Meta drops Muse Spark 1.1
The new version is better across the board, with a massive upgrade to agentic coding capabilities.
Meta has also launched its official API alongside the model.
💰 Pricing & Perks:
Cost: $1.25 (Input) / $4.25 (Output) per 1M tokens.
Bonus: All new accounts get $20 in free API credits to start building right away.
Hardened for Nuclear Scenarios: BAE Systems’ New Space Chip
BAE Systems announced that its Endura space processor has survived some of the harshest radiation testing required for national security satellites.
The chip was put through two extreme phases: natural space radiation and a much harsher "strategic" environment—essentially simulating a nuclear blast near the spacecraft.
Here is what makes this silicon so special
The Technology Inside
Unlike older architectures that split functions across multiple microchips, Endura is a true System-on-Chip (SoC):
All-in-one: Processor, memory, and communication links are packed onto a single die.
Versatile: Features embedded FPGA blocks for specialized edge computing and secure boot capabilities.
The Process: Fabricated using BAE's radiation-hardened 45nm tech at the trusted GlobalFoundries facility in NY (a commercial Silicon-on-Insulator platform upgraded for space).
From Mars to Missile Defense
BAE's chips are no strangers to extreme environments—their previous-gen hardware is currently powering NASA's Curiosity and Perseverance rovers on Mars.
Launched in 2024 as Software Development Units, Endura is already being integrated into classified defense missions. James LaRosa, BAE’s Program Director, confirmed they are now working with multiple contractors to qualify the chip for strategic missile defense systems.
AI is failing at lunar science. Big time.
A new study led by the Southwest Research Institute (SwRI) reveals major flaws and discrepancies in lunar crater catalogs generated by Artificial Intelligence.
Here is why scientists are sounding the alarm
📉 The Deficit in Accuracy
The research team benchmarked 8 different AI-generated catalogs of lunar craters against traditional, human-compiled records. The results are highly concerning:
Under strict scientific criteria, AI accuracy plummets significantly.
In the worst cases, the error rates in AI catalogs were over 10 times higher than those made by humans.
Why does this matter?
For lunar science, it’s not just about spotting a crater. It is critical to precisely measure their exact count, coordinates, and diameters.
The Problem: Scientists estimate the age of the Moon's surface based on crater counts (crater-counting chronology).
The Risk: If an AI algorithm accidentally duplicates the same craters, it could mistakenly estimate the lunar surface to be twice as old as it actually is.
What's the solution?
The authors aren't saying we should ban machine learning. Automated crater recognition can save scientists years of tedious manual labor.
However, they are calling for unified verification standards and independent validation. Before AI catalogs can be safely trusted for peer-reviewed space research, we need strict quality control.
1687: The day physics changed forever
Isaac Newton published his Philosophia Naturalis Principia Mathematica, laying down the three laws of motion and the law of universal gravitation. These very principles remain the bedrock of celestial mechanics used to calculate satellite orbits and interplanetary trajectories today.
Fun fact: The printing was entirely funded out of pocket by astronomer Edmond Halley (yes, the comet guy), and the Royal Society's licensing permission was signed by Samuel Pepys.
1966: Testing Apollo's fuel in zero-G
NASA launched the uncrewed AS-203 mission aboard a Saturn IB rocket to study how liquid hydrogen behaves in weightlessness. The experiment proved the S-IVB stage could successfully restart its engine after drifting in orbit—a crucial capability without which the Apollo missions could never have flown from Earth to the Moon.
The twist: The rocket didn't carry a spacecraft, just an aerodynamic nose cone. The fuel tank was packed with 88 sensors and two TV cameras. At the end of the mission, engineers intentionally overpressurized the tank until it ruptured, just six hours after launch.
2016: Juno arrives at Jupiter
NASA’s Juno spacecraft successfully entered Jupiter's orbit after a critical burn at 03:53 UTC. The main engine fired for 2,102 seconds, slowing the probe down by ~542 m/s to let the gas giant’s gravity capture it.
The record: Juno became the farthest spacecraft from the Sun to run entirely on solar power instead of a nuclear RTG. To capture the faint sunlight, it uses three massive solar arrays, each about 9 meters (30 feet) long!
2023: End of an era for Ariane 5
The legendary Ariane 5 rocket flew its final mission (VA261) from Kourou, French Guiana, successfully deploying the Heinrich Hertz and Syracuse 4B satellites. This 117th launch closed a spectacular 27-year history for the heavy-lifter that famously launched the James Webb Space Telescope.
Poetic ending: The two nations behind the payloads for this farewell flight—Germany and France—were the exact first two contributors to the Ariane program decades ago.
🛰️ Blacker than space: Saving astronomy from Starlink flare
Satellites in Low Earth Orbit are ruining ground-based astronomy. Reflected sunlight leaves bright streaks across telescope images, photobombing distant galaxies and dangerous asteroids.
Astrophysicists at the University of Surrey have a solution: coating spacecraft in Vantablack 310, an ultra-black material that absorbs 98% of light and eliminates harsh, blinding glints.
📉 The alarming numbers:
There are 11,800+ satellites in LEO right now (~7,100 are Starlinks).
Global mega-constellation plans target 1.7+ million spacecraft in the future.
For the upcoming Vera C. Rubin Observatory, this light pollution could slash star detection by 7.5% and cost an extra $22M.
What's next?
SpaceX has tried visors and mirror films, but Vantablack 310 performs uniquely well across all viewing angles while enduring harsh space conditions.
The material is about to face a real-world test. It will fly on Jovian-1, a joint student CubeSat mission, to see if it survives the orbital environment and if Earth-based telescopes notice the drop in brightness.
GPT 5.6 is coming as early as tomorrow
OpenAI has promised to release all three models of the lineup this Thursday, and in the meantime, they are expanding the preview to companies outside the US.
The Coronal Heating Paradox
Why is the Sun's corona heated to 1–3 million degrees, while its actual surface registers a mere 5500 °C? This is one of the oldest unsolved mysteries in heliophysics. For decades, scientists have searched for the answer among electrons, ions, magnetic fields, and plasma waves. Now, Syed Ayaz, a graduate student at the UAH Center for Space Plasma, has introduced an ingredient to this equation that was previously thought to be impossible—dust.
The Probe as a Dust Detector
The initial logic seemed ironclad: dust grains are a million times more massive than ions and simply shouldn't survive the inferno near the Sun. However, data from the Parker Solar Probe revealed otherwise. Dust is very much present and active much closer to the star than expected.
Amusingly, there is no dedicated dust detector onboard the spacecraft. Instead, when a tiny dust grain slams into the probe at breakneck speed, it instantly vaporizes, creating a micro-cloud of charged particles. This cloud produces a sharp voltage spike on the antennas of the FIELDS instrument. Effectively, the entire spacecraft functions as one giant dust detector.
Dust and Alfvén Waves
Charged dust grains are not passive passengers. Once they acquire a charge, they begin to interact with electromagnetic fields and influence kinetic Alfvén waves, which carry electromagnetic energy through the corona and transfer it to particles. It is precisely these waves that accelerate and heat the plasma.
Ayaz demonstrated that dust affects these waves in two contrasting ways:
Mass Influence: The mass of the dust adds inertia to the plasma, slowing the waves down and allowing them to carry energy further out into the corona before dissipating.
Charge Influence: Conversely, the charge strengthens the coupling between the wave, the field, and the particles, causing the energy to be released locally as heat.
Which of these forces dominates ultimately dictates exactly where the Sun heats its own atmosphere.
Rewriting the Textbooks
While the model remains preliminary, Parker edges closer to the star with every orbit, venturing further than any spacecraft before it. If dust truly proves to be the missing puzzle piece in the coronal heating problem, the traditional picture—painted solely with electrons and ions—will have to be completely rewritten.
The Nitrogen Challenge in Lunar Farming
The primary obstacle to lunar agriculture is not a lack of light or water. Lunar regolith is chemically dead: it contains no organic matter, no ammonia, and no nitrates—meaning zero accessible nitrogen, without which plants simply cannot grow. To solve this, researchers from Tohoku University and the Japanese space agency, JAXA, have developed a method to synthesize nitrogen fertilizer right on-site, directly from the air.
Harnessing Plasma Technology
The process relies on plasma technology. A portable device passes atmospheric air through an electrical discharge to selectively produce dinitrogen pentoxide (\bm{N_2O_5}) gas, while consuming less than 100 watts of power. This gas dissolves in water with nearly 100% efficiency, converting into nitrate—the exact form of nitrogen that plants can readily absorb. The entire chemical process runs on electricity from solar panels, requiring absolutely no fossil fuels.
Transforming the Soil Chemistry
When this nitrate-infused water was added to a lunar regolith simulant, it triggered several beneficial reactions simultaneously:
pH Correction: Raw regolith is highly alkaline, with a pH of 9.09, which is nearly toxic to plant roots. The nitrate water brought it down to an almost ideal 6.76.
Nutrient Release: The solution unlocked essential calcium, magnesium, and potassium ions trapped within the soil.
Toxicity Suppression: Conversely, it suppressed toxic aluminum, which typically destroys root systems.
The visual results speak for themselves. Three months after planting, the rice grown with the nitrate solution showed significantly better development than the control plants grown in pure water. A month later, the rice began heading (flowering), proving that the fertilizer functions as a genuine nitrogen source rather than just a soil acidifier.
An Unexpected Growth-Control Bonus
Researchers discovered an unexpected bonus when spraying the foliage with the gas. \bm{N_2O_5} activates hormonal pathways that boost plant immunity while inhibiting the elongation of stems and internodes. In low gravity, where stems tend to stretch upward uncontrollably, managing crop height is a major engineering challenge. Remarkably, the very same gas that feeds the plants also solves this structural problem.
The Ring Current Mystery
For decades, physicists have debated the origin of the ring current ions—a belt of charged particles drifting around Earth thousands of kilometers above the equator. Two sources are known: the solar wind and Earth's ionosphere, the charged upper layer of the atmosphere. Usually, both contribute during geomagnetic storms. Logic suggested that if a storm was triggered by a dense solar wind, the contribution of solar ions should be significant.
May 2024 turned this logic upside down. A massive sunspot released a series of powerful coronal mass ejections; clouds of magnetized plasma merged on their way to Earth and slammed into the magnetosphere. This gave birth to the storm later named the Gannon storm—its \bm{SYM\text{-}H} index plummeted to \bm{-518\text{ nT}} (nanoteslas), making it the second strongest storm since 1981. The last time a storm of this magnitude occurred was back in November 2004.
Breakthrough Data from Arase
The Japanese satellite Arase, which has been orbiting directly in the ring current's formation zone since 2016, had been waiting for such an opportunity for over seven years. It passed through the current twice—at the beginning of the storm and near its peak—and measured the ion composition directly, something that had never been done during a superstorm before. The data was unequivocal: about 85% of the ions were oxygen from Earth's ionosphere, while the solar wind's contribution was minimal.
Oxygen is nearly an order of magnitude heavier than the protons from the solar wind, and it appears that this mass amplified the magnetic disturbance and compressed the peak of the current unusually close to the planet. At an altitude of about 16,000 kilometers, Arase recorded a 40% drop in magnetic field strength—much closer to Earth than any similar dips previously observed. In that same zone, the high-energy electrons that typically circulate there vanished completely.
Shaking the Foundations of Space Weather Forecasting
This discovery challenges the very foundation of space weather forecasting. Current models calculate storm intensity primarily based on the state of the solar wind, but it turns out that the Earth's atmosphere's readiness to release its own ions can be just as decisive.
To fill this gap, Japanese scientists are already developing the concept for the FACTORS mission—a pair of satellites designed to track exactly how atmospheric ions escape into the magnetosphere. Without this half of the equation, accurately predicting the next superstorm will be impossible.
Canadarm2 Repair Mission
Astronauts Jessica Meir and Chris Williams will venture outside the station for a six-and-a-half-hour spacewalk with a single objective: to replace a joint on the Canadarm2 robotic arm. Last month, the manipulator began exhibiting anomalous behavior; during routine operations, engineers detected elevated motor current. A faulty wrist joint was identified as the culprit. Fortunately, a spare was already onboard the station—Canadarm2 was designed from the outset with field-replaceable units.
This will mark the 280th spacewalk in ISS program history and the first in an upcoming series. Three more EVAs are scheduled through August: installing components for a new solar array, repairing electrical jumpers, and replacing a communications antenna. For Williams, this will be his second spacewalk—his first was in March, which also focused on solar panel assembly.
Aging Station and Budget Concerns
The repairs coincide with warning signs from the Aerospace Safety Advisory Panel (ASAP), NASA's independent safety board. Its chair, former astronaut Susan Helms, called the management of critical spare parts a major headache for the aging station. She warned that the temptation to slash the ISS budget ahead of its planned 2030 decommissioning remains a persistent issue. While engineers still have the risks under control, "the margin for doing so has shrunk to a concerning level." Another pressing worry is the inventory of EMU spacesuits: there are currently four operational suits, with a fifth arriving this fall, while supply chain issues for components only exacerbate equipment wear and tear.
NASA's Stance and the Russian Module Leak
ISS Operations Manager Bill Spetch downplayed these fears, maintaining that spare parts are sufficient and the spacesuits are in excellent shape. He also commented on the leak in the PrK transfer tunnel of the Russian Zvezda module. On June 5, the crew was ordered to shelter inside the docked Crew Dragon spacecraft while Russian cosmonauts prepared for a repair operation that NASA feared could compromise the module's structural integrity. Ultimately, the repair was not performed. Currently, the PrK is not leaking; the pressure is stable, having been intentionally lowered.
Meanwhile, the Crew Dragon remains configured with a fifth seat for Williams, who arrived aboard a Soyuz spacecraft back in November—just in case the crew needs to shelter again.
1997 NC1
1997 NC1 is a space rock ranging from 0.75 to 1.65 kilometers in width—roughly the size of two to four Empire State Buildings stacked on top of each other. On Saturday morning, it reached its closest point to Earth at 2.6 million kilometers. This is nearly seven times further than the Moon, so there was never any real danger.
The object was discovered nearly three decades ago by an asteroid tracking system in Hawaii, and its orbit has been precisely calculated ever since. In fact, this is exactly why NASA, ESA, and other agencies maintain a catalog of near-Earth objects: knowing who is passing by and when, well in advance, is far more critical than the flyby event itself. For an object of this size, even a tiny margin of error in its trajectory is a big deal.
How to Spot It
Anyone with binoculars or a small telescope could have spotted the asteroid as a tiny, moving dot of light against the backdrop of stars. While a modest sight, it is a rare one. The last time an object of similar size safely passed Earth was in 2022—asteroid 1994 PC1—and it came even closer.
A Once-in-a-Lifetime Event: The next time 1997 NC1 comes this close to Earth will not be until the year 2133. This means that absolutely no one reading this post today will live to see its return.
The Strategic Value
The real value lies not in the rocket itself, but in what Rocket Lab acquires. Iridium brings an operational constellation in Low Earth Orbit (LEO), over 2.55 million active subscribers, and a globally coordinated L-band frequency spectrum. This is the very connectivity that maritime fleets, aviation, and the military rely on in areas completely devoid of ground-based cell towers. Additionally, it provides alternative PNT (Positioning, Navigation, and Timing) architecture—a backup system for scenarios where GPS is jammed or unavailable.
Previously, Rocket Lab built rockets and satellites, but they flew primarily for external customers. Now, the company is closing the entire loop: designing, manufacturing, launching, and operating its own constellation. This eliminates third-party launch costs during constellation deployment and replenishment, keeping the launch margin entirely in-house. As Rocket Lab founder Peter Beck put it, the goal is to sustain the existing network while building new markets on top of it.
Financial Transformation
Financially, this completely reshapes the company's profile. In 2025, Iridium generated $871.7 million in revenue with an Operational EBITDA of $495 million—representing a 57% margin and the stable cash flow that Rocket Lab previously lacked. The cash portion of the deal is backed by a $3.6 billion bridge loan commitment from Deutsche Bank and Wells Fargo.
The Long-Term Bet: The next generation of Iridium is poised to expand toward direct-to-device (smartphone) connectivity and critical defense infrastructure. If the deal closes as planned by mid-2027, Rocket Lab will transform from a launch contractor into a fully integrated communications operator with millions of active subscribers worldwide.
1961 📸
The Thor-Ablestar rocket launched from Cape Canaveral carrying the US Navy's Transit 4A navigation satellite, which became the world's first spacecraft powered by a nuclear source—the SNAP-3 radioisotope thermoelectric generator fueled by plutonium-238. This type of generator has no moving parts and generates electricity directly from the heat of natural radioactive decay, allowing it to operate for decades without maintenance. It was thanks to this that Transit 4A served in orbit for fifteen years.
Today, such power sources are rarely installed on near-Earth spacecraft: solar panels have become cheap and efficient, while plutonium-238 remains so rare and expensive that only a few grams are produced per year, and launching radioactive cargo requires extraordinary safety measures. Therefore, nuclear power is now reserved for missions to the outer planets and interstellar space where sunlight is too scarce—ranging from the Voyagers to the Curiosity and Perseverance Mars rovers.
1995
In the STS-71 mission, the Space Shuttle Atlantis docked with an orbital station for the first time among American spacecraft and performed the first-ever crew rotation directly in orbit. Notably, during the joint flight, the docked complex became the largest man-made object in space at that time.
2024
Asteroid 2024 MK, ranging in size from 120 to 270 meters, passed by Earth at a distance of about 290,000 kilometers—closer than the Moon's orbit. What is concerning is that it was discovered just thirteen days before its closest approach: ground-based telescopes are almost incapable of spotting objects approaching from the direction of the Sun in advance. This is precisely why astronomers are pinning their hopes on new early detection tools—the Vera C. Rubin Observatory and NASA's future NEO Surveyor infrared space telescope, designed to track hazardous near-Earth objects directly from orbit.
A $4.7 Billion Fine
This post isn't strictly about AI, but since Google plays a key role in our field, it's hard to ignore.
The company has officially lost its final appeal in the EU’s highest court, upholding a record antitrust fine of €4.12 billion (~$4.68 billion). The case dates back to 2018, when the European Commission accused Google of restricting competition by requiring the pre-installation of Google Search and Chrome on Android devices as a condition for licensing Google Play.
Think of it as a hefty "tax" on Gemini's user base and their Google account data. It will be interesting to see what changes follow this ruling.
Habr just rolled out the implementation details for GigaChat 3.5 Ultra.
The devs open-sourced a 432B parameter model, fully trained in FP8 and run through a new online RL stage following classic SFT and DPO. The model got a massive upgrade in coding, math, and agentic scenarios, while generation speeds jumped up to 2.2x thanks to dual MTP heads. Throughput under load is also up by 20%.
Benchmark-wise, the model performs on par with DeepSeek V3.2, despite being 1.5x smaller. The full breakdown of the architectural framework and stability recipes are available in the article. Links to the model under the MIT license are already up on HuggingFace and GitVerse
Uranus and Neptune have borne the nickname "ice giants" for decades. The logic seemed ironclad: beneath a hydrogen-helium atmosphere lies a gargantuan mantle of water, ammonia, and methane ice, with a rocky core at the center. The only problem is that this picture has never fully aligned with what we actually observe. The magnetic fields of both planets behave strangely, and their heat distribution does not fit the icy model. This is compounded by the fact that we have only seen them up close once — Voyager 2 flew past Uranus in 1986 and Neptune in 1989. No spacecraft has visited them since.
A team from UCLA, led by Edward Young, ran a series of computer simulations of the interior structures — and came up with a completely different answer. Instead of ice, the depths of both planets harbor an ocean of molten magma composed of silicates, iron, and hydrogen.
The structure in the new model is three-layered. On top is a hydrogen-helium envelope that transfers heat upward and radiates it into space. Below that is a boundary layer made of a mixture of hydrogen, helium, magnesium, silicon monoxide, and oxygen. Even deeper lies the magma ocean itself, where in a transition zone, silicate melt literally "rains" back down. The model was tested against six observable parameters, adjusting three free parameters using the Markov Chain Monte Carlo method — including the pressure of the transition zone and the strength of convection in the envelope.
The magma theory is also supported by what these planets were originally made of. Kuiper Belt objects, which are considered close relatives to the building blocks of Uranus and Neptune, are predominantly rocky rather than icy. Protoplanetary disks around other stars are also ice-poor, suggesting that ice there is the exception rather than the rule.
There is also a bonus beyond the Solar System. The authors propose Uranus and Neptune as analogs for "sub-Neptunes" — the most common type of exoplanets, with radii between 1 and 4.5 times that of Earth, which simply do not exist in our own system for comparison. Concept missions like the Uranus Orbiter and Probe and Neptune Odyssey could verify all of this. However, neither has been approved yet — so for now, the debate over the composition of their interiors will have to rely on the old Voyager data.
To venture further into the Solar System, next-generation spacecraft may first need to refuel in Earth orbit. This requires a device that NASA calls a cryo-coupler — essentially a refueling nozzle that connects the spacecraft's tank with an orbital fuel depot. These depots are set to become the "gas stations in space." The first prototype has just undergone testing at NASA's Marshall Space Flight Center.
The challenge lies in the fact that liquid hydrogen and liquid oxygen must be kept at hundreds of degrees below zero — otherwise, they will simply boil off. Furthermore, transferring such fuel between two vehicles in orbit is something no one has ever done before. Travis Belcher, the project manager at Marshall, bluntly calls this one of the toughest engineering challenges in spaceflight, noting that materials, seals, and mechanisms must operate at their absolute limits.
Ground-based couplers, used to fuel the SLS before Artemis launches, won't work here. They disconnect with a single tug during liftoff, are manually connected before each flight, and are too bulky and not designed for space. In contrast, the coupler from L3Harris docks and undocks multiple times and is fully automated — meaning astronauts won't have to perform spacewalks to transfer the fuel.
The NASA and L3Harris team ran liquid nitrogen at minus 196 °C through the connection, testing how the metal reacts to thermal contraction and sharp temperature fluctuations. The second test focused on geometry: one half of the coupler was placed on a robotic stage that moves and rotates in any direction, simulating a misaligned docking. In reality, the spacecraft and the depot are unlikely to align perfectly.
For now, this is the earliest stage — engineers are verifying basic functionality, and couplers for specific missions will be designed and tested more rigorously later. Meanwhile, at NASA's Glenn Research Center, another part of the same dream is being worked on: the CryoFILL project, which aims to learn how to extract and liquefy oxygen directly on the Moon to refuel landers using local ice. Without a reliable way to transfer ultra-low-temperature fuel, both ideas remain nothing more than blueprints.