Last year, China saw a boom in foundation models, the do-everything large language models that underpin the AI revolution. This year, the focus has shifted to AI agents—systems that are less about responding to users’ queries and more about autonomously accomplishing things for them.  There are now a host of Chinese startups building these general-purpose digital tools, which can answer emails, browse the internet to plan vacations, and even design an interactive website. Many of these have emerged in just the last two months, following in the footsteps of Manus—a general AI agent that sparked weeks of social media frenzy for invite codes after its limited-release launch in early March.  These emerging AI agents aren’t large language models themselves. Instead, they’re built on top of them, using a workflow-based structure designed to get things done. A lot of these systems also introduce a different way of interacting with AI. Rather than just chatting back and forth with users, they are optimized for managing and executing multistep tasks—booking flights, managing schedules, conducting research—by using external tools and remembering instructions.  China could take the lead on building these kinds of agents. The country’s tightly integrated app ecosystems, rapid product cycles, and digitally fluent user base could provide a favorable environment for embedding AI into daily life.  For now, its leading AI agent startups are focusing their attention on the global market, because the best Western models don’t operate inside China’s firewalls. But that could change soon: Tech giants like ByteDance and Tencent are preparing their own AI agents that could bake automation directly into their native super-apps, pulling data from their vast ecosystem of programs that dominate many aspects of daily life in the country.  As the race to define what a useful AI agent looks like unfolds, a mix of ambitious startups and entrenched tech giants are now testing how these tools might actually work in practice—and for whom. Set the standard It’s been a whirlwind few months for Manus, which was developed by the Wuhan-based startup Butterfly Effect. The company raised $75 million in a funding round led by the US venture capital firm Benchmark, took the product on an ambitious global roadshow, and hired dozens of new employees.  Even before registration opened to the public in May, Manus had become a reference point for what a broad, consumer‑oriented AI agent should accomplish. Rather than handling narrow chores for businesses, this “general” agent is designed to be able to help with everyday tasks like trip planning, stock comparison, or your kid’s school project.  Unlike previous AI agents, Manus uses a browser-based sandbox that lets users supervise the agent like an intern, watching in real time as it scrolls through web pages, reads articles, or codes actions. It also proactively asks clarifying questions, supports long-term memory that would serve as context for future tasks. "Manus represents a promising product experience for AI agents," says Ang Li, cofounder and CEO of Simular, a startup based in Palo Alto, California, that’s building computer use agents, AI agents that control a virtual computer. “I believe Chinese startups have a huge advantage when it comes to designing consumer products, thanks to cutthroat domestic competition that leads to fast execution and greater attention to product details.” In the case of Manus, the competition is moving fast. Two of the most buzzy follow‑ups, Genspark and Flowith, for example, are already boasting benchmark scores that match or edge past Manus’s.  Genspark, led by former Baidu executives Eric Jing and Kay Zhu, links many small “super agents” through what it calls multi‑component prompting. The agent can switch among several large language models, accepts both images and text, and carries out tasks from making slide decks to placing phone calls. Whereas Manus relies heavily on Browser Use, a popular open-source product that lets agents operate a web browser in a virtual window like a human, Genspark directly integrates with a wide array of tools and APIs. Launched in April, the company says that it already has over 5 million users and over $36 million in yearly revenue. Flowith, the work of a young team that first grabbed public attention in April 2025 at a developer event hosted by the popular social media app Xiaohongshu, takes a different tack. Marketed as an “infinite agent,” it opens on a blank canvas where each question becomes a node on a branching map. Users can backtrack, take new branches, and store results in personal or sharable “knowledge gardens”—a design that feels more like project management software (think Notion) than a typical chat interface. Every inquiry or task builds its own mind-map-like graph, encouraging a more nonlinear and creative interaction with AI. Flowith’s core agent, NEO, runs in the cloud and can perform scheduled tasks like sending emails and compiling files. The founders want the app to be a “knowledge marketbase”, and aims to tap into the social aspect of AI with the aspiration of becoming “the OnlyFans of AI knowledge creators”. What they also share with Manus is the global ambition. Both Genspark and Flowith have stated that their primary focus is the international market. A global address Startups like Manus, Genspark, and Flowith—though founded by Chinese entrepreneurs—could blend seamlessly into the global tech scene and compete effectively abroad. Founders, investors, and analysts that MIT Technology Review has spoken to believe Chinese companies are moving fast, executing well, and quickly coming up with new products.  Money reinforces the pull to launch overseas. Customers there pay more, and there are plenty to go around. “You can price in USD, and with the exchange rate that’s a sevenfold multiplier,” Manus cofounder Xiao Hong quipped on a podcast. “Even if we’re only operating at 10% power because of cultural differences overseas, we’ll still make more than in China.” But creating the same functionality in China is a challenge. Major US AI companies including OpenAI and Anthropic have opted out of mainland China because of geopolitical risks and challenges with regulatory compliance. Their absence initially created a black market as users resorted to VPNs and third-party mirrors to access tools like ChatGPT and Claude. That vacuum has since been filled by a new wave of Chinese chatbots—DeepSeek, Doubao, Kimi—but the appetite for foreign models hasn’t gone away.  Manus, for example, uses Anthropic’s Claude Sonnet—widely considered the top model for agentic tasks. Manus cofounder Zhang Tao has repeatedly praised Claude’s ability to juggle tools, remember contexts, and hold multi‑round conversations—all crucial for turning chatty software into an effective executive assistant. But the company’s use of Sonnet has made its agent functionally unusable inside China without a VPN. If you open Manus from a mainland IP address, you’ll see a notice explaining that the team is “working on integrating Qwen’s model,” a special local version that is built on top of Alibaba’s open-source model.  An engineer overseeing ByteDance’s work on developing an agent, who spoke to MIT Technology Review anonymously to avoid sanction, said that the absence of Claude Sonnet models “limits everything we do in China.” DeepSeek’s open models, he added, still hallucinate too often and lack training on real‑world workflows. Developers we spoke with rank Alibaba’s Qwen series as the best domestic alternative, yet most say that switching to Qwen knocks performance down a notch. Jiaxin Pei, a postdoctoral researcher at Stanford’s Institute for Human‑Centered AI, thinks that gap will close: “Building agentic capabilities in base LLMs has become a key focus for many LLM builders, and once people realize the value of this, it will only be a matter of time." For now, Manus is doubling down on audiences it can already serve. In a written response, the company said its “primary focus is overseas expansion,” noting that new offices in San Francisco, Singapore, and Tokyo have opened in the past month. A super‑app approach Although the concept of AI agents is still relatively new, the consumer-facing AI app market in China is already crowded with major tech players. DeepSeek remains the most widely used, while ByteDance’s Doubao and Moonshot’s Kimi have also become household names. However, most of these apps are still optimized for chat and entertainment rather than task execution. This gap in the local market has pushed China’s big tech firms to roll out their own user-facing agents, though early versions remain uneven in quality and rough around the edges.  ByteDance is testing Coze Space, an AI agent based on its own Doubao model family that lets users toggle between “plan” and “execute” modes, so they can either directly guide the agent’s actions or step back and watch it work autonomously. It connects up to 14 popular apps, including GitHub, Notion, and the company’s own Lark office suite. Early reviews say the tool can feel clunky and has a high failure rate, but it clearly aims to match what Manus offers. Meanwhile, Zhipu AI has released a free agent called AutoGLM Rumination, built on its proprietary ChatGLM models. Shanghai‑based Minimax has launched Minimax Agent. Both products look almost identical to Manus and demo basic tasks such as building a simple website, planning a trip, making a small Flash game, or running quick data analysis. Despite the limited usability of most general AI agents launched within China, big companies have plans to change that. During a May 15 earnings call, Tencent president Liu Zhiping teased an agent that would weave automation directly into China’s most ubiquitous app, WeChat.  Considered the original super-app, WeChat already handles messaging, mobile payments, news, and millions of mini‑programs that act like embedded apps. These programs give Tencent, its developer, access to data from millions of services that pervade everyday life in China, an advantage most competitors can only envy. Historically, China’s consumer internet has splintered into competing walled gardens—share a Taobao link in WeChat and it resolves as plaintext, not a preview card. Unlike the more interoperable Western internet, China’s tech giants have long resisted integration with one another, choosing to wage platform war at the expense of a seamless user experience. But the use of mini‑programs has given WeChat unprecedented reach across services that once resisted interoperability, from gym bookings to grocery orders. An agent able to roam that ecosystem could bypass the integration headaches dogging independent startups. Alibaba, the e-commerce giant behind the Qwen model series, has been a front-runner in China's AI race but has been slower to release consumer-facing products. Even though Qwen was the most downloaded open-source model on Hugging Face in 2024, it didn’t power a dedicated chatbot app until early 2025. In March, Alibaba rebranded its cloud storage and search app Quark into an all-in-one AI search tool. By June, Quark had introduced DeepResearch—a new mode that marks its most agent-like effort to date.  ByteDance and Alibaba did not reply to MIT Technology Review’s request for comments. “Historically, Chinese tech products tend to pursue the all-in-one, super-app approach, and the latest Chinese AI agents reflect just that,” says Li of Simular, who previously worked at Google DeepMind on AI-enabled work automation. “In contrast, AI agents in the US are more focused on serving specific verticals.” Pei, the researcher at Stanford, says that existing tech giants could have a huge advantage in bringing the vision of general AI agents to life—especially those with built-in integration across services. “The customer-facing AI agent market is still very early, with tons of problems like authentication and liability,” he says. “But companies that already operate across a wide range of services have a natural advantage in deploying agents at scale.”

ZDNET's key takeaways The Eufy E28 robot vacuum, mop, and spot cleaner combination is available for $999.The mop performs better than more expensive flagships, and the water tank system doubles as a portable spot cleaner with a self-cleaning hose.Unplugging the spot cleaner also unplugs the charging station, and the spot cleaner requires you to brush it to scrub. more buying choices The Eufy E28 Omni robot vacuum and mop just hit its lowest price ever at $850, a total of $450 off with an on-page Amazon coupon.Robot vacuum manufacturers are constantly trying to outdo one another. One company develops a detachable cordless vacuum for the robot's dock, and the next makes the robot's dustbin a detachable cordless vacuum. Now, Eufy is upping the ante with the first robot vacuum to also have a detachable spot cleaner for carpet and upholstery. Also: Finally, I found a robot and handheld vacuum combo that's ideal for apartment dwellersThe Eufy Omni E28 is a robot vacuum and mop with a portable spot cleaner for carpets and upholstery. You can run your robot vacuum as you would any other, but you can also pick up the top with a built-in retractable handle and do some deep cleaning on carpets, rugs, or upholstery. I've been testing this robot vacuum over the past couple of weeks, and I'm happy to report on its performance.  details View at Amazon Essentially, the deep cleaner you carry is the Omni station's clean and dirty water tank, so it's where the robot autonomously sources clean water for its mop when it's at the dock. The deep cleaner has a handle for spot-cleaning with a hose, so you can set it down near where you want to clean and plug it in without holding it as you clean. Aside from detachable cordless vacuums, a mop roller system is another big thing for robot vacuums; many makers, including SwitchBot, Yeedi, and Ecovacs, are moving away from detachable rotating mop pads and using mop rollers instead. Eufy's also done this before with the flagship S1 Pro.  Maria Diaz/ZDNETThe Omni E28 also has a mop roller, which looks similar to the S1 Pro but not identical. The Eufy Omni S1 Pro is unequivocally the best mopping robot vacuum I've ever tested, so I was excited to test the new Omni E28, which features a very similar system. Thankfully, the Omni E28 didn't disappoint. The large mop roller covers the length of the vacuum's width and leaves no streaks behind during mopping. Streaky floors are one of my biggest pet peeves when testing robot vacuum and mop combinations, and they're a more common occurrence than I'd like. Also: My favorite indoor security camera has no subscription fees and is on sale right nowEufy is also launching an E25 robot vacuum, which features the same HydroJet self-cleaning mop system as the E28. The robot's mop roller is continuously scraped clean as it spins, and it keeps clean and dirty water in separate tanks within its body, so it only mops with clean water. The mop also exerts more downward pressure to ensure deep cleaning than the S1 Pro, with 1.5kg (3.3 lbs) instead of 1kg.  Maria Diaz/ZDNETHowever, continuously spraying the roller with clean water inside the robot keeps it moist. Scraping off the dirt and wringing out the dirty water as the roller spins ensures your floors are clean instead of streaky or filmy. The Eufy Omni E28, like the S1 Pro, is the closest thing to a manual mopping result you can get from a robot vacuum and mop. Also: I love a vacuum and mop to clean dry and wet messes, especially when it's on saleThe E28 and E25 robot vacuums have 20,000Pa of suction power and anti-tangle technology, including a DuoSpiral double roller brush to remove pet hair and avoid entanglements. The robots are self-emptying robot vacuums with a self-washing mop roller. The only difference is the portable spot cleaner, which only the E28 has. The Eufy Omni E28 robot vacuum and mop has dual brush rollers and a roller mop. Maria Diaz/ZDNETThe spot cleaner proved to be very effective. It's heavy, like most spot cleaners when full of water, but it has a retractable handle on top that makes it easy to carry around. You also don't have to press any buttons or move anything to release it; just unplug it, pull the retractable handle, and go. I didn't like that the spot cleaner powers the base station, so if you take it somewhere else in your home to clean, the robot's charging station is left without power until you return it. This is fine for quick cleanups, but it can be annoying when a forgetful house member takes the spot cleaner upstairs and leaves it in a room for two days after they're done (totally not me). Also: This simple Amazon tablet became one of my biggest smart home upgrades yet (and it's on sale)The spot cleaner has a self-cleaning hose attached, but no other attachments. The hose ends in a static brush head that sprays clean water, which you use to clean messes on soft surfaces, like carpets, rugs, and upholstery. Since it isn't motorized, you must manually brush the rug or fabric with the brush head to scrub it clean. Once you're done, you just have to press a button along one end of the brush head to make the spot cleaner cycle clean water through the hose without spilling, cleaning the hose for you.  These pictures were taken three minutes apart. The left (before) shows yogurt stains on an ottoman. I sprayed a liquid spot cleaner and went over it with the brush and water; the results are on the right (after). Maria Diaz/ZDNETThere's no separate detergent tank for the spot cleaner, though you could add some detergent directly into the clean water tank. I recommend using just water in the clean water tank and spraying your preferred carpet or upholstery cleaning solution directly on the stains to pre-treat them. You can then scrub the mess clean with the Eufy E28 brush and rinse out the detergent. ZDNET's buying adviceThe Eufy Omni E28 is designed to solve a big problem for many US customers: the need for regular floor cleanings and the ability to quickly clean up messes on soft surfaces in a single device. This device isn't meant to replace your existing carpet cleaner, but it's perfect for consumers who may need a spot cleaner and are also in the market for one of the best robot vacuum and mops available.Also: Spring cleaning takes me no time with my favorite smart home devicesI'm not a fan of fully carpeted living spaces, but my home has carpeted bedrooms, one of which we use as our TV room. With three young kids who have movie nights and play time in that TV room, its carpet unfortunately sees a lot of spills. I was in the market for a spot cleaner for a while, but decided to buy a full carpet cleaner instead, which I use to clean our carpets at least once every quarter.  The E28 cleaned up the soy sauce stains comparably to the larger carpet cleaner. Maria Diaz/ZDNETBut there are always little messes in between -- whether it's spilled ketchup on the carpet, a muddy shoeprint on the entryway rug, or yogurt on the fabric ottoman. So I appreciate having the Eufy Omni E28's spot cleaner always handy. Instead of dragging out a heavy carpet cleaning machine and filling it with water to clean a 6-inch in diameter spill, I can just unplug and grab the E28, clean the mess, and return it to the dock with little work on my end. It's also priced quite well for a first-of-its-kind device with a flagship-level robot vacuum and mop. The Eufy Omni E28 robot vacuum, mop, and spot cleaning combination is now available on Eufy's website and Amazon for $1,000. The Eufy E25 robot vacuum and mop without the spot cleaner will be available in June for $900. When will this deal expire? While many sales events feature deals for a specific length of time, deals are on a limited-time basis, making them subject to expire at any time. ZDNET remains committed to finding, sharing, and updating the best offers to help you maximize your savings so you can feel as confident in your purchases as we feel in our recommendations. Our ZDNET team of experts constantly monitors the deals we feature to keep our stories up-to-date. If you missed out on this deal, don't worry -- we're always sourcing new savings opportunities at ZDNET.com. Show more Featured

Last year, China saw a boom in foundation models, the do-everything large language models that underpin the AI revolution. This year, the focus has shifted to AI agents—systems that are less about responding to users’ queries and more about autonomously accomplishing things for them.  There are now a host of Chinese startups building these general-purpose digital tools, which can answer emails, browse the internet to plan vacations, and even design an interactive website. Many of these have emerged in just the last two months, following in the footsteps of Manus—a general AI agent that sparked weeks of social media frenzy for invite codes after its limited-release launch in early March.  These emerging AI agents aren’t large language models themselves. Instead, they’re built on top of them, using a workflow-based structure designed to get things done. A lot of these systems also introduce a different way of interacting with AI. Rather than just chatting back and forth with users, they are optimized for managing and executing multistep tasks—booking flights, managing schedules, conducting research—by using external tools and remembering instructions.  China could take the lead on building these kinds of agents. The country’s tightly integrated app ecosystems, rapid product cycles, and digitally fluent user base could provide a favorable environment for embedding AI into daily life.  For now, its leading AI agent startups are focusing their attention on the global market, because the best Western models don’t operate inside China’s firewalls. But that could change soon: Tech giants like ByteDance and Tencent are preparing their own AI agents that could bake automation directly into their native super-apps, pulling data from their vast ecosystem of programs that dominate many aspects of daily life in the country.  As the race to define what a useful AI agent looks like unfolds, a mix of ambitious startups and entrenched tech giants are now testing how these tools might actually work in practice—and for whom. Set the standard It’s been a whirlwind few months for Manus, which was developed by the Wuhan-based startup Butterfly Effect. The company raised $75 million in a funding round led by the US venture capital firm Benchmark, took the product on an ambitious global roadshow, and hired dozens of new employees.  Even before registration opened to the public in May, Manus had become a reference point for what a broad, consumer‑oriented AI agent should accomplish. Rather than handling narrow chores for businesses, this “general” agent is designed to be able to help with everyday tasks like trip planning, stock comparison, or your kid’s school project.  Unlike previous AI agents, Manus uses a browser-based sandbox that lets users supervise the agent like an intern, watching in real time as it scrolls through web pages, reads articles, or codes actions. It also proactively asks clarifying questions, supports long-term memory that would serve as context for future tasks. "Manus represents a promising product experience for AI agents," says Ang Li, cofounder and CEO of Simular, a startup based in Palo Alto, California, that’s building computer use agents, AI agents that control a virtual computer. “I believe Chinese startups have a huge advantage when it comes to designing consumer products, thanks to cutthroat domestic competition that leads to fast execution and greater attention to product details.” In the case of Manus, the competition is moving fast. Two of the most buzzy follow‑ups, Genspark and Flowith, for example, are already boasting benchmark scores that match or edge past Manus’s.  Genspark, led by former Baidu executives Eric Jing and Kay Zhu, links many small “super agents” through what it calls multi‑component prompting. The agent can switch among several large language models, accepts both images and text, and carries out tasks from making slide decks to placing phone calls. Whereas Manus relies heavily on Browser Use, a popular open-source product that lets agents operate a web browser in a virtual window like a human, Genspark directly integrates with a wide array of tools and APIs. Launched in April, the company says that it already has over 5 million users and over $36 million in yearly revenue. Flowith, the work of a young team that first grabbed public attention in April 2025 at a developer event hosted by the popular social media app Xiaohongshu, takes a different tack. Marketed as an “infinite agent,” it opens on a blank canvas where each question becomes a node on a branching map. Users can backtrack, take new branches, and store results in personal or sharable “knowledge gardens”—a design that feels more like project management software (think Notion) than a typical chat interface. Every inquiry or task builds its own mind-map-like graph, encouraging a more nonlinear and creative interaction with AI. Flowith’s core agent, NEO, runs in the cloud and can perform scheduled tasks like sending emails and compiling files. The founders want the app to be a “knowledge marketbase”, and aims to tap into the social aspect of AI with the aspiration of becoming “the OnlyFans of AI knowledge creators”. What they also share with Manus is the global ambition. Both Genspark and Flowith have stated that their primary focus is the international market. A global address Startups like Manus, Genspark, and Flowith—though founded by Chinese entrepreneurs—could blend seamlessly into the global tech scene and compete effectively abroad. Founders, investors, and analysts that MIT Technology Review has spoken to believe Chinese companies are moving fast, executing well, and quickly coming up with new products.  Money reinforces the pull to launch overseas. Customers there pay more, and there are plenty to go around. “You can price in USD, and with the exchange rate that’s a sevenfold multiplier,” Manus cofounder Xiao Hong quipped on a podcast. “Even if we’re only operating at 10% power because of cultural differences overseas, we’ll still make more than in China.” But creating the same functionality in China is a challenge. Major US AI companies including OpenAI and Anthropic have opted out of mainland China because of geopolitical risks and challenges with regulatory compliance. Their absence initially created a black market as users resorted to VPNs and third-party mirrors to access tools like ChatGPT and Claude. That vacuum has since been filled by a new wave of Chinese chatbots—DeepSeek, Doubao, Kimi—but the appetite for foreign models hasn’t gone away.  Manus, for example, uses Anthropic’s Claude Sonnet—widely considered the top model for agentic tasks. Manus cofounder Zhang Tao has repeatedly praised Claude’s ability to juggle tools, remember contexts, and hold multi‑round conversations—all crucial for turning chatty software into an effective executive assistant. But the company’s use of Sonnet has made its agent functionally unusable inside China without a VPN. If you open Manus from a mainland IP address, you’ll see a notice explaining that the team is “working on integrating Qwen’s model,” a special local version that is built on top of Alibaba’s open-source model.  An engineer overseeing ByteDance’s work on developing an agent, who spoke to MIT Technology Review anonymously to avoid sanction, said that the absence of Claude Sonnet models “limits everything we do in China.” DeepSeek’s open models, he added, still hallucinate too often and lack training on real‑world workflows. Developers we spoke with rank Alibaba’s Qwen series as the best domestic alternative, yet most say that switching to Qwen knocks performance down a notch. Jiaxin Pei, a postdoctoral researcher at Stanford’s Institute for Human‑Centered AI, thinks that gap will close: “Building agentic capabilities in base LLMs has become a key focus for many LLM builders, and once people realize the value of this, it will only be a matter of time." For now, Manus is doubling down on audiences it can already serve. In a written response, the company said its “primary focus is overseas expansion,” noting that new offices in San Francisco, Singapore, and Tokyo have opened in the past month. A super‑app approach Although the concept of AI agents is still relatively new, the consumer-facing AI app market in China is already crowded with major tech players. DeepSeek remains the most widely used, while ByteDance’s Doubao and Moonshot’s Kimi have also become household names. However, most of these apps are still optimized for chat and entertainment rather than task execution. This gap in the local market has pushed China’s big tech firms to roll out their own user-facing agents, though early versions remain uneven in quality and rough around the edges.  ByteDance is testing Coze Space, an AI agent based on its own Doubao model family that lets users toggle between “plan” and “execute” modes, so they can either directly guide the agent’s actions or step back and watch it work autonomously. It connects up to 14 popular apps, including GitHub, Notion, and the company’s own Lark office suite. Early reviews say the tool can feel clunky and has a high failure rate, but it clearly aims to match what Manus offers. Meanwhile, Zhipu AI has released a free agent called AutoGLM Rumination, built on its proprietary ChatGLM models. Shanghai‑based Minimax has launched Minimax Agent. Both products look almost identical to Manus and demo basic tasks such as building a simple website, planning a trip, making a small Flash game, or running quick data analysis. Despite the limited usability of most general AI agents launched within China, big companies have plans to change that. During a May 15 earnings call, Tencent president Liu Zhiping teased an agent that would weave automation directly into China’s most ubiquitous app, WeChat.  Considered the original super-app, WeChat already handles messaging, mobile payments, news, and millions of mini‑programs that act like embedded apps. These programs give Tencent, its developer, access to data from millions of services that pervade everyday life in China, an advantage most competitors can only envy. Historically, China’s consumer internet has splintered into competing walled gardens—share a Taobao link in WeChat and it resolves as plaintext, not a preview card. Unlike the more interoperable Western internet, China’s tech giants have long resisted integration with one another, choosing to wage platform war at the expense of a seamless user experience. But the use of mini‑programs has given WeChat unprecedented reach across services that once resisted interoperability, from gym bookings to grocery orders. An agent able to roam that ecosystem could bypass the integration headaches dogging independent startups. Alibaba, the e-commerce giant behind the Qwen model series, has been a front-runner in China's AI race but has been slower to release consumer-facing products. Even though Qwen was the most downloaded open-source model on Hugging Face in 2024, it didn’t power a dedicated chatbot app until early 2025. In March, Alibaba rebranded its cloud storage and search app Quark into an all-in-one AI search tool. By June, Quark had introduced DeepResearch—a new mode that marks its most agent-like effort to date.  ByteDance and Alibaba did not reply to MIT Technology Review’s request for comments. “Historically, Chinese tech products tend to pursue the all-in-one, super-app approach, and the latest Chinese AI agents reflect just that,” says Li of Simular, who previously worked at Google DeepMind on AI-enabled work automation. “In contrast, AI agents in the US are more focused on serving specific verticals.” Pei, the researcher at Stanford, says that existing tech giants could have a huge advantage in bringing the vision of general AI agents to life—especially those with built-in integration across services. “The customer-facing AI agent market is still very early, with tons of problems like authentication and liability,” he says. “But companies that already operate across a wide range of services have a natural advantage in deploying agents at scale.”

Last year, China saw a boom in foundation models, the do-everything large language models that underpin the AI revolution. This year, the focus has shifted to AI agents—systems that are less about responding to users’ queries and more about autonomously accomplishing things for them.  There are now a host of Chinese startups building these general-purpose digital tools, which can answer emails, browse the internet to plan vacations, and even design an interactive website. Many of these have emerged in just the last two months, following in the footsteps of Manus—a general AI agent that sparked weeks of social media frenzy for invite codes after its limited-release launch in early March.  These emerging AI agents aren’t large language models themselves. Instead, they’re built on top of them, using a workflow-based structure designed to get things done. A lot of these systems also introduce a different way of interacting with AI. Rather than just chatting back and forth with users, they are optimized for managing and executing multistep tasks—booking flights, managing schedules, conducting research—by using external tools and remembering instructions.  China could take the lead on building these kinds of agents. The country’s tightly integrated app ecosystems, rapid product cycles, and digitally fluent user base could provide a favorable environment for embedding AI into daily life.  For now, its leading AI agent startups are focusing their attention on the global market, because the best Western models don’t operate inside China’s firewalls. But that could change soon: Tech giants like ByteDance and Tencent are preparing their own AI agents that could bake automation directly into their native super-apps, pulling data from their vast ecosystem of programs that dominate many aspects of daily life in the country.  As the race to define what a useful AI agent looks like unfolds, a mix of ambitious startups and entrenched tech giants are now testing how these tools might actually work in practice—and for whom. Set the standard It’s been a whirlwind few months for Manus, which was developed by the Wuhan-based startup Butterfly Effect. The company raised $75 million in a funding round led by the US venture capital firm Benchmark, took the product on an ambitious global roadshow, and hired dozens of new employees.  Even before registration opened to the public in May, Manus had become a reference point for what a broad, consumer‑oriented AI agent should accomplish. Rather than handling narrow chores for businesses, this “general” agent is designed to be able to help with everyday tasks like trip planning, stock comparison, or your kid’s school project.  Unlike previous AI agents, Manus uses a browser-based sandbox that lets users supervise the agent like an intern, watching in real time as it scrolls through web pages, reads articles, or codes actions. It also proactively asks clarifying questions, supports long-term memory that would serve as context for future tasks. “Manus represents a promising product experience for AI agents,” says Ang Li, cofounder and CEO of Simular, a startup based in Palo Alto, California, that’s building computer use agents, AI agents that control a virtual computer. “I believe Chinese startups have a huge advantage when it comes to designing consumer products, thanks to cutthroat domestic competition that leads to fast execution and greater attention to product details.” In the case of Manus, the competition is moving fast. Two of the most buzzy follow‑ups, Genspark and Flowith, for example, are already boasting benchmark scores that match or edge past Manus’s.  Genspark, led by former Baidu executives Eric Jing and Kay Zhu, links many small “super agents” through what it calls multi‑component prompting. The agent can switch among several large language models, accepts both images and text, and carries out tasks from making slide decks to placing phone calls. Whereas Manus relies heavily on Browser Use, a popular open-source product that lets agents operate a web browser in a virtual window like a human, Genspark directly integrates with a wide array of tools and APIs. Launched in April, the company says that it already has over 5 million users and over $36 million in yearly revenue. Flowith, the work of a young team that first grabbed public attention in April 2025 at a developer event hosted by the popular social media app Xiaohongshu, takes a different tack. Marketed as an “infinite agent,” it opens on a blank canvas where each question becomes a node on a branching map. Users can backtrack, take new branches, and store results in personal or sharable “knowledge gardens”—a design that feels more like project management software (think Notion) than a typical chat interface. Every inquiry or task builds its own mind-map-like graph, encouraging a more nonlinear and creative interaction with AI. Flowith’s core agent, NEO, runs in the cloud and can perform scheduled tasks like sending emails and compiling files. The founders want the app to be a “knowledge marketbase”, and aims to tap into the social aspect of AI with the aspiration of becoming “the OnlyFans of AI knowledge creators”. What they also share with Manus is the global ambition. Both Genspark and Flowith have stated that their primary focus is the international market. A global address Startups like Manus, Genspark, and Flowith—though founded by Chinese entrepreneurs—could blend seamlessly into the global tech scene and compete effectively abroad. Founders, investors, and analysts that MIT Technology Review has spoken to believe Chinese companies are moving fast, executing well, and quickly coming up with new products.  Money reinforces the pull to launch overseas. Customers there pay more, and there are plenty to go around. “You can price in USD, and with the exchange rate that’s a sevenfold multiplier,” Manus cofounder Xiao Hong quipped on a podcast. “Even if we’re only operating at 10% power because of cultural differences overseas, we’ll still make more than in China.” But creating the same functionality in China is a challenge. Major US AI companies including OpenAI and Anthropic have opted out of mainland China because of geopolitical risks and challenges with regulatory compliance. Their absence initially created a black market as users resorted to VPNs and third-party mirrors to access tools like ChatGPT and Claude. That vacuum has since been filled by a new wave of Chinese chatbots—DeepSeek, Doubao, Kimi—but the appetite for foreign models hasn’t gone away.  Manus, for example, uses Anthropic’s Claude Sonnet—widely considered the top model for agentic tasks. Manus cofounder Zhang Tao has repeatedly praised Claude’s ability to juggle tools, remember contexts, and hold multi‑round conversations—all crucial for turning chatty software into an effective executive assistant. But the company’s use of Sonnet has made its agent functionally unusable inside China without a VPN. If you open Manus from a mainland IP address, you’ll see a notice explaining that the team is “working on integrating Qwen’s model,” a special local version that is built on top of Alibaba’s open-source model.  An engineer overseeing ByteDance’s work on developing an agent, who spoke to MIT Technology Review anonymously to avoid sanction, said that the absence of Claude Sonnet models “limits everything we do in China.” DeepSeek’s open models, he added, still hallucinate too often and lack training on real‑world workflows. Developers we spoke with rank Alibaba’s Qwen series as the best domestic alternative, yet most say that switching to Qwen knocks performance down a notch. Jiaxin Pei, a postdoctoral researcher at Stanford’s Institute for Human‑Centered AI, thinks that gap will close: “Building agentic capabilities in base LLMs has become a key focus for many LLM builders, and once people realize the value of this, it will only be a matter of time.” For now, Manus is doubling down on audiences it can already serve. In a written response, the company said its “primary focus is overseas expansion,” noting that new offices in San Francisco, Singapore, and Tokyo have opened in the past month. A super‑app approach Although the concept of AI agents is still relatively new, the consumer-facing AI app market in China is already crowded with major tech players. DeepSeek remains the most widely used, while ByteDance’s Doubao and Moonshot’s Kimi have also become household names. However, most of these apps are still optimized for chat and entertainment rather than task execution. This gap in the local market has pushed China’s big tech firms to roll out their own user-facing agents, though early versions remain uneven in quality and rough around the edges.  ByteDance is testing Coze Space, an AI agent based on its own Doubao model family that lets users toggle between “plan” and “execute” modes, so they can either directly guide the agent’s actions or step back and watch it work autonomously. It connects up to 14 popular apps, including GitHub, Notion, and the company’s own Lark office suite. Early reviews say the tool can feel clunky and has a high failure rate, but it clearly aims to match what Manus offers. Meanwhile, Zhipu AI has released a free agent called AutoGLM Rumination, built on its proprietary ChatGLM models. Shanghai‑based Minimax has launched Minimax Agent. Both products look almost identical to Manus and demo basic tasks such as building a simple website, planning a trip, making a small Flash game, or running quick data analysis. Despite the limited usability of most general AI agents launched within China, big companies have plans to change that. During a May 15 earnings call, Tencent president Liu Zhiping teased an agent that would weave automation directly into China’s most ubiquitous app, WeChat.  Considered the original super-app, WeChat already handles messaging, mobile payments, news, and millions of mini‑programs that act like embedded apps. These programs give Tencent, its developer, access to data from millions of services that pervade everyday life in China, an advantage most competitors can only envy. Historically, China’s consumer internet has splintered into competing walled gardens—share a Taobao link in WeChat and it resolves as plaintext, not a preview card. Unlike the more interoperable Western internet, China’s tech giants have long resisted integration with one another, choosing to wage platform war at the expense of a seamless user experience. But the use of mini‑programs has given WeChat unprecedented reach across services that once resisted interoperability, from gym bookings to grocery orders. An agent able to roam that ecosystem could bypass the integration headaches dogging independent startups. Alibaba, the e-commerce giant behind the Qwen model series, has been a front-runner in China’s AI race but has been slower to release consumer-facing products. Even though Qwen was the most downloaded open-source model on Hugging Face in 2024, it didn’t power a dedicated chatbot app until early 2025. In March, Alibaba rebranded its cloud storage and search app Quark into an all-in-one AI search tool. By June, Quark had introduced DeepResearch—a new mode that marks its most agent-like effort to date.  ByteDance and Alibaba did not reply to MIT Technology Review’s request for comments. “Historically, Chinese tech products tend to pursue the all-in-one, super-app approach, and the latest Chinese AI agents reflect just that,” says Li of Simular, who previously worked at Google DeepMind on AI-enabled work automation. “In contrast, AI agents in the US are more focused on serving specific verticals.” Pei, the researcher at Stanford, says that existing tech giants could have a huge advantage in bringing the vision of general AI agents to life—especially those with built-in integration across services. “The customer-facing AI agent market is still very early, with tons of problems like authentication and liability,” he says. “But companies that already operate across a wide range of services have a natural advantage in deploying agents at scale.”

Macworld If you’re an Apple user we already know you like the finer things in life, and we’d put money on the fact your iPhone is controlling half the gadgets in your home. Why? Well, why not. It just works, and this seamless integration with smart home kit frees up your time and removes many of the frustrations of everyday life. But there is one smart gadget you might not have considered yet, and this robot can take care of the most gruelling chore of all: cleaning out the pool. Investing in a robot pool cleaner is a smart choice. Imagine being able to enjoy the luxury of a swimming pool, but with none of the hard work pool maintenance typically requires. The Dreame Z1 Pro is a great pick, unlocking plenty of extra free time to actually enjoy your pool. Imagine this: you’ve spent a very loooong week at the office, but now it’s Saturday, the weather has warmed up, and you can’t think of anything better than spending the weekend lounging in the pool you worked so hard to pay for. You grab a towel, get into your swimsuit, oil up… and then start picking out leaves that have fallen into the water. Oh, and here are some bugs, let’s get rid of those. And we’ll tackle that nice slimey film on top of the water. Did you scrub the pool walls and floors lately? Suddenly this luxurious dip you had been longing for feels like hassle. Let Dreame take the hard work out of pool maintenance It doesn’t have to be this way. Dreame can help, and the Z1 Pro ensures you enjoy hands-off pool maintenance anytime you need it. It deploys all of its 8,000GPH suction power to clean your pool top to bottom, picking up anything that’s fallen in the water and scrubbing the waterline. Much like a robot vacuum will tackle cleaning your floors, this one will stick to the bottom or sides of the pool and meticulously clean the surface, too. The robot’s sensors will quickly map out the pool while AI smarts divide it into areas so it can better tackle cleaning. The Z1 Pro even knows how to avoid obstacles such as drain covers, lights, and so on. This wireless robot pool cleaner can run for about 180 minutes on a single charge, cleaning up to 2,160 square feet in the process. That’s quite a lot of pool to cover! While it can normally do its own thing and clean wherever it deems necessary, you can also assume full control right from your iPhone or the bundled remote. Hands-free pool cleaning If you’re running an emergency clean just before guests arrive and you don’t want them to see the robot, use the remote to call it back to you. This LiFi-connected remote is suitable for fresh- and saltwater pools and can communicate with the robot even while it’s underwater, allowing you to assign tasks or steer the cleaner. Once the cleaning job is done, the robot will park itself at the edge of the pool so you can pick it up and set it up to recharge. Easy! To set up the robot and get cleaning reports, also install the Dreamehome app on your iPhone. You’ll see everything from the map to the multiple cleaning modes available, as well as the cleaning logs. What are you waiting for? Dreame’s Z1 Pro costs $1,499 at Dreame’s online shop, but right now it’s discounted to $1,099–and you can save a further 15% with the code PROMO15. We think you might just find the time saved on pool maintenance is worth more to you than the discount, however. Take a further 15% off Dreame Z1 Pro Use Promo Code PROMO15

When people talk about UX, it’s usually about the things they can see and interact with, like wireframes and prototypes, smart interactions, and design tools like Figma, Miro, or Maze. Some of the outputs are even glamorized, like design systems, research reports, and pixel-perfect UI designs. But here’s the truth I’ve seen again and again in over two decades of working in UX: none of that moves the needle if there is no collaboration. Great UX doesn’t happen in isolation. It happens through conversations with engineers, product managers, customer-facing teams, and the customer support teams who manage support tickets. Amazing UX ideas come alive in messy Miro sessions, cross-functional workshops, and those online chats (e.g., Slack or Teams) where people align, adapt, and co-create. Some of the most impactful moments in my career weren’t when I was “designing” in the traditional sense. They have been gaining incredible insights when discussing problems with teammates who have varied experiences, brainstorming, and coming up with ideas that I never could have come up with on my own. As I always say, ten minds in a room will come up with ten times as many ideas as one mind. Often, many ideas are the most useful outcome. There have been times when a team has helped to reframe a problem in a workshop, taken vague and conflicting feedback, and clarified a path forward, or I’ve sat with a sales rep and heard the same user complaint show up in multiple conversations. This is when design becomes a team sport, and when your ability to capture the outcomes multiplies the UX impact. Why This Article Matters Now The reason collaboration feels so urgent now is that the way we work since COVID has changed, according to a study published by the US Department of Labor. Teams are more cross-functional, often remote, and increasingly complex. Silos are easier to fall into, due to distance or lack of face-to-face contact, and yet alignment has never been more important. We can’t afford to see collaboration as a “nice to have” anymore. It’s a core skill, especially in UX, where our work touches so many parts of an organisation. Let’s break down what collaboration in UX really means, and why it deserves way more attention than it gets. What Is Collaboration In UX, Really? Let’s start by clearing up a misconception. Collaboration is not the same as cooperation. Cooperation: “You do your thing, I’ll do mine, and we’ll check in later.” Collaboration: “Let’s figure this out together and co-own the outcome.” Collaboration, as defined in the book Communication Concepts, published by Deakin University, involves working with others to produce outputs and/or achieve shared goals. The outcome of collaboration is typically a tangible product or a measurable achievement, such as solving a problem or making a decision. Here’s an example from a recent project: Recently, I worked on a fraud alert platform for a fintech business. It was a six-month project, and we had zero access to users, as the product had not yet hit the market. Also, the users were highly specialised in the B2B finance space and were difficult to find. Additionally, the team members I needed to collaborate with were based in Malaysia and Melbourne, while I am located in Sydney. Instead of treating that as a dead end, we turned inward: collaborating with subject matter experts, professional services consultants, compliance specialists, and customer support team members who had deep knowledge of fraud patterns and customer pain points. Through bi-weekly workshops using a Miro board, iterative feedback loops, and sketching sessions, we worked on design solution options. I even asked them to present their own design version as part of the process. After months of iterating on the fraud investigation platform through these collaboration sessions, I ended up with two different design frameworks for the investigator’s dashboard. Instead of just presenting the “best one” and hoping for buy-in, I ran a voting exercise with PMs, engineers, SMEs, and customer support. Everyone had a voice. The winning design was created and validated with the input of the team, resulting in an outcome that solved many problems for the end user and was owned by the entire team. That’s collaboration! It is definitely one of the most satisfying projects of my career. On the other hand, I recently caught up with an old colleague who now serves as a product owner. Her story was a cautionary tale: the design team had gone ahead with a major redesign of an app without looping her in until late in the game. Not surprisingly, the new design missed several key product constraints and business goals. It had to be scrapped and redone, with her now at the table. That experience reinforced what we all know deep down: your best work rarely happens in isolation. As illustrated in my experience, true collaboration can span many roles. It’s not just between designers and PMs. It can also include QA testers who identify real-world issues, content strategists who ensure our language is clear and inclusive, sales representatives who interact with customers on a daily basis, marketers who understand the brand’s voice, and, of course, customer support agents who are often the first to hear when something goes wrong. The best outcomes arrive when we’re open to different perspectives and inputs. Why Collaboration Is So Overlooked? If collaboration is so powerful, why don’t we talk about it more? In my experience, one reason is the myth of the “lone UX hero”. Many of us entered the field inspired by stories of design geniuses revolutionising products on their own. Our portfolios often reflect that as well. We showcase our solo work, our processes, and our wins. Job descriptions often reinforce the idea of the solo UX designer, listing tool proficiency and deliverables more than soft skills and team dynamics. And then there’s the team culture within many organisations of “just get the work done”, which often leads to fewer meetings and tighter deadlines. As a result, a sense of collaboration is inefficient and wasted. I have also experienced working with some designers where perfectionism and territoriality creep in — “This is my design” — which kills the open, communal spirit that collaboration needs. When Collaboration Is The User Research In an ideal world, we’d always have direct access to users. But let’s be real. Sometimes that just doesn’t happen. Whether it’s due to budget constraints, time limitations, or layers of bureaucracy, talking to end users isn’t always possible. That’s where collaboration with team members becomes even more crucial. The next best thing to talking to users? Talking to the people who talk to users. Sales teams, customer success reps, tech support, and field engineers. They’re all user researchers in disguise! On another B2C project, the end users were having trouble completing the key task. My role was to redesign the onboarding experience for an online identity capture tool for end users. I was unable to schedule interviews with end users due to budget and time constraints, so I turned to the sales and tech support teams. I conducted multiple mini-workshops to identify the most common onboarding issues they had heard directly from our customers. This led to a huge “aha” moment: most users dropped off before the document capture process. They may have been struggling with a lack of instruction, not knowing the required time, or not understanding the steps involved in completing the onboarding process. That insight reframed my approach, and we ultimately redesigned the flow to prioritize orientation and clear instructions before proceeding to the setup steps. Below is an example of one of the screen designs, including some of the instructions we added. This kind of collaboration is user research. It’s not a substitute for talking to users directly, but it’s a powerful proxy when you have limited options. But What About Using AI? Glad you asked! Even AI tools, which are increasingly being used for idea generation, pattern recognition, or rapid prototyping, don’t replace collaboration; they just change the shape of it. AI can help you explore design patterns, draft user flows, or generate multiple variations of a layout in seconds. It’s fantastic for getting past creative blocks or pressure-testing your assumptions. But let’s be clear: these tools are accelerators, not oracles. As an innovation and strategy consultant Nathan Waterhouse points out, AI can point you in a direction, but it can’t tell you which direction is the right one in your specific context. That still requires human judgment, empathy, and an understanding of the messy realities of users and business goals. You still need people, especially those closest to your users, to validate, challenge, and evolve any AI-generated idea. For instance, you might use ChatGPT to brainstorm onboarding flows for a SaaS tool, but if you’re not involving customer support reps who regularly hear “I didn’t know where to start” or “I couldn’t even log in,” you’re just working with assumptions. The same applies to engineers who know what is technically feasible or PMs who understand where the business is headed. AI can generate ideas, but only collaboration turns those ideas into something usable, valuable, and real. Think of it as a powerful ingredient, but not the whole recipe. How To Strengthen Your UX Collaboration Skills? If collaboration doesn’t come naturally or hasn’t been a focus, that’s okay. Like any skill, it can be practiced and improved. Here are a few ways to level up: Cultivate curiosity about your teammates.Ask engineers what keeps them up at night. Learn what metrics your PMs care about. Understand the types of tickets the support team handles most frequently. The more you care about their challenges, the more they'll care about yours. Get comfortable facilitating.You don’t need to be a certified Design Sprint master, but learning how to run a structured conversation, align stakeholders, or synthesize different points of view is hugely valuable. Even a simple “What’s working? What’s not?” retro can be an amazing starting point in identifying where you need to focus next. Share early, share often.Don’t wait until your designs are polished to get input. Messy sketches and rough prototypes invite collaboration. When others feel like they’ve helped shape the work, they’re more invested in its success. Practice active listening.When someone critiques your work, don’t immediately defend. Pause. Ask follow-up questions. Reframe the feedback. Collaboration isn’t about consensus; it’s about finding a shared direction that can honour multiple truths. Co-own the outcome.Let go of your ego. The best UX work isn’t “your” work. It’s the result of many voices, skill sets, and conversations converging toward a solution that helps users. It’s not “I”, it’s “we” that will solve this problem together. Conclusion: UX Is A Team Sport Great design doesn’t emerge from a vacuum. It comes from open dialogue, cross-functional understanding, and a shared commitment to solving real problems for real people. If there’s one thing I wish every early-career designer knew, it’s this: Collaboration is not a side skill. It’s the engine behind every meaningful design outcome. And for seasoned professionals, it’s the superpower that turns good teams into great ones. So next time you’re tempted to go heads-down and just “crank out a design,” pause to reflect. Ask who else should be in the room. And invite them in, not just to review your work, but to help create it. Because in the end, the best UX isn’t just what you make. It’s what you make together. Further Reading On SmashingMag “Presenting UX Research And Design To Stakeholders: The Power Of Persuasion,” Victor Yocco “Transforming The Relationship Between Designers And Developers,” Chris Day “Effective Communication For Everyday Meetings,” Andrii Zhdan “Preventing Bad UX Through Integrated Design Workflows,” Ceara Crawshaw

June 3, 20255 min readAre We Ready for Death in Space?NASA has quietly taken steps to prepare for a death in space. We need to ask how nations will deal with this inevitability now, as more people start traveling off the planetBy Peter Cummings edited by Lee Billings SciePro/Science Photo Library/Getty ImagesIn 2012 NASA stealthily slipped a morgue into orbit.No press release. No fanfare. Just a sealed, soft-sided pouch tucked in a cargo shipment to the International Space Station (ISS) alongside freeze-dried meals and scientific gear. Officially, it was called the Human Remains Containment Unit (HRCU). To the untrained eye it looked like a shipping bag for frozen cargo. But to NASA it marked something far more sobering: a major advance in preparing for death beyond Earth.As a kid, I obsessed over how astronauts went to the bathroom in zero gravity. Now, decades later, as a forensic pathologist and a perennial applicant to NASA’s astronaut corps, I find myself fixated on a darker, more haunting question:On supporting science journalismIf you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.What would happen if an astronaut died out there? Would they be brought home, or would they be left behind? If they expired on some other world, would that be their final resting place? If they passed away on a spacecraft or space station, would their remains be cast off into orbit—or sent on an escape-velocity voyage to the interstellar void?NASA, it turns out, has begun working out most of these answers. And none too soon. Because the question itself is no longer if someone will die in space—but when.A Graying CorpsNo astronaut has ever died of natural causes off-world. In 1971 the three-man crew of the Soviet Soyuz 11 mission asphyxiated in space when their spacecraft depressurized shortly before its automated atmospheric reentry—but their deaths were only discovered once the spacecraft landed on Earth. Similarly, every U.S. spaceflight fatality to date has occurred within Earth’s atmosphere—under gravity, oxygen and a clear national jurisdiction. That matters, because it means every spaceflight mortality has played out in familiar territory.But planned missions are getting longer, with destinations beyond low-Earth orbit. And NASA’s astronaut corps is getting older. The average age now hovers around 50—an age bracket where natural death becomes statistically relevant, even for clean-living fitness buffs. Death in space is no longer a thought experiment. It’s a probability curve—and NASA knows it.In response, the agency is making subtle but decisive moves. The most recent astronaut selection cycle was extended—not only to boost intake but also to attract younger crew members capable of handling future long-duration missions.NASA’s Space MorgueIf someone were to die aboard the ISS today, their body would be placed in the HRCU, which would then be sealed and secured in a nonpressurized area to await eventual return to Earth.The HRCU itself is a modified version of a military-grade body bag designed to store human remains in hazardous environments. It integrates with refrigeration systems already aboard the ISS to slow decomposition and includes odor-control filters and moisture-absorbent linings, as well as reversed zippers for respectful access at the head. There are straps to secure the body in a seat for return, and patches for name tags and national flags.Cadaver tests conducted in 2019 at Sam Houston State University have proved the system durable. Some versions held for over 40 days before decomposition breached the barrier. NASA even drop-tested the bag from 19 feet to simulate a hard landing.But it’s never been used in space. And since no one yet knows how a body decomposes in true microgravity (or, for that matter, on the moon), no one can really say whether the HRCU would preserve tissue well enough for a forensic autopsy.This is a troubling knowledge gap, because in space, a death isn’t just a tragic loss—it’s also a vital data point. Was an astronaut’s demise from a fluke of their physiology, or an unavoidable stroke of cosmic bad luck—or was it instead a consequence of flaws in a space habitat’s myriad systems that might be found and fixed? Future lives may depend on understanding what went wrong, via a proper postmortem investigation.But there’s no medical examiner in orbit. So NASA trains its crews in something called the In-Mission Forensic Sample Collection protocol. The space agency’s astronauts may avoid talking about it, but they all have it memorized: Document everything, ideally with real-time guidance from NASA flight surgeons. Photograph the body. Collect blood and vitreous fluid, as well as hair and tissue samples. Only then can the remains be stowed in the HRCU.NASA has also prepared for death outside the station—on spacewalks, the moon or deep space missions. If a crew member perishes in vacuum but their remains are retrieved, the body is wrapped in a specially designed space shroud.The goal isn’t just a technical matter of preventing contamination. It’s psychological, too, as a way of preserving dignity. Of all the “firsts” any space agency hopes to achieve, the first-ever human corpse drifting into frame on a satellite feed is not among them.If a burial must occur—in lunar regolith or by jettisoning into solar orbit—the body will be dutifully tracked and cataloged, treated forevermore as a hallowed artifact of space history.Such gestures are also of relevance to NASA’s plans for off-world mourning; grief and memorial protocols are now part of official crew training. If a death occurs, surviving astronauts are tasked with holding a simple ceremony to honor the fallen—then to move on with their mission.Uncharted RealmsSo far we’ve only covered the “easy” questions. NASA and others are still grappling with harder ones.Consider the issue of authority over a death and mortal remains. On the ISS, it’s simple: the deceased astronaut’s home country retains jurisdiction. But that clarity fades as destinations grow more distant and the voyages more diverse: What really happens on space-agency missions to the moon, or to Mars? How might rules change for commercial or multinational spaceflights—or, for that matter, the private space stations and interplanetary settlements that are envisioned by Elon Musk, Jeff Bezos and other tech multibillionaires?NASA and its partners have started drafting frameworks, like the Artemis Accords—agreements signed by more than 50 nations to govern behavior in space. But even those don’t address many intimate details of death.What happens, for instance, if foul play is suspected?The Outer Space Treaty, a legal document drafted in 1967 under the United Nations that is humanity’s foundational set of rules for orbit and beyond, doesn’t say.Of course, not everything can be planned for in advance. And NASA has done an extraordinary job of keeping astronauts in orbit alive. But as more people venture into space, and as the frontier stretches to longer voyages and farther destinations, it becomes a statistical certainty that sooner or later someone won’t come home.When that happens, it won’t just be a tragedy. It will be a test. A test of our systems, our ethics and our ability to adapt to a new dimension of mortality. To some, NASA’s preparations for astronautical death may seem merely morbid, even silly—but that couldn’t be further from the truth.Space won’t care of course, whenever it claims more lives. But we will. And rising to that grim occasion with reverence, rigor and grace will define not just policy out in the great beyond—but what it means to be human there, too.

June 3, 20258 min readWhite House Budget Plan Would Devastate U.S. Space ScienceScientists are rallying to reverse ruinous proposed cuts to both NASA and the National Science FoundationBy Nadia Drake edited by Lee BillingsFog shrouds the iconic Vehicle Assembly Building at NASA’s Kennedy Space Center in Florida in this photograph from February 25, 2025. Gregg Newton/AFP via GettyLate last week the Trump Administration released its detailed budget request for fiscal year 2026 —a request that, if enacted, would be the equivalent of carpet-bombing the national scientific enterprise.“This is a profound, generational threat to scientific leadership in the United States,” says Casey Dreier, chief of space policy at the Planetary Society, a science advocacy group. “If implemented, it would fundamentally undermine and potentially devastate the most unique capabilities that the U.S. has built up over a half-century.”The Trump administration’s proposal, which still needs to be approved by Congress, is sure to ignite fierce resistance from scientists and senators alike. Among other agencies, the budget deals staggering blows to NASA and the National Science Foundation (NSF), which together fund the majority of U.S. research in astronomy, astrophysics, planetary science, heliophysics and Earth science —all space-related sciences that have typically mustered hearty bipartisan support.On supporting science journalismIf you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.The NSF supports ground-based astronomy, including such facilities as the Nobel Prize–winning gravitational-wave detectors of the Laser Interferometer Gravitational-Wave Observatory (LIGO), globe-spanning arrays of radio telescopes, and cutting-edge observatories that stretch from Hawaii to the South Pole. The agency faces a lethal 57 percent reduction to its $9-billion budget, with deep cuts to every program except those in President Trump’s priority areas, which include artificial intelligence and quantum information science. NASA, which funds space-based observatories, faces a 25 percent reduction, dropping the agency’s $24.9-billion budget to $18.8 billion. The proposal beefs up efforts to send humans to the moon and to Mars, but the agency’s Science Mission Directorate —home to Mars rovers, the Voyager interstellar probes, the James Webb Space Telescope (JWST), the Hubble Space Telescope, and much more —is looking at a nearly 50 percent reduction, with dozens of missions canceled, turned off or operating on a starvation diet.“It’s an end-game scenario for science at NASA,” says Joel Parriott, director of external affairs and public policy at the American Astronomical Society. “It’s not just the facilities. You’re punching a generation-size hole, maybe a multigenerational hole, in the scientific and technical workforce. You don’t just Cryovac these people and pull them out when the money comes back. People are going to move on.”Adding to the chaos, on Saturday President Trump announced that billionaire entrepreneur and private astronaut Jared Isaacman was no longer his pick for NASA administrator—just days before the Senate was set to confirm Isaacman’s nomination. Initial reports—which have now been disputed—explained the president’s decision as stemming from his discovery that Isaacman recently donated money to Democratic candidates. Regardless of the true reason, the decision leaves both NASA and the NSF, whose director abruptly resigned in April, with respective placeholder “acting” leaders at the top. That leadership vacuum significantly weakens the agencies’ ability to fight the proposed budget cuts and advocate for themselves. “What’s more inefficient than a rudderless agency without an empowered leadership?” Dreier asks.Actions versus WordsDuring his second administration, President Trump has repeatedly celebrated U.S. leadership in space. When he nominated Isaacman last December, Trump noted “NASA’s mission of discovery and inspiration” and looked to a future of “groundbreaking achievements in space science, technology and exploration.” More recently, while celebrating Hubble’s 35th anniversary in April, Trump called the telescope “a symbol of America’s unmatched exploratory might” and declared that NASA would “continue to lead the way in fueling the pursuit of space discovery and exploration.” The administration’s budgetary actions speak louder than Trump’s words, however. Instead of ushering in a new golden age of space exploration—or even setting up the U.S. to stay atop the podium—the president’s budget “narrows down what the cosmos is to moon and Mars and pretty much nothing else,” Dreier says. “And the cosmos is a lot bigger, and there’s a lot more to learn out there.”Dreier notes that when corrected for inflation, the overall NASA budget would be the lowest it’s been since 1961. But in April of that year, the Soviet Union launched the first human into orbit, igniting a space race that swelled NASA’s budget and led to the Apollo program putting American astronauts on the moon. Today China’s rapidprogress and enormous ambitions in space would make the moment ripe for a 21st-century version of this competition, with the U.S. generously funding its own efforts to maintain pole position. Instead the White House’s budget would do the exact opposite.“The seesaw is sort of unbalanced,” says Tony Beasley, director of the NSF-funded National Radio Astronomy Observatory (NRAO). “On the one side, we’re saying, ‘Well, China’s kicking our ass, and we need to do something about that.’ But then we’re not going to give any money to anything that might actually do that.”How NASA will achieve a crewed return to the moon and send astronauts to Mars—goals that the agency now considers part of “winning the second space race”—while also maintaining its leadership in science is unclear.“This is Russ Vought’s budget,” Dreier says, referring to the director of the White House’s Office of Management and Budget (OMB), an unelected bureaucrat who has been notorious for his efforts to reshape the U.S. government by weaponizing federal funding. “This isn’t even Trump’s budget. Trump’s budget would be good for space. This one undermines the president’s own claims and ambitions when it comes to space.”“Low Expectations” at the High FrontierRumors began swirling about the demise of NASA science in April, when a leaked OMB document described some of the proposed cuts and cancellations. Those included both the beleaguered, bloated Mars Sample Return (MSR) program and the on-time, on-budget Nancy Grace Roman Space Telescope, the next astrophysics flagship mission.The top-line numbers in the more fleshed-out proposal are consistent with that document, and MSR would still be canceled. But Roman would be granted a stay of execution: rather than being zeroed out, it would be put on life support.“It’s a reprieve from outright termination, but it’s still a cut for functionally no reason,” Dreier says. “In some ways, [the budget] is slightly better than I was expecting. But I had very low expectations.”In the proposal, many of the deepest cuts would be made to NASA science, which would sink from $7.3 billion to $3.9 billion. Earth science missions focused on carbon monitoring and climate change, as well as programs aimed at education and workforce diversity, would be effectively erased by the cuts. But a slew of high-profile planetary science projects would suffer, too, with cancellations proposed for two future Venus missions, the Juno mission that is currently surveilling Jupiter, the New Horizons mission that flew by Pluto and two Mars orbiters. (The Dragonfly mission to Saturn’s moon Titan would survive, as would the flagship Europa Clipper spacecraft, which launched last October.) NASA’s international partnerships in planetary science fare poorly, too, as the budget rescinds the agency’s involvement with multiple European-led projects, including a Venus mission and Mars rover.The proposal is even worse for NASA astrophysics—the study of our cosmic home—which “really takes it to the chin,” Dreier says, with a roughly $1-billion drop to just $523 million. In the president’s proposal, only three big astrophysics missions would survive: the soon-to-launch Roman and the already-operational Hubble and JWST. The rest of NASA’s active astrophysics missions, which include the Chandra X-ray Observatory, the Fermi Gamma-Ray Space Telescope and the Transiting Exoplanet Survey Satellite (TESS), would be severely pared back or zeroed out. Additionally, the budget would nix NASA’s contributions to large European missions, such as a future space-based gravitational-wave observatory.“This is the most powerful fleet of missions in the history of the study of astrophysics from space,” says John O’Meara, chief scientist at the W. M. Keck Observatory in Hawaii and co-chair of a recent senior review panel that evaluated NASA’s astrophysics missions. The report found that each reviewed mission “continues to be capable of producing important, impactful science.” This fleet, O’Meara adds, is more than the sum of its parts, with much of its power emerging from synergies among multiple telescopes that study the cosmos in many different types, or wavelengths, of light.By hollowing out NASA’s science to ruthlessly focus on crewed missions, the White House budget might be charitably viewed as seeking to rekindle a heroic age of spaceflight—with China’s burgeoning space program as the new archrival. But even for these supposedly high-priority initiatives, the proposed funding levels appear too anemic and meager to give the U.S. any competitive edge. For example, the budget directs about $1 billion to new technology investments to support crewed Mars missions while conservative estimates have projected that such voyages would cost hundreds of billions of dollars more.“It cedes U.S. leadership in space science at a time when other nations, particularly China, are increasing their ambitions,” Dreier says. “It completely flies in the face of the president’s own stated goals for American leadership in space.”Undermining the FoundationThe NSF’s situation, which one senior space scientist predicted would be “diabolical” when the NASA numbers leaked back in April, is also unsurprisingly dire. Unlike NASA, which is focused on space science and exploration, the NSF’s programs span the sweep of scientific disciplines, meaning that even small, isolated cuts—let alone the enormous ones that the budget has proposed—can have shockingly large effects on certain research domains.“Across the different parts of the NSF, the programs that are upvoted are the president’s strategic initiatives, but then everything else gets hit,” Beasley says.Several large-scale NSF-funded projects would escape more or less intact. Among these are the panoramic Vera C. Rubin Observatory, scheduled to unveil its first science images later this month, and the Atacama Large Millimeter/submillimeter Array (ALMA) radio telescope. The budget also moves the Giant Magellan Telescope, which would boast starlight-gathering mirrors totaling more than 25 meters across, into a final design phase. All three of those facilities take advantage of Chile’s pristine dark skies. Other large NSF-funded projects that would survive include the proposed Next Generation Very Large Array of radio telescopes in New Mexico and several facilities at the South Pole, such as the IceCube Neutrino Observatory.If this budget is enacted, however, NSF officials anticipate only funding a measly 7 percent of research proposals overall rather than 25 percent; the number of graduate research fellowships awarded would be cleaved in half, and postdoctoral fellowships in the physical sciences would drop to zero. NRAO’s Green Bank Observatory — home to the largest steerable single-dish radio telescope on the planet — would likely shut down. So would other, smaller observatories in Arizona and Chile. The Thirty Meter Telescope, a humongous, perennially embattled project with no clear site selection, would be canceled. And the budget proposes closing one of the two gravitational-wave detectors used by the LIGO collaboration—whose observations of colliding black holes earned the 2017 Nobel Prize in Physics—even though both detectors need to be online for LIGO’s experiment to work. Even factoring in other operational detectors, such as Virgo in Europe and the Kamioka Gravitational Wave Detector (KAGRA) in Japan, shutting down half of LIGO would leave a gaping blind spot in humanity’s gravitational-wave view of the heavens.“The consequences of this budget are that key scientific priorities, on the ground and in space, will take at least a decade longer—or not be realized at all,” O’Meara says. “The universe is telling its story at all wavelengths. It doesn’t care what you build, but if you want to hear that story, you must build many things.”Dreier, Parriott and others are anticipating fierce battles on Capitol Hill. And already both Democratic and Republican legislators have issued statement signaling that they won’t support the budget request as is. “This sick joke of a budget is a nonstarter,” said Representative Zoe Lofgren of California, ranking member of the House Committee on Science, Space, and Technology, in a recent statement. And in an earlier statement, Senator Susan Collins of Maine, chair of the powerful Senate Committee on Appropriations, cautioned that “the President’s Budget Request is simply one step in the annual budget process.”The Trump administration has “thrown a huge punch here, and there will be a certain back-reaction, and we’ll end up in the middle somewhere,” Beasley says. “The mistake you can make right now is to assume that this represents finalized decisions and the future—because it doesn’t.”

The Carbon MORPHEUS is so sleek and streamlined its wind-tunnel test is probably just a bunch of parallel lines. Designed by EPTA, this mid-engine two-seater supercar skips nostalgia and dives straight into aggressive futurism. It looks engineered by a design studio that sees speed not as a metric, but as a shape. The design brief? “Design is the language of dreams.” Which sounds poetic until you realize what they’ve dreamt up could moonlight as a Batmobile in a Nolan film – just sharper, lower, and probably too dangerous for even Bruce Wayne. A manifesto of Italian elegance, they call it. But it leans more toward brutalist cyberpunk with a splash of Lamborghini’s wild DNA and a whisper of Bertone-era silhouette language. That tapered greenhouse? Pure retro-futurism, the kind that could only come from a country that worships both speed and sculpture in equal measure. Designer: EPTA Design The bodywork is entirely carbon fiber, and not just as a flex. The weave is visible from every angle, wrapping around exaggerated wheel arches and angry origami-like creases. It’s not just aesthetics either – it’s structure, soul, and statement. The reflections rolling off the surface are almost liquid, a high-gloss hallucination of light and intent. Look closely and you’ll see the word “VEPTA” near the front wheels, perhaps a clue or a red herring in EPTA Design’s growing mythos. Around the back, things get wild. Four vertical exhausts erupt from a tail section that looks like a spaceship’s afterburner array. There’s no wing, but the rear deck itself swoops like it’s ready to generate lift – or downforce, depending on how many laws of physics it’s trying to bend. A golden center-lock nut anchors the gloss-black wheels, which are wide, multi-spoked, and performance-obsessed. Brake calipers peek out in red, adding a splash of aggression to an otherwise stealth-fighter palette. From above, the canopy has a near-seamless wrap of glass and carbon. It’s tight. Tense. Like a jet cockpit shrunken down and perched just ahead of a carbon fiber storm. The symmetry here is clinical, the panel gaps razor-thin. It doesn’t just look fast; it looks like it was born in a vacuum chamber under military surveillance. EPTA says this is a “mid-engine two-seater supercar,” which automatically tells you what kind of proportions and intent they’re working with. But beyond the silhouette and stance, this thing is abstract sculpture on wheels. It’s less about raw specs and more about shaping emotion through velocity. And yet, it’s not lost in concept art fantasy. The presence is real, the panels are real, the reflections tell you this isn’t just a digital mockup. There’s heat coming out of those pipes in one of the rear shots. That’s combustion. That’s ambition made tactile. You don’t commit this hard to an idea unless you’re serious about chasing it into reality.The post The MORPHEUS Supercar Concept Could Make Batman Rethink His Ride first appeared on Yanko Design.

Humans learn the norms, values and behaviors of society from each other — and Bernt Børnich, founder and CEO of 1X Technologies, thinks robots should learn like this, too. “For robots to be truly intelligent and show nuances like being careful around your pet, holding the door open for an elderly person and generally behaving like we want robots to behave, they have to live and learn among us,” Børnich told the AI Podcast. 1X Technologies is committed to building fully autonomous humanoid robots, with a focus on safety, affordability and adaptability. Børnich explained how 1X Technologies uses a combination of reinforcement learning, expert demonstrations and real-world data to enable its robots to continuously learn and adapt to new situations. NEO, the company’s upcoming robot, can perform household tasks like vacuuming, folding laundry, tidying and retrieving items. It’s built with operational safety at its core, using tendon-driven mechanisms inspired by the human musculoskeletal system to achieve low energy consumption. Børnich highlights the potential for robots to enhance human productivity by helping handle mundane tasks, freeing people up to focus more on interpersonal connections and creative activities. Learn more about the latest in physical AI and robotics at NVIDIA GTC Paris, which takes place from June 10-12. Register to attend humanoid-related sessions, including: “An Introduction to Humanoid Robots” for a deep dive into NVIDIA Isaac GR00T “How Physical AI Is Shaping the Next Generation of Industrial Robots” to learn how robotics leaders are making robots more intuitive and effective across industries Time Stamps 05:18 – 1X Technologies’ approach to robot safety. 11:36 – How world models enable robots to search backwards from the goal. 16:51 – How robots can free humans up for more meaningful activities. 22:29 – NEO answers the door so Børnich can interview a candidate. You Might Also Like…  How World Foundation Models Will Advance Physical AI With NVIDIA’s Ming-Yu Liu AI models that can accurately simulate and predict outcomes in physical, real-world environments will enable the next generation of physical AI systems. Ming-Yu Liu, vice president of research at NVIDIA and an IEEE Fellow, explains the significance of world foundation models — powerful neural networks that can simulate physical environments. Roboflow Helps Unlock Computer Vision for Every Kind of AI Builder Roboflow’s mission is to make the world programmable through computer vision. By simplifying computer vision development, the company helps bridge the gap between AI and people looking to harness it. Cofounder and CEO Joseph Nelson discusses how Roboflow empowers users in manufacturing, healthcare and automotive to solve complex problems with visual AI. Imbue CEO Kanjun Qiu on Transforming AI Agents Into Personal Collaborators Kanjun Qiu, CEO of Imbue, explores the emerging era where individuals can create and use their own AI agents. Drawing a parallel to the PC revolution of the late 1970s and ‘80s, Qiu discusses how modern AI systems are evolving to work collaboratively with users, enhancing their capabilities rather than just automating tasks.