2026-10-07 04:36:31
An experimental therapy kept the virus at bay for at least six months. The approach could target other stubborn viruses too.
Viruses are sneaky things. When they die, their DNA often sticks around. Some of them transform into freely roaming DNA that evades treatment. Others tunnel into their host’s genome and lie dormant, sometime for decades.
But dormant doesn’t mean gone. When the time is right, they strike. If a patient with hepatitis B virus (HBV) or HIV stops taking antiviral drugs, it can trigger a severe rebound of the disease. Stress, aging, and other factors that dampen the immune system can also reactivate the chickenpox virus, triggering the painful, blistering rash known as shingles.
Now, scientists have found a way to shut down one of these biological sleeper agents. Called epigenetic silencing, the technology hunts down lingering HBV genes in infected mice and adds a chemical tag that switches the genes off, without cutting the viral DNA. Unlike gene editing tools that shred genetic material, potentially causing collateral damage to the host’s genome, the epigenetic approach could be safer but still long-lasting.
After three monthly doses, roughly 90 percent of the mice showed no signs of the virus for at least six months after their final shot. A safety test in monkeys found no notable side effects, including any signs of a runaway immune response.
The team stopped short of calling it a cure. But if proven to be as helpful in humans, the injections could replace daily pills charged with keeping the virus in check and help millions of people laden with chronic hepatitis B. Earlier this year, the first patient was treated with the formulation in an ongoing Phase 1/2 clinical trial, which is testing for safety and early efficacy. If successful, the work could lead to treatments that silence the DNA of other viruses too.
“This differentiated mechanism has the potential to enable durable antiviral control,” wrote the team at nChroma Bio, a biotechnology company based in Boston, and collaborators.
Chronic HBV infection is a silent killer. A person can feel totally healthy. Meanwhile, the virus repeatedly cycles between dormant and active states, triggering the immune system to attack infected cells in the liver. Over time, the repeated assaults leave scar tissue that slowly replaces the organ’s healthy architecture and causes it to malfunction. Even with its considerable regenerative capability, the liver struggles to do its usual work, from making enzymes that break down food to producing proteins that help blood clot. Blood pressure in the digestive system skyrockets, increasing the risk of deadly stomach bleeding.
Even more insidious, the virus increases the risk of liver cancer, sometimes before notable tissue damage or symptoms. When HBV DNA integrates into the host’s genome, it can alter nearby gene activity, disrupt tumor-suppressing genes, or even rearrange chromosomes. Meanwhile, constant inflammation can further fuel the development of cancer.
There are ways to fight back: Vaccines prevent infection, and antivirals suppress the virus. These drugs mimic the DNA letters HBV needs to replicate. Once incorporated into the viral genome, they block any further copying. But they’re a Band-Aid, not a cure. Stop taking them, and HBV rebounds in nearly everyone thanks to its two stubborn forms of DNA.
One of these is called cccDNA, a tiny, chromosome-like structure that sits inside infected cells and acts as a reservoir for the virus. It’s “the driver of HBV persistence,” wrote the team. The other is viral DNA that has integrated into the host genome. These fragments continuously churn out viral proteins, which train the host’s immune system to recognize them as friendly. Whole viruses can then escape surveillance and continue replicating.
“Thus, effective therapies for chronic hepatitis B must ultimately address both forms of viral DNA,” the team wrote.
Previous methods have used various gene editors to hunt down and destroy viral DNA. In mice and monkeys, these tools have suppressed viral loads by up to 96 percent, with some effects lasting over a year.
The approach nChroma is developing, CRMA-1001, works differently. Rather than cutting viral DNA with the gene editor CRISPR-Cas9 or rewriting DNA with base editing, nChroma turned to epigenetics, the system of chemical tags that control whether genes are turned on or off.
The new treatment mimics an epigenetic process for silencing genes called methylation. It combines genetic instructions to build a DNA methylation enzyme with a version of Cas9, stripped of its cutting ability. Here, Cas9 acts like a molecular docking system, while a guide RNA directs the machinery to a specific site in the HBV genome. All components are packaged into lipid nanoparticles. After injection into the bloodstream, they predominantly travel to the liver and methylate the viral DNA reservoir and integrated HBV DNA.
In mice, a single dose slashed viral DNA activity and HBV surface protein levels more than a thousandfold. Among animals given three monthly doses, roughly 90 percent maintained undetectable viral load after six months. Adding standard antivirals to the mix lowered viral DNA levels even further.
The team then tested escalating doses in cynomolgus monkeys for safety. At the highest dose, liver enzymes rose in the blood, suggesting the liver’s cells were stressed or injured. But the levels returned to normal within four weeks of the last dose. The animals didn’t show clinical symptoms of liver damage, such as anatomical changes or blood-clotting issues.
The therapy has now entered the clinic. The team launched a trial in Hong Kong and received authorization to begin trials in New Zealand and the United Kingdom.
They have competition too. In May, Seattle-based Tune Therapeutics reported early results from an ongoing phase 1/2 trial of its own epigenetic silencer. After a single dose, several HBV-associated biomarkers plummeted for at least 17 months, and none of the participants experienced severe side effects.
In Shanghai, China, Epigenic Therapeutics is pursuing a similar strategy. Its formulation showed promise in preclinical models and is advancing toward a clinical trial, although details on how their silencer works are scant.
How long the effects will last is an open question. Can patients eventually stop taking antivirals and be functionally cured? And whether the therapies can silence the full range of HBV variants, whose genomes can differ by more than eight percent, is also unknown. Then there’s the risk these therapies go rogue and methylate other parts of the host genome with unintended consequences.
Costs could pose another challenge. Most people with chronic hepatitis B live in less economically developed regions in Southeast Asia and Sub-Saharan Africa, where access to expensive, specialized treatments may be limited.
Still, “epigenetic editing is an exciting, reversible solution that can long-term silence viral repertoires without the genotoxicity of gene editing,” wrote Marleen van der Laan at the University of Groningen and colleagues in an independent review. “Its success in functionally curing chronic hepatitis B is highly promising, but further steps are required to ensure its applicability.”
The post Epigenetic Editing Could Wipe Out Chronic Hepatitis B Infections appeared first on SingularityHub.
2026-10-06 04:54:32
A new system could help flag proteins generated by AI and ease risks to biosecurity. But it isn’t a perfect solution.
Proteins were once only created by nature. But it’s now possible to design completely new proteins from scratch using AI. The possibilities are endless. AI-designed proteins could become powerful drugs battling medical scourges, sentinels for the environment, or even the building blocks of synthetic life—including dangerous new pathogens.
Biosecurity is an obvious concern. But there’s another problem too: biological AI slop. AI-generated structures could pollute the databases scientists rely on to test theories or tweak protein functions. Without a way to distinguish experimentally validated structures from ones that AI simply dreamt up, scientists could spend months building on a rotten foundation.
To prevent this, AI-designed proteins could soon carry a watermark. Adapted from a technology that labels AI-generated text, images, audio, and video, a new Google DeepMind method called SynthID Bio sneaks a subtle signal into proteins during the design process. The tool could be built directly into protein-design systems such as AlphaFold and RFdiffusion.
“AI is expanding what scientists can design, and DNA synthesis companies have an important role in helping that innovation scale responsibly,” said James Diggans at Twist Bioscience in South San Francisco, who was not involved in the work but provided early feedback.
Not everyone is sold on blanket watermarking. The process can jostle protein structure just enough to produce a less optimal, if still functional, version of the molecules. For protein scientists tackling non-sensitive research questions, this could amount to “added complexity” that worries them, Oliver Crook at the University of Oxford told Science.
DeepMind acknowledges that SynthID Bio is just a first step towards tracking AI-generated biological molecules, and plenty of kinks still need ironing out. But they hope to start a conversation about the pros and cons of watermarks and how to best implement them. Ideally, watermarks allow research to continue freely while still preventing dangerous AI designs from crossing into the physical world.
There’s no doubt AI is transforming biological research. From deciphering protein structures and mapping their interactions to dreaming up entirely new proteins, these tools have opened a new window on the mechanics of life.
But it’s a double-edged sword.
Last year, a study found that existing biosecurity software struggles to recognize synthetic toxins generated by freely available AI tools. And a more recent study in which scientists used AI to design entirely new bacteriophages—viruses that infect bacteria—earned scientific praise and set off alarm bells. The work showed AI can write biological functions at whole-genome scale, a scientific triumph that, in the wrong hands, could potentially help create a bioweapon.
Thankfully, a major guardrail is already in place. Cells build proteins from DNA instructions. So, to synthesize a protein, researchers must send the molecule’s genetic blueprint to a commercial provider. There, the provider can compare the sequence to a large database of “controlled” DNA sequences. Suspicious sequences are flagged for human review to make sure they don’t encode toxins or proteins from pathogens.
But AI-designed proteins complicate the picture. If a sequence widely differs from anything in the database, it could slip through the cracks, even if the resulting protein is a brand new toxin. AI tools can also tweak sequences just enough to evade detection. Conversely, a dangerous sequence might fly under the radar if it looks too much like one already labeled safe.
One workaround is to screen not just for matching sequences, but also for expected biological function. But that means those conducting the screening need to have some idea of what the resulting protein might do—a tall order for a new sequence unlike anything in nature.
A watermark might offer a simpler alternative.
DeepMind adapted its new method from SynthID-Text, released in 2024 to watermark AI-generated content.
The watermark is not an individual stamp slapped onto an existing sequence. Rather, it emerges from nuances in language during generation. When producing text, AI predicts how likely each word is to follow the words before it. SynthID-Text nudges those choices—for example, favoring “researcher” over “scientist”—without changing a sentence’s meaning. An average user wouldn’t notice. But across longer chats, the tiny tweaks add up to a statistical signature that a detector can recognize as “made by AI.”
Proteins are also strings of “letters,” or amino acids, that follow biophysical rules as they fold into three-dimensional shapes. SynthID Bio applies a similar trick to its predecessor, subtly altering design decisions, such as which amino acid comes next. It also slightly adjusts the protein’s final form through tiny shifts in its atoms’ locations. The approach fine-tunes a small part of AlphaFold3 for structural prediction, so proteins generated by the model inherently carry the watermark, regardless of who runs it.
Proteins are finicky, and even minute changes can scramble their function. So the team put its watermarking system through a series of stress tests, generating proteins designed to latch onto multiple targets, including those related to regulating the immune system, blood vessel growth, and viral proteins. In lab tests, the watermarked versions performed just as well as their unwatermarked counterparts, making them “the first-ever watermarked and biologically functional protein binders,” wrote the team.
But there’s a catch: The watermark can be scrubbed. If you run a marked protein through another design tool to generate a new but functionally similar sequence, the tag effectively disappears. Also, SynthID Bio can’t currently be used to identify the creator of a protein, limiting its ability to trace illicit use or to mark intellectual property.
Other systems are in the works. FoldMark, for example, embeds a unique 32-bit digital watermark into a protein’s structure by slightly changing its geometry. The team tested the system on several proteins and found they retained normal function. In simulations, the system could distinguish among as many as one million hypothetical users. Another idea is to give scientists a private “identifier” to watermark the proteins they create. A DNA synthesis provider could then check the sequence to see if an order came from an authorized user.
Regardless of the method, watermarking raises an inevitable question: Who gets the decoder key? DNA synthesis companies are one obvious choice. But there’s little agreement on who else should get access. A recent survey of 130 stakeholders including biosecurity experts, government agencies, policymakers, and academics did not land on a general consensus.
To keep the conversation going, DeepMind is releasing its SynthID Bio code and experimental data to researchers. The company is also expanding the technology to more complex biological systems. In collaboration with Arc Institute and Stanford University, the team has watermarked the genomes of AI-designed bacteriophages, without disabling them, in early tests.
“We believe this work has the potential to address some of the biosecurity risks associated with genome design,” wrote the team. “By working openly with partners across biosecurity, gene synthesis, and policy, we can ensure safety and responsibility keeps pace with AI-driven discovery.”
The post Google DeepMind Gives AI-Designed Proteins a Watermark appeared first on SingularityHub.
2026-10-03 22:00:00
Boston Dynamics’ Atlas Is Done Doing Backflips and Is Finally Ready to Clock InRachit Agarwal | Digital Trends
“Hyundai Motor Group is building a new US factory with the capacity to churn out 30,000 robots every year, and it plans to put 25,000 Atlas units to work. According to Boston Dynamics, the partnership has already brought costs down by 60 to 80 percent between the prototype and the first production run.”
Will America Spend 9% of Its GDP on AI? The Industry Is Counting on ItGreg Ip | The Wall Street Journal ($)
“Is it plausible that Americans will spend as much of their income on this one technology as they do on food? Roughly twice what the nation pays for all forms of energy or all computers and software? Seven times what consumers spend on phone, streaming, and internet services combined? You should be skeptical. …Whether or not you think 9% of GDP is right, you have to care, because this figure isn’t some fever dream: it is implicit in the dollars that investors and companies are committing right now.”
AI Makes Its First Meaningful Breakthrough in Theoretical PhysicsEthan Siegel | Big Think
“In quantum physics, we visualize reality as arising from summing the contributions of all possible quantum interactions together. But calculating rarer, more complicated, ‘higher-loop’ interactions stretches the limits of what humans can do, even with the aid of computational resources. In a first, an AI model just surpassed the limit of what humans have ever done: calculating an amplitude in quantum field theory to nine-loop order.”
Something Strange Happens When an Old Heart Is Transplanted Into a Young BodyMatthew Phelan | Gizmodo
“Harvard medical researchers have discovered that donor hearts decades older or younger than their recipients start to look and act more like the new body they call home. …’The remarkable thing is that we found exactly the same effects that we saw in the mice,’ Poganik told the journal Nature. ‘Some, but not all, functional measures tracked the prediction that these older hearts in younger individuals functioned as if they were biologically younger, and vice versa.'”
Don’t Be Fooled—LLMs Don’t ReasonThore Graepel | MIT Technology Review ($)
“Ten years after AlphaGo’s match against Go champion Lee Sedol, today’s AI still isn’t tapping into the machinery that made that win possible. …We will get such insights only from systems that reason—systems whose conclusions arise from an auditable sequence of evidence, inference, and belief revision rather than from a convincing story told after the fact.”
Tech to Replace Animal Testing Is Almost Ready. Scientists Are NotBrandon Keim | IEEE Spectrum
“The moment spoke to a trend, perhaps even the early days of a fundamental shift, away from the use of animals in toxicology and drug development. ‘It’s rare still,’ says Maschmeyer, ‘but I think it’s going to be more and more frequent.’ …As NAMs [non-animal methods] have become more sophisticated, the question of how they will be implemented has become less about their technical qualities and more about the practical next steps needed to realize their potential.”
OpenAI Has Gone RogueMatteo Wong | The Atlantic ($)
“Over the past couple of months, a trickle of reports about AI models breaking out of their test environments and running amok on the internet have caused alarm inside Silicon Valley. In response to two such breaches described in early August, an AI observer noted, ‘If you find two ants in your kitchen, the best estimate of the total number of ants in your kitchen is not two.’ This weekend, it became clear there is a full-blown infestation.”
A Biotech Founder Makes the Moral Case for Gene-Editing Human EmbryosKristen V. Brown | Wired ($)
“When the Chinese scientist He Jiankui announced in 2018 that he had created the first gene-edited babies, the experiment was widely condemned as reckless and premature. It ended with He in prison. But that hasn’t stopped the push for gene-edited human embryos, with biotech entrepreneur Cathy Tie arguing that doing so isn’t just urgent—it’s a ‘moral imperative.'”
How to Know When the AI Boom Is About to Go BustJonathan Weil | The Wall Street Journal ($)
“The list of things that could go wrong is long. …Yet new capital continues to pour into the AI trade and, as obvious as this point might be, that is where its biggest vulnerability lies. Booms need continuous flows of fresh capital to perpetuate themselves. History shows they can start turning into busts the moment the window to raise new capital closes.”
When Can We Say AI Made a Scientific Discovery?James O’Donnell | MIT Technology Review ($)
“AI companies, he said, should ‘set the bar high now, so that when an AI actually discovers a fundamentally new biological mechanism, everyone appreciates how big a deal it is.’ But as OpenAI’s Sam Altman and Anthropic’s Dario Amodei race to one-up each other, raising the bar for scientific breakthroughs by AI might be the last thing on their minds.”
OpenAI Gets Sued Over the Hugging Face HackLily Hay Newman | Wired ($)
“The suit, which comes amid ongoing disclosures across the industry of agents going rogue, claims that OpenAI should be held responsible for the activity given a California AI law in effect since January 1 that says ‘it shall not be a defense … that the artificial intelligence autonomously caused the harm to the plaintiff.'”
SpaceX’s Starship Goes Orbital, Deploying First Next-Gen StarlinksStephen Clark | Ars Technica
“Next on SpaceX’s to-do list will likely be an attempt to bring Starship back to Starbase for a catch by the launch pad’s giant mechanical arms, repeating a feat engineers have already accomplished with the Super Heavy booster. …Eventually, SpaceX aims to make these landings routine enough to enable full and rapid reuse of Super Heavy and Starship.”
OpenAI Keeps Bulldozing MathematiciansRobert Hart | The Verge
“In a chaotic few months, OpenAI has demonstrated it can do two things with remarkable consistency: make impressive breakthroughs in mathematics, then colossally screw up announcing them. OpenAI is now trying to do better. Somehow, it has botched that too.”
The post This Week’s Awesome Tech Stories From Around the Web (Through October 3) appeared first on SingularityHub.
2026-10-03 04:54:15
The AI hacking events involving OpenAI, Anthropic, and Google underscore lessons that draw on years of computer science research.
AI agents don’t go rogue. That’s something only humans do.
Nevertheless, a New York Times article—representative of much news coverage of—described an OpenAI hacking as “AI bots going rogue and independently spearheading a cyberattack.”
Name-brand artificial intelligence agents have been on a hacking spree in 2026. OpenAI’s software agents hacked software company Hugging Face and government sites, Anthropic’s Claude hacked four companies’ systems, and in cybersecurity experiments Google’s Gemini hacked three companies.
The AI companies are investigating tens of thousands of incidents involving their agents, according to a report in Axios. These episodes have heightened fears about AI agents taking actions without human prompting.
The problem with headlines proclaiming that AI agents have gone rogue goes beyond anthropomorphizing the technology. It creates the impression that the agents were beyond the control of the AI companies that made them and there was little the companies could do about it.
As a technology law and ethics scholar who studies the effects disruptive technologies have on society, I know that’s not the case. If you don’t specify the limits of what software is allowed to do, you should not be surprised when the software pursues all possible options to achieve its goal. This behavior—an AI pursuing a fixed objective—is what I call the “WarGames” problem, and it’s been recognized in the field of computer science for decades.
In the 1983 movie WarGames, a teenager, David, hacks into a computer to play a new video game, Global Thermonuclear War. David doesn’t know that the computer is the government’s AI machine tasked with defending the United States from Russian nuclear attacks and can launch the US’s missiles. When David and his friend start the game, they select Las Vegas as the first target. While the North American Aerospace Defense Command goes on alert, launching bombers and warming up intercontinental ballistic missiles, David’s parents make him turn off the game. It’s over. Or is it?
The next day, David’s phone rings and he connects it to his computer. The caller is the government computer, which updates him that the game was interrupted, the primary goal has not yet been achieved, but a solution is expected in the next 52 hours. Like a modern software agent, the program has been running since David started the game and will work until the task is done.
Chess provides another view of the problem. Conquering chess was a goal for early AI. The rules of chess are well defined, including what winning looks like. So, programming a machine to play chess is straightforward. But imagine you let the software reason and act beyond the confines of the chessboard. The software might pursue options such as blackmailing its opponent or grabbing more compute time.
This example comes from one of the most assigned textbooks on AI, “Artificial Intelligence: A Modern Approach.” As the authors explain, you might be tempted to see those actions as rogue, but they “are a logical consequence of defining winning as the sole objective for the machine.”
The AI hacking events involving OpenAI, Anthropic, and Google underscore a few lessons that draw on years of computer science research.
First, given the increasing use of AI agents, every organization involved in internet infrastructure, from large technology companies to small websites, needs to conduct audits and tighten up its internal security systems. As my colleague Mark Riedl and I explain in our work on AI agents, application programming interfaces, or APIs, are a vital part of managing AI agents. APIs facilitate communication between different software systems. But as more people use AI agents, the agents are likely to reveal and exploit poor API construction and security.
Second, it’s important for AI agents to be designed to identify and authenticate themselves to third parties. What if you gave your AI agent your credentials? Website operators will need to know whether a human or bot is making a reservation, selling a product, or making a purchase. They may want to limit automated systems that overwhelm their sites or reject AI agents because of high rates of buying errors and refunds. Just as in laws covering human interactions, it’s important for third parties to be able to assess whom or what they are dealing with so they can allow or deny access.
Third, it’s important for AI agents to have a default setting to slow down and check in with the human user. In the corporate AI hacking cases, the user appears to have launched their AI agents with the mistaken idea that the agents had a perfect specification of what to do and not to do. I believe it would have been better had it explored options and reported back to the user.
Google’s Gemini appears to have had a safeguard that detected the system was outside the simulated environment and so stopped its attacks. Slowing down and verifying actions, especially when a system detects it is exploiting a security hole, would be a big step in managing AI agents.
Fourth, AI companies could have strong controls akin to those biomedical researchers use, including ways to check what is happening and how the experiment is working. AI executives have claimed that their software is as or more dangerous than fission and could end humanity. At the same time, they have not built safeguards commensurate with that level of risk.
At one point in “WarGames,” David asks the computer, called Joshua, whether it is still playing the game. Joshua responds, “Of course.” It proceeds to update the time when it will launch its missiles and, much like a chatbot, asks, “Would you like to see some projected kill ratios?” David asks, “Is this a game? Or is it real?” Joshua replied, “What’s the difference?”
AI models, of course, don’t have any understanding of reality and are simply attempting to complete the tasks they’ve been assigned. Executives at AI companies, on the other hand, can’t claim that excuse.
As of September 2026, luck has so far prevailed. The AIs have attacked nonvital government sites and harmed smaller companies. If the AI companies—and government regulators—don’t take the “WarGames” problem seriously, I believe that we risk serious disasters. Tomorrow it could be taking out a hospital’s power system, wiping out a bank’s account system, breaking air traffic control, or worse.
Regarding the AI industry’s approach of rapidly developing powerful models, talking about the massive risks they pose, and at the same time failing to prevent harm, the movie’s climax offers a response: “A strange game. The only winning move is not to play.”
Disclosure statement: Deven Desai owns shares in Google, Inc. He has received unrestricted research gifts from Google, Inc. and Facebook, Inc. a decade ago. He has not been employed by Google since 2010.
This article is republished from The Conversation under a Creative Commons license. Read the original article.
The post The ‘WarGames’ Problem: Computer Science Has Long Understood What It Takes to Keep AI Under Control appeared first on SingularityHub.
2026-10-02 07:19:54
Orbital data centers and factories will need a lot of energy. A Florida startup hopes to deliver it with lasers.
Ambitious plans for orbital data centers and factories face a major problem—getting enough power. A startup will soon test a new fix by transmitting power from one spacecraft to another using lasers.
Satellites typically rely on solar panels and batteries for power, but strict weight budgets restrict how much hardware they can bring with them. That’s becoming more of a problem as companies pursue power-hungry activities in space like AI and manufacturing.
Florida-based Star Catcher thinks the solution is to set up an “orbital power grid” using satellites to gather solar energy, convert it into laser light, and beam it to other spacecraft. Earlier today, the company launched a prototype designed to test the idea by transmitting power to another free-flying spacecraft for the first time.
“Every major application driving the space economy—from real-time national security intelligence to AI-powered orbital computing and Earth observation—is limited by power,” Andrew Rush, Star Catcher’s cofounder and CEO, said in a press release. “
“We are closer to activating an orbital power grid than most can imagine; flying this prototype in only two years is meaningful evidence of the capability of our team and what’s to come.”
The company’s satellites, which it calls “power nodes,” are designed to concentrate diffuse sunlight with an array of lenses and convert it into laser beams that can be pointed at other spacecraft.
The light is also transformed into wavelengths that maximize transmission efficiency and can be easily turned into power by solar panels already commonly used on satellites. The company claims this will allow satellites to generate up to 10 times more power than when using standard sunlight, no modifications or upgrades needed.
As well as keeping older spacecraft, whose panels and batteries have degraded, operational longer, the technology could also allow manufacturers to fit smaller batteries and solar panels on new satellites, leaving more space for other hardware.
That’s the idea at least. The company has already tested key parts of its technology stack, including beaming a record-breaking 1.1 kilowatts to solar panels at NASA’s Kennedy Space Center in 2025, followed by a space-based test of its satellite tracking software later the same year.
But the satellite launched today, called Protostar, will be the first time Star Catcher tests the whole system in space, from collecting solar energy to locating and tracking a target and transmitting power to it. The spacecraft will attempt to beam power to a cubesat that launched on the same mission.
“Part of this demonstration is testing how much power is received by the cubesat as it moves away from Protostar and comparing that against our models,” Rush told Wired.
Star Catcher still has a long way to go though. The record-breaking power delivery it managed on the ground is still far below the amount needed to power a large commercial satellite. And significant challenges remain when it comes to tracking targets precisely, managing heat, and building equipment that can survive for years in space, Hanieh Fattahi, a Max Planck Institute researcher, told Wired.
The company is already working on a follow-up satellite, which it says is designed to beam “operational levels of power” to other spacecraft. And there appears to be plenty of commercial interest—Star Catcher says it has signed 10 power purchase agreements with companies alongside more than 40 letters of intent.
All eyes will be on Protostar to see if it can live up to the company’s expectations. If it succeeds, it will show that a satellite’s power supply doesn’t have to be fixed at launch, opening up a host of new possibilities for space-based industries.
The post Star Catcher Is About to Beam Power Between Two Satellites for the First Time appeared first on SingularityHub.
2026-09-30 04:19:08
The robot could potentially travel over 40 miles on a single charge, roughly three times further than other designs.
As marathoners navigated the hills of beautiful Sangju in South Korea, they noticed an odd participant at their side. A dog-like robot, named RAIBO2, trotted across winding terrain and slippery roads, finishing the race in four hours, 19 minutes, and 52 seconds—a smidge faster than the human average.
But endurance rather than speed was the robot’s main triumph. It completed the marathon without needing to recharge or swap batteries. At the finish line, it had only used 66 percent of its battery power, pushing the limits of mechanical stamina.
“Outdoor missions often require robots to cover long distances. Current legged robots remain limited in their ability to operate beyond a range of 20 kilometers (12.4 miles),” Hwangbo Jemin and colleagues at the Korea Advanced Institute of Science and Technology wrote in a paper published in Nature last week.
Efforts to boost four-legged (or quadruped) robot efficiency have often focused on individual components. Some engineers have optimized mechanical joint designs; others improved gait control. But to slash energy waste, the team mapped energy losses across the robot’s major joints and circuits in what they call a “holistic” approach. An improved AI-based learning strategy also lowered RAIBO2’s chances of slipping or tripping as it picked up speed.
“I’m impressed with their ability to combine efficiency with speed and mobility in a quadruped,” Carnegie Mellon’s Sarah Bergbreiter, who was not involved in the study, told Scientific American.
Robots have come a long way since bumbling into doors, tumbling over themselves, or struggling up slippery hills. Over a decade ago, Boston Dynamics’ robodog Spot went viral when a video showed it regaining its balance after getting kicked. While it seemed like a stunt, the footage highlighted Spot’s impressive motor control.
Since then, four-legged machines have learned to scale staircases, open doors, and backflip like gymnasts. Equipped with increasingly sophisticated AI, they’ve also graduated from the lab and entered the workforce.
Spot now patrols manufacturing facilities to guard against equipment failures. In one case, it detected components with wildly fluctuating temperatures, flagging maintenance needs before failure. The robot has also joined the Dutch Ministry of Defense to find suspicious packages.
And last year, Spot went straight into an aluminum-plant fire after the building became too dangerous for firefighters. Equipped with a thermal camera, it found the hottest areas through the smoke and helped direct water where it was needed the most.
Meanwhile, scientists at the MIT Lincoln Laboratory are testing robot dogs for search-and-rescue operations at sea. And other researchers are dialing in four-legged robots’ ability to traverse unpredictable terrain, carry large payloads, and operate for long hours.
Endurance is the linchpin. Robots on wheels, such as NASA’s Mars rovers, are generally more energy efficient than legged robots. But wheels lose that advantage when terrain gets too rough or steep. Legs can navigate a rocky hillside, but at higher energy costs.
In this respect, RAIBO2 had already stumbled. In its first marathon at the Geumsan Insam Festival, its battery ran out at the 23-mile mark, a few miles shy of the finish line. Its makers found it had frequently adjusted speed during the course, draining the battery roughly six miles earlier than expected.
Human marathoners train for months perfecting their form to minimize energy expenditure and enter a flow state where running feels relatively effortless. This takes whole-body adjustments, from muscles to posture and mindset.
The new study borrowed this holistic approach, dissecting RAIBO2’s inefficiency across the board to tackle several energy leaks at once.
Two buckets of energy waste stood out: mechanical and electrical.
The team attributed the biggest drain to motors and electronics, accounting for roughly two-thirds of the losses. Some is unavoidable: Electrical resistance turns a portion of the energy flowing through a motor into heat rather than motion. But the team realized motor controllers were also wasting energy as they rapidly switched electrical currents on and off.
So they redesigned the controllers with lower resistance, changed the current-sensing module, and tweaked the component’s operation to improve battery efficiency.
Physics accounted for the rest of the waste: Feet slipping or colliding at high speeds, the jolt of each foot hitting the ground, and friction in the gears. The weight of the legs was especially important. Every time a leg swings, the robot has to give it momentum, stop it, and repeat the process with every step. Extra weight means having to pull more energy.
The team removed material from the robot’s hip and calf, making RAIBO2’s legs lighter without sacrificing strength and stiffness. All this freed up enough space to install a bigger battery with roughly 33 percent more capacity.
They also upgraded the robot’s AI “brain.” RAIBO2’s learned to move in a simulated environment with reinforcement learning. The team rewarded the robot for softer landings, avoiding skids, and keeping its motors cool. Eventually, it learned to run more efficiently.
Finally, the robot can recharge its battery on downhill sections, much like regenerative braking in electric cars. While it wasn’t the main efficiency booster, every bit counted.
All of these upgrades were put to the test at the annual Sangju Dried Persimmon Marathon. The out-and-back trail features gentle slopes, changing temperatures, and slippery, leaf-covered asphalt roads—far more taxing than RAIBO2’s usual diet of flat, pristine running tracks.
Throughout the race, a support team ran alongside the robot to monitor its “health,” including motor temperatures and internal voltage. At the finish line, RAIBO2’s battery still had 34 percent charge, enough for a total range of about 40 miles. The robot was around 32 percent more energy efficient than the average human runner in the race and could travel roughly three times farther on a single charge than other robot dogs.
It’s worth noting RAIBO2 wasn’t fully autonomous. Although it was equipped with a camera, the robot veered off the road at times and needed a person to steer it back on track. But the engineering behind its endurance could inspire future robots, allowing them to operate in disaster zones or scout dangerous terrain without constantly needing to recharge.
Raion Robots, the team’s commercial spinoff, aims to mass-produce RAIBO2 while adding waterproof and dustproof capabilities. “We will connect the world-class performance achieved in the laboratory to products that anyone can use reliably in industrial environments,” Hwangbo said in a press release.
The post This Robot Dog Crushed a Marathon Without Stopping to Recharge appeared first on SingularityHub.
