvidia/nemotron-3-ultra-550b-a55b/modelcard
• NVIDIA Research: Nemotron 3 Ultra
https://research.nvidia.com/labs/nemotron/Nemotron-3-Ultra/
• NVIDIA Technical Blog
https://developer.nvidia.com/blog/nvidia-nemotron-3-ultra-powers-faster-more-efficient-reasoning-for-long-running-agents/
#Nemotron3 #NVIDIA #MambaArchitecture #AI #OpenWeights #StateSpaceModels #Transformers
NVIDIA NIM
nemotron-3-ultra-550b-a55b Model by NVIDIA | NVIDIA NIM
Open, efficient hybrid Mamba-Transformer MoE with 1M context, excelling in agentic reasoning, coding, planning, tool calling, and more
⚡8
🧬 Scientists May Have Found a New Way to Mass-Produce Cancer-Fighting Immune Cells
For more than a decade, cancer immunotherapy has been dominated by T cells. CAR-T therapies can be powerful against some blood cancers, but they remain expensive, highly personalized, and much harder to use against solid tumors.
Now researchers at USC Stem Cell have shifted attention to a different immune lineage: granulocyte-monocyte progenitors, or GMPs — early precursor cells that can produce macrophages, the immune system’s “first responders.”
Macrophages are especially interesting because they naturally enter tumors, engulf abnormal cells, and help coordinate immune responses. But mature macrophages are difficult to grow in large numbers, hard to genetically engineer, and not ideal for freezing and storage.
The USC team worked one step earlier — with GMPs before they mature. Using a defined chemical cocktail, they managed to keep mouse and human GMPs in a progenitor-like state while allowing them to expand long-term in the lab. That is important because long-term self-renewal in the blood system was traditionally associated mainly with true hematopoietic stem cells, not more committed progenitors.
Then the researchers engineered these GMPs with CAR receptors so they could recognize cancer cells. They also added a CAR-Fc design that can recruit other immune cells and help activate broader anti-tumor responses.
In mouse experiments, the engineered GMPs settled into bone marrow and other blood-forming tissues, where they continuously generated tumor-infiltrating macrophages and other myeloid cells. The cells suppressed CD19-positive leukemia and HER2-positive solid tumors, and the dual CAR-Fc design showed stronger effects in allogeneic cancer models.
The same platform also restored antibacterial defense in mice with chronic granulomatous disease, an inherited immune deficiency disorder.
Why this matters: this is not just another CAR-T variant. It is a possible manufacturing breakthrough for an entirely different branch of the immune system — one that may be better suited for solid tumors and potentially easier to produce as an off-the-shelf therapy.
The important caveat: these are still preclinical results in mice. The real test will be whether the platform is safe, durable, and effective in humans.
But the idea is powerful: instead of only engineering mature immune cells, scientists may be learning how to grow a renewable upstream factory that keeps producing cancer-fighting cells from inside the body.
Paper: Cell
https://www.cell.com/cell/fulltext/S0092-8674(26)00643-4
Summary: ScienceDaily
https://www.sciencedaily.com/releases/2026/06/260620100317.htm
#Immunotherapy #Cancer #StemCells #CellTherapy #Biotech
For more than a decade, cancer immunotherapy has been dominated by T cells. CAR-T therapies can be powerful against some blood cancers, but they remain expensive, highly personalized, and much harder to use against solid tumors.
Now researchers at USC Stem Cell have shifted attention to a different immune lineage: granulocyte-monocyte progenitors, or GMPs — early precursor cells that can produce macrophages, the immune system’s “first responders.”
Macrophages are especially interesting because they naturally enter tumors, engulf abnormal cells, and help coordinate immune responses. But mature macrophages are difficult to grow in large numbers, hard to genetically engineer, and not ideal for freezing and storage.
The USC team worked one step earlier — with GMPs before they mature. Using a defined chemical cocktail, they managed to keep mouse and human GMPs in a progenitor-like state while allowing them to expand long-term in the lab. That is important because long-term self-renewal in the blood system was traditionally associated mainly with true hematopoietic stem cells, not more committed progenitors.
Then the researchers engineered these GMPs with CAR receptors so they could recognize cancer cells. They also added a CAR-Fc design that can recruit other immune cells and help activate broader anti-tumor responses.
In mouse experiments, the engineered GMPs settled into bone marrow and other blood-forming tissues, where they continuously generated tumor-infiltrating macrophages and other myeloid cells. The cells suppressed CD19-positive leukemia and HER2-positive solid tumors, and the dual CAR-Fc design showed stronger effects in allogeneic cancer models.
The same platform also restored antibacterial defense in mice with chronic granulomatous disease, an inherited immune deficiency disorder.
Why this matters: this is not just another CAR-T variant. It is a possible manufacturing breakthrough for an entirely different branch of the immune system — one that may be better suited for solid tumors and potentially easier to produce as an off-the-shelf therapy.
The important caveat: these are still preclinical results in mice. The real test will be whether the platform is safe, durable, and effective in humans.
But the idea is powerful: instead of only engineering mature immune cells, scientists may be learning how to grow a renewable upstream factory that keeps producing cancer-fighting cells from inside the body.
Paper: Cell
https://www.cell.com/cell/fulltext/S0092-8674(26)00643-4
Summary: ScienceDaily
https://www.sciencedaily.com/releases/2026/06/260620100317.htm
#Immunotherapy #Cancer #StemCells #CellTherapy #Biotech
Cell
Expansion and CAR engineering of granulocyte-monocyte progenitors for cellular immunotherapy
This study establishes granulocyte-monocyte progenitors as a renewable upstream platform
for engineered myeloid cell therapy and introduces a CAR-Fc strategy that couples
direct tumor targeting to host antigen-presenting cell activation.
for engineered myeloid cell therapy and introduces a CAR-Fc strategy that couples
direct tumor targeting to host antigen-presenting cell activation.
👍6🔥6👀6⚡2😁2
This media is not supported in your browser
VIEW IN TELEGRAM
🌞 The Sun Is Waking Up Fast — A Major X-Class Solar Flare Could Happen at Any Moment
Solar activity has accelerated dramatically over the past 48 hours, and space weather scientists are watching closely.
The number of solar flares has surged from 5 per day two days ago, to 8 yesterday, and 17 within the last 24 hours. This morning, the Sun produced its first M-class flares of the current activity spike — leaving only the most powerful category, X-class, yet to appear.
What’s making scientists especially cautious is the location of the Sun’s largest active region of 2026. It is currently facing almost directly toward Earth. Surprisingly, despite its enormous size, it has not yet produced a single X-class flare. Earlier this year, another large sunspot group generated five X-class flares, including the year’s strongest event, X8.1. That makes the current quietness look more like the calm before the storm than a sign of stability.
Space-based observations reveal an even more intriguing picture. Two giant sunspot groups that appear separate on the solar surface are actually connected high above it in the corona, forming a single, highly complex magnetic system. These intertwined magnetic fields continuously exchange energy. On one hand, this can relieve local magnetic stress. On the other, it effectively creates one enormous energy reservoir capable of producing an exceptionally powerful eruption.
Predicting exactly when that energy will be released remains one of the biggest challenges in solar physics. Most of the Sun’s magnetic field lies hidden beneath the visible surface, beyond direct observation, making even the most sophisticated computer models unreliable for systems this complex.
For now, an X-class solar flare could occur at virtually any time. If accompanied by a coronal mass ejection directed toward Earth, it could trigger strong geomagnetic storms, spectacular auroras at unusually low latitudes, and temporary disruptions to satellites, radio communications, and navigation systems.
🔭 The Sun is reminding us that even after centuries of observation, our nearest star can still surprise us.
#Science #Astronomy #Sun #SolarFlare #SpaceWeather #SolarStorm #Heliophysics
Solar activity has accelerated dramatically over the past 48 hours, and space weather scientists are watching closely.
The number of solar flares has surged from 5 per day two days ago, to 8 yesterday, and 17 within the last 24 hours. This morning, the Sun produced its first M-class flares of the current activity spike — leaving only the most powerful category, X-class, yet to appear.
What’s making scientists especially cautious is the location of the Sun’s largest active region of 2026. It is currently facing almost directly toward Earth. Surprisingly, despite its enormous size, it has not yet produced a single X-class flare. Earlier this year, another large sunspot group generated five X-class flares, including the year’s strongest event, X8.1. That makes the current quietness look more like the calm before the storm than a sign of stability.
Space-based observations reveal an even more intriguing picture. Two giant sunspot groups that appear separate on the solar surface are actually connected high above it in the corona, forming a single, highly complex magnetic system. These intertwined magnetic fields continuously exchange energy. On one hand, this can relieve local magnetic stress. On the other, it effectively creates one enormous energy reservoir capable of producing an exceptionally powerful eruption.
Predicting exactly when that energy will be released remains one of the biggest challenges in solar physics. Most of the Sun’s magnetic field lies hidden beneath the visible surface, beyond direct observation, making even the most sophisticated computer models unreliable for systems this complex.
For now, an X-class solar flare could occur at virtually any time. If accompanied by a coronal mass ejection directed toward Earth, it could trigger strong geomagnetic storms, spectacular auroras at unusually low latitudes, and temporary disruptions to satellites, radio communications, and navigation systems.
🔭 The Sun is reminding us that even after centuries of observation, our nearest star can still surprise us.
#Science #Astronomy #Sun #SolarFlare #SpaceWeather #SolarStorm #Heliophysics
👍22🔥11👀6😁4
🦴 A Single Injection Reversed Osteoarthritis in Animals — Human Trials Could Be Next
Osteoarthritis affects roughly one in six people over 30 worldwide. It slowly destroys cartilage — the smooth tissue that keeps bones from grinding against each other — and can eventually damage the bone underneath. Today, most treatments still focus on managing pain, reducing inflammation, or replacing the joint when damage becomes severe.
A Colorado research team is trying to change that.
Scientists from CU Boulder, CU Anschutz, and Colorado State University have developed two experimental regenerative therapies designed not just to relieve symptoms, but to push damaged joints to repair themselves.
The first approach repurposes an existing FDA-approved drug and delivers it through a patented particle system injected directly into the joint. Instead of releasing everything at once, the particles provide therapeutic bursts over several months.
The second is more like a biological repair kit: a cocktail of engineered proteins injected arthroscopically, where it hardens in place and recruits the body’s own progenitor cells to patch defects in cartilage and bone.
The early results are striking. In animal studies, a single injection restored arthritic and injured joints to a healthy state within four to eight weeks. The biomaterial patch produced what lead researcher Stephanie Bryant described as “full regeneration and repair of the defect.” The therapies also showed regenerative effects in human cells taken from patients undergoing joint replacement surgery.
ARPA-H has now advanced the project into its next phase, backed by up to $33.5 million under its NITRO program. A new company, Renovare Therapeutics, has also been formed to move the technology toward commercialization. If the next studies go according to plan, first human trials could begin within about 18 months.
The caveat is important: this is not a treatment available to patients yet. The strongest results so far come from animal models and lab-tested human cells, and the animal findings are still expected to be published in a peer-reviewed journal. Safety, dosing, durability, and real clinical benefit in people all remain to be proven.
But the direction is fascinating.
For hundreds of millions of people with osteoarthritis, the current choice is often painfully limited: manage symptoms for years, or eventually replace the joint. A minimally invasive therapy that tells the body to rebuild cartilage and bone would be a major shift — from pain control to true regeneration.
The big takeaway: the future of chronic joint disease may not be better painkillers. It may be teaching the body how to repair itself again.
📄📄 CU Boulder article: https://www.colorado.edu/today/2026/04/06/simple-shot-shows-promise-reverse-osteoarthritis-within-weeks
📰 ScienceDaily summary: https://www.sciencedaily.com/releases/2026/06/260619101356.htm
#Osteoarthritis #RegenerativeMedicine #Arthritis #Biotech #MedicalScience
Osteoarthritis affects roughly one in six people over 30 worldwide. It slowly destroys cartilage — the smooth tissue that keeps bones from grinding against each other — and can eventually damage the bone underneath. Today, most treatments still focus on managing pain, reducing inflammation, or replacing the joint when damage becomes severe.
A Colorado research team is trying to change that.
Scientists from CU Boulder, CU Anschutz, and Colorado State University have developed two experimental regenerative therapies designed not just to relieve symptoms, but to push damaged joints to repair themselves.
The first approach repurposes an existing FDA-approved drug and delivers it through a patented particle system injected directly into the joint. Instead of releasing everything at once, the particles provide therapeutic bursts over several months.
The second is more like a biological repair kit: a cocktail of engineered proteins injected arthroscopically, where it hardens in place and recruits the body’s own progenitor cells to patch defects in cartilage and bone.
The early results are striking. In animal studies, a single injection restored arthritic and injured joints to a healthy state within four to eight weeks. The biomaterial patch produced what lead researcher Stephanie Bryant described as “full regeneration and repair of the defect.” The therapies also showed regenerative effects in human cells taken from patients undergoing joint replacement surgery.
ARPA-H has now advanced the project into its next phase, backed by up to $33.5 million under its NITRO program. A new company, Renovare Therapeutics, has also been formed to move the technology toward commercialization. If the next studies go according to plan, first human trials could begin within about 18 months.
The caveat is important: this is not a treatment available to patients yet. The strongest results so far come from animal models and lab-tested human cells, and the animal findings are still expected to be published in a peer-reviewed journal. Safety, dosing, durability, and real clinical benefit in people all remain to be proven.
But the direction is fascinating.
For hundreds of millions of people with osteoarthritis, the current choice is often painfully limited: manage symptoms for years, or eventually replace the joint. A minimally invasive therapy that tells the body to rebuild cartilage and bone would be a major shift — from pain control to true regeneration.
The big takeaway: the future of chronic joint disease may not be better painkillers. It may be teaching the body how to repair itself again.
📄📄 CU Boulder article: https://www.colorado.edu/today/2026/04/06/simple-shot-shows-promise-reverse-osteoarthritis-within-weeks
📰 ScienceDaily summary: https://www.sciencedaily.com/releases/2026/06/260619101356.htm
#Osteoarthritis #RegenerativeMedicine #Arthritis #Biotech #MedicalScience
CU Boulder Today
A simple shot shows promise to reverse osteoarthritis within weeks
A CU Boulder-led team has developed a suite of new therapies aimed at reversing osteoarthritis in a single injection. With animal studies showing promise and
🔥16👍7😁5👀4
🔭 Astronomers Have Just Opened a New Map of the Black Hole Universe
The LIGO–Virgo–KAGRA collaboration has released GWTC-5.0 — the largest gravitational-wave catalog ever assembled.
It adds 161 new detections from the O4b observing run, bringing the total number of observed gravitational-wave events since 2015 to 390. What began as a handful of almost unbelievable signals has now become a real population study of black holes.
The jump is dramatic. The fourth observing run alone now accounts for roughly 75% of all gravitational-wave events detected so far. In LIGO’s first observing run, scientists detected just three events over about four months. Now the network is catching around 3–4 black hole mergers per week.
And the data are starting to reveal something deeper: black holes do not all form the same way.
Some likely come from pairs of massive stars born together. Others may meet later inside dense stellar clusters. The most intriguing group are “second-generation” black holes — objects that were themselves created in earlier black hole mergers, then merged again. Their clue is unusually rapid spin, which can act like a fingerprint of these repeated cosmic collisions.
A few highlights from the new catalog:
— 390 gravitational-wave events in total
— 161 new detections from O4b
— Black hole masses clustering around ~10 and ~35 solar masses
— Evidence for second-generation black holes from hierarchical mergers
— Best sky localization yet: one event narrowed down to just 6 square degrees
— Clearest black hole signal ever recorded, with a signal-to-noise ratio of 76.9
— A new gravitational-wave measurement of the Hubble constant with ~25% improved precision
Why it matters: gravitational-wave astronomy is no longer just about detecting rare cosmic “chirps.” It is becoming a census of an invisible universe — one that can study black hole populations, test general relativity, probe stellar evolution, and even help measure the expansion rate of the cosmos.
The big takeaway: astronomy no longer needs light alone. We are now listening to spacetime itself — and the signal is turning into a map.
📄 LIGO press release: https://www.ligo.caltech.edu/news/ligo20260526
📖 Summary: https://www.sciencealert.com/lost-world-of-gravitational-waves-reveals-the-origins-of-black-holes
#GravitationalWaves #BlackHoles #LIGO #Astrophysics #Cosmology
The LIGO–Virgo–KAGRA collaboration has released GWTC-5.0 — the largest gravitational-wave catalog ever assembled.
It adds 161 new detections from the O4b observing run, bringing the total number of observed gravitational-wave events since 2015 to 390. What began as a handful of almost unbelievable signals has now become a real population study of black holes.
The jump is dramatic. The fourth observing run alone now accounts for roughly 75% of all gravitational-wave events detected so far. In LIGO’s first observing run, scientists detected just three events over about four months. Now the network is catching around 3–4 black hole mergers per week.
And the data are starting to reveal something deeper: black holes do not all form the same way.
Some likely come from pairs of massive stars born together. Others may meet later inside dense stellar clusters. The most intriguing group are “second-generation” black holes — objects that were themselves created in earlier black hole mergers, then merged again. Their clue is unusually rapid spin, which can act like a fingerprint of these repeated cosmic collisions.
A few highlights from the new catalog:
— 390 gravitational-wave events in total
— 161 new detections from O4b
— Black hole masses clustering around ~10 and ~35 solar masses
— Evidence for second-generation black holes from hierarchical mergers
— Best sky localization yet: one event narrowed down to just 6 square degrees
— Clearest black hole signal ever recorded, with a signal-to-noise ratio of 76.9
— A new gravitational-wave measurement of the Hubble constant with ~25% improved precision
Why it matters: gravitational-wave astronomy is no longer just about detecting rare cosmic “chirps.” It is becoming a census of an invisible universe — one that can study black hole populations, test general relativity, probe stellar evolution, and even help measure the expansion rate of the cosmos.
The big takeaway: astronomy no longer needs light alone. We are now listening to spacetime itself — and the signal is turning into a map.
📄 LIGO press release: https://www.ligo.caltech.edu/news/ligo20260526
📖 Summary: https://www.sciencealert.com/lost-world-of-gravitational-waves-reveals-the-origins-of-black-holes
#GravitationalWaves #BlackHoles #LIGO #Astrophysics #Cosmology
LIGO Lab | Caltech
GWTC-5.0: Updated LIGO–Virgo–KAGRA Catalog sets new records in precision gravitational wave astronomy
The LIGO–Virgo–KAGRA Collaboration released today a new catalog of gravitational wave events. A total of 161 events, detected between April 2024 and the end of January 2025, have been added to the collection, bringing the total number of gravitational wave…
🔥11👍9👀5😁4
⏳ Scientists Found a Way to Make Quantum Time Look Like It Runs Backward
At our everyday scale, time has a clear direction. Eggs break, coffee cools, and no one has ever un-spilled a glass of water.
But at the quantum scale, the story is stranger. Many fundamental equations work just as well forward or backward in time. The arrow of time appears when a system is measured — because measurement randomly disturbs its state.
Now researchers from Los Alamos National Laboratory, NIST, and the University of Maryland have developed a theoretical control method that can reshape that arrow.
Their idea is based on a “control Hamiltonian” — a precisely designed sequence of fields and pulses that mimics and counteracts the random disturbance caused by quantum measurement. With the right feedback, the system’s trajectory can be made to look as if it is moving backward rather than forward.
To be clear: this is not a time machine.
No one reversed time for people, objects, or the universe. The work shows that, in a monitored quantum system, the appearance of time’s direction can be weakened, blurred, or even inverted.
The team also used the framework to design a quantum version of Maxwell’s demon — a measurement-powered engine that could, in principle, extract useful energy from the act of observation itself. That energy might one day help drive quantum processes or be stored in a quantum battery.
The caveat: this is still mostly a theoretical result. Experimental tests are planned for superconducting qubits, where fast feedback and quantum Maxwell’s demon setups are already technically plausible.
Why it matters: better control over quantum measurement could help with quantum state preparation, more stable quantum computers, and new ways to manage energy at the smallest scales.
The big takeaway: the arrow of time may feel absolute in daily life — but in the quantum world, it can become something engineers may learn to control.
📄 Original paper: https://link.aps.org/doi/10.1103/l18s-9vmh
📖 Los Alamos summary: https://www.lanl.gov/media/news/0319-reshaping-quantum-arrow
📖 ScienceDaily summary: https://www.sciencedaily.com/releases/2026/06/260625014802.htm
#QuantumPhysics #ArrowOfTime #QuantumComputing #Physics
At our everyday scale, time has a clear direction. Eggs break, coffee cools, and no one has ever un-spilled a glass of water.
But at the quantum scale, the story is stranger. Many fundamental equations work just as well forward or backward in time. The arrow of time appears when a system is measured — because measurement randomly disturbs its state.
Now researchers from Los Alamos National Laboratory, NIST, and the University of Maryland have developed a theoretical control method that can reshape that arrow.
Their idea is based on a “control Hamiltonian” — a precisely designed sequence of fields and pulses that mimics and counteracts the random disturbance caused by quantum measurement. With the right feedback, the system’s trajectory can be made to look as if it is moving backward rather than forward.
To be clear: this is not a time machine.
No one reversed time for people, objects, or the universe. The work shows that, in a monitored quantum system, the appearance of time’s direction can be weakened, blurred, or even inverted.
The team also used the framework to design a quantum version of Maxwell’s demon — a measurement-powered engine that could, in principle, extract useful energy from the act of observation itself. That energy might one day help drive quantum processes or be stored in a quantum battery.
The caveat: this is still mostly a theoretical result. Experimental tests are planned for superconducting qubits, where fast feedback and quantum Maxwell’s demon setups are already technically plausible.
Why it matters: better control over quantum measurement could help with quantum state preparation, more stable quantum computers, and new ways to manage energy at the smallest scales.
The big takeaway: the arrow of time may feel absolute in daily life — but in the quantum world, it can become something engineers may learn to control.
📄 Original paper: https://link.aps.org/doi/10.1103/l18s-9vmh
📖 Los Alamos summary: https://www.lanl.gov/media/news/0319-reshaping-quantum-arrow
📖 ScienceDaily summary: https://www.sciencedaily.com/releases/2026/06/260625014802.htm
#QuantumPhysics #ArrowOfTime #QuantumComputing #Physics
Physical Review X
Reshaping the Quantum Arrow of Time
A quantum control Hamiltonian is developed that can blur or even reverse the perceived arrow of time in monitored systems, enabling new ways to emulate backward-in-time dynamics and extract energy to power measurement-driven engines.
⚡16
🧠 Scientists May Have Found a Missing Link in How Alzheimer’s Kills Brain Cells
For decades, researchers have known that toxic proteins build up in the brains of people with Alzheimer’s disease and other neurodegenerative disorders. But one key question has remained stubbornly difficult:
How exactly does that protein buildup turn into the death of neurons?
A team from King’s College London and the UK Dementia Research Institute has now identified a previously overlooked form of cell death called karyoptosis.
Unlike apoptosis — the well-known “programmed cell death” pathway — karyoptosis appears to attack the cell’s control center first: the nucleus. Under proteotoxic stress, when misfolded proteins accumulate and clump together, the nuclear structure begins to fail. The nucleus shrivels, its supporting scaffold destabilizes, and the cell eventually dies.
The researchers analyzed around 3,000 brain cells from 28 patients with either end-stage Alzheimer’s disease or frontotemporal dementia. In the frontal cortex of Alzheimer’s patients, 35% of cells showed signs of karyoptosis, compared with 15% in healthy older controls.
The key molecular switch appears to involve p38 MAP kinase and LaminB1.
LaminB1 normally helps maintain the structure of the nuclear envelope — the “scaffolding” that keeps the nucleus intact. When toxic proteins accumulate, p38 MAP kinase can destabilize LaminB1, triggering nuclear collapse. In laboratory experiments, blocking this pathway reduced markers of karyoptosis in neurons.
That does not mean scientists have found a cure for Alzheimer’s. This is still early-stage work, and the protective effect has been shown mainly in experimental models, not in living human patients.
But the shift in thinking is important.
Instead of only trying to remove toxic proteins such as amyloid or tau, future therapies might also try to protect neurons from the damage those proteins trigger. In other words: even if the toxic proteins are still there, it may be possible to interrupt the death signal before the cell is lost.
The same mechanism also appears relevant to frontotemporal dementia and ALS models, suggesting karyoptosis may be part of a broader pathway across neurodegenerative disease.
Dementia affects around 57 million people worldwide, and Alzheimer’s disease accounts for roughly 60–70% of cases. Any new route to slowing neuron loss matters — but this one is still a biological roadmap, not a treatment.
The big takeaway:
Sometimes the most important question is not only “what causes the damage?”
It is: what chain reaction does the damage trigger — and can we stop it in time?
Original paper: https://www.nature.com/articles/s41467-026-73802-w
Summary: https://www.sciencedaily.com/releases/2026/06/260626124701.htm
#Alzheimers #Neuroscience #Dementia #BrainHealth #Science
For decades, researchers have known that toxic proteins build up in the brains of people with Alzheimer’s disease and other neurodegenerative disorders. But one key question has remained stubbornly difficult:
How exactly does that protein buildup turn into the death of neurons?
A team from King’s College London and the UK Dementia Research Institute has now identified a previously overlooked form of cell death called karyoptosis.
Unlike apoptosis — the well-known “programmed cell death” pathway — karyoptosis appears to attack the cell’s control center first: the nucleus. Under proteotoxic stress, when misfolded proteins accumulate and clump together, the nuclear structure begins to fail. The nucleus shrivels, its supporting scaffold destabilizes, and the cell eventually dies.
The researchers analyzed around 3,000 brain cells from 28 patients with either end-stage Alzheimer’s disease or frontotemporal dementia. In the frontal cortex of Alzheimer’s patients, 35% of cells showed signs of karyoptosis, compared with 15% in healthy older controls.
The key molecular switch appears to involve p38 MAP kinase and LaminB1.
LaminB1 normally helps maintain the structure of the nuclear envelope — the “scaffolding” that keeps the nucleus intact. When toxic proteins accumulate, p38 MAP kinase can destabilize LaminB1, triggering nuclear collapse. In laboratory experiments, blocking this pathway reduced markers of karyoptosis in neurons.
That does not mean scientists have found a cure for Alzheimer’s. This is still early-stage work, and the protective effect has been shown mainly in experimental models, not in living human patients.
But the shift in thinking is important.
Instead of only trying to remove toxic proteins such as amyloid or tau, future therapies might also try to protect neurons from the damage those proteins trigger. In other words: even if the toxic proteins are still there, it may be possible to interrupt the death signal before the cell is lost.
The same mechanism also appears relevant to frontotemporal dementia and ALS models, suggesting karyoptosis may be part of a broader pathway across neurodegenerative disease.
Dementia affects around 57 million people worldwide, and Alzheimer’s disease accounts for roughly 60–70% of cases. Any new route to slowing neuron loss matters — but this one is still a biological roadmap, not a treatment.
The big takeaway:
Sometimes the most important question is not only “what causes the damage?”
It is: what chain reaction does the damage trigger — and can we stop it in time?
Original paper: https://www.nature.com/articles/s41467-026-73802-w
Summary: https://www.sciencedaily.com/releases/2026/06/260626124701.htm
#Alzheimers #Neuroscience #Dementia #BrainHealth #Science
Nature
Karyoptosis mediates cell death and neurodegeneration upon proteotoxic stress
Nature Communications - This study characterizes karyoptosis as a mechanism of neuronal cell death triggered by proteotoxic stress and executed through disintegration of the nuclear lamina, which...
🔥5
🔒 GPT-5.6 Is Here — But Most People Can’t Use It Yet
OpenAI has launched GPT-5.6 — its next-generation model family — but this is not a normal public rollout.
The new lineup comes in three tiers:
Sol — the flagship model, built for the hardest reasoning, coding, cyber and science tasks
Terra — a balanced lower-cost model, positioned as competitive with GPT-5.5 at roughly half the price
Luna — the fastest and cheapest tier for high-volume workloads
But here is the real story: during the preview, GPT-5.6 is not available in ChatGPT and is not open through public self-service access. OpenAI says Sol, Terra and Luna are available only through API and Codex to a select group of trusted partners and organizations.
Why?
Cybersecurity.
OpenAI says GPT-5.6 Sol is its most capable model yet for cybersecurity tasks, including long-horizon vulnerability research and exploitation. The company also says it previewed the model and its release plans to the U.S. government before launch, and — at the government’s request — started with a limited preview for trusted partners whose participation had been shared with the government.
That is a very different model launch from the old “release first, regulate later” era.
To be clear: OpenAI says GPT-5.6 Sol does not cross its “Cyber Critical” threshold. In tests involving Chromium and Firefox, the model identified bugs and exploitation primitives, but did not autonomously produce a full-chain exploit under the tested conditions.
Still, the direction is obvious: frontier AI is moving into dual-use territory, where the same capability can help defenders find vulnerabilities faster — or help attackers do the same.
And this is not happening in isolation.
Anthropic’s Claude Fable 5 and Mythos 5 were temporarily suspended in June after U.S. export controls, then restored after new safeguards and further government coordination. Mythos 5 remains restricted to approved organizations through Project Glasswing, a controlled-access program focused on defensive cybersecurity.
The Five Eyes cyber agencies have also warned that frontier AI models could transform offensive and defensive cyber capabilities on a timeline of “months, not years.”
Nikolas Bush Take:
This may also accelerate the race for open-weight alternatives. Reflection AI recently locked in a compute deal worth up to $6.3 billion through 2029 to build open-weight frontier models. That looks like a direct market response to the risk of depending entirely on closed, gated systems.
But there is a darker possibility too: if open-weight models reach similar cyber capability, governments may move to restrict them as well.
The GPT-5.6 launch is not just another AI product update.
It may be an early signal of a new frontier AI regime: more powerful, more useful, more regulated — and much less predictable.
Sources:
https://openai.com/index/previewing-gpt-5-6-sol/
https://help.openai.com/en/articles/20001325-a-preview-of-gpt-56-sol-terra-and-luna
https://www.anthropic.com/news/redeploying-fable-5
https://www.anthropic.com/glasswing
https://www.cyber.gov.au/about-us/view-all-content/news/five-eyes-cyber-security-agencies-statement
https://techcrunch.com/2026/06/22/spacex-inks-compute-deal-with-reflection-ai-an-open-source-ai-lab/
OpenAI has launched GPT-5.6 — its next-generation model family — but this is not a normal public rollout.
The new lineup comes in three tiers:
Sol — the flagship model, built for the hardest reasoning, coding, cyber and science tasks
Terra — a balanced lower-cost model, positioned as competitive with GPT-5.5 at roughly half the price
Luna — the fastest and cheapest tier for high-volume workloads
But here is the real story: during the preview, GPT-5.6 is not available in ChatGPT and is not open through public self-service access. OpenAI says Sol, Terra and Luna are available only through API and Codex to a select group of trusted partners and organizations.
Why?
Cybersecurity.
OpenAI says GPT-5.6 Sol is its most capable model yet for cybersecurity tasks, including long-horizon vulnerability research and exploitation. The company also says it previewed the model and its release plans to the U.S. government before launch, and — at the government’s request — started with a limited preview for trusted partners whose participation had been shared with the government.
That is a very different model launch from the old “release first, regulate later” era.
To be clear: OpenAI says GPT-5.6 Sol does not cross its “Cyber Critical” threshold. In tests involving Chromium and Firefox, the model identified bugs and exploitation primitives, but did not autonomously produce a full-chain exploit under the tested conditions.
Still, the direction is obvious: frontier AI is moving into dual-use territory, where the same capability can help defenders find vulnerabilities faster — or help attackers do the same.
And this is not happening in isolation.
Anthropic’s Claude Fable 5 and Mythos 5 were temporarily suspended in June after U.S. export controls, then restored after new safeguards and further government coordination. Mythos 5 remains restricted to approved organizations through Project Glasswing, a controlled-access program focused on defensive cybersecurity.
The Five Eyes cyber agencies have also warned that frontier AI models could transform offensive and defensive cyber capabilities on a timeline of “months, not years.”
Nikolas Bush Take:
The most important shift here is not the benchmark. It is access.
For years, the promise of AI was democratization: more people getting access to more intelligence at lower cost. But the frontier is starting to split into layers:
— models everyone can use
— models enterprises can buy
— models only vetted partners can access
— models governments may effectively influence before release
The cybersecurity argument is real. A system that can help find serious vulnerabilities at scale is not just another productivity tool. It is dual-use infrastructure.
But the governance question is just as real:
Who decides who is “trusted”?
Are the criteria transparent?
Can decisions be appealed?
Are restrictions temporary — or do they become the new default?
This may also accelerate the race for open-weight alternatives. Reflection AI recently locked in a compute deal worth up to $6.3 billion through 2029 to build open-weight frontier models. That looks like a direct market response to the risk of depending entirely on closed, gated systems.
But there is a darker possibility too: if open-weight models reach similar cyber capability, governments may move to restrict them as well.
The GPT-5.6 launch is not just another AI product update.
It may be an early signal of a new frontier AI regime: more powerful, more useful, more regulated — and much less predictable.
Sources:
https://openai.com/index/previewing-gpt-5-6-sol/
https://help.openai.com/en/articles/20001325-a-preview-of-gpt-56-sol-terra-and-luna
https://www.anthropic.com/news/redeploying-fable-5
https://www.anthropic.com/glasswing
https://www.cyber.gov.au/about-us/view-all-content/news/five-eyes-cyber-security-agencies-statement
https://techcrunch.com/2026/06/22/spacex-inks-compute-deal-with-reflection-ai-an-open-source-ai-lab/
OpenAI
Previewing GPT-5.6 Sol: a next-generation model
OpenAI previews GPT-5.6 Sol, a next-generation model with stronger capabilities in coding, science, and cybersecurity, paired with its most advanced safety stack.
👍9👀4⚡1