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.”

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.”

Stranger Things fans are worried about the ultimate fate of main character Eleven, played by Millie Bobby Brown, and even some think the teen might not make it out alive at the end of the series.Eleven has been an integral part of the Duffer Brothers’ smash hit Netflix series since it first hit streaming in the summer of 2016.Viewers immediately gravitated toward the show for its spooky atmosphere and mystery-centered plot, nostalgic ’80s vibes and lovable cast of Goonies-esque teen characters.Fans have loved Eleven ever since she made her first appearance in Season 1, Episode 1, “The Vanishing of Will Byers,” and they've watched the unsure, traumatized and quiet young girl transform into a confident, spunky teen with powerful telekinetic abilities over the course of four seasons.Now though, with the series’ fifth and final season set to air later this year, longtime fans are worried about what the end of the show might spell for Eleven, now also known as Jane Hopper.Does Eleven Die in Stranger Things?Nothing about the fate of the core Stranger Things characters is known for sure at this time. However, that hasn’t stopped viewers from theorizing and speculating.During an appearance on U.K. talk show The Jonathan Ross Show in March 2024, Millie Bobby Brown may have inadvertently hinted that her character dies at the end of the show thanks to some questionable phrasing.While discussing the final season, the actress hinted, “I know how she ...” before catching herself and correcting, “I know what happens to my character.”The initial wording of “I know how she” sparked fans’ ears, many of whom thought the actress almost blurted out, “I know how she dies.”Brown also worried fans during a 2024 interview with Capital Radio, when she admitted she discovered her character’s fate after “kind of [forcing] myself into the writers’ room.”“I saw my ending and thought, ‘Oh,’ and then I walked away very slowly,” she cryptically teased.For years fans have speculated about the ending of Stranger Things, particularly about which of the core group might not make it out alive.Some fan theories suggest that Eleven is ultimately doomed, and might be forced to lock herself in the Upside Down forever to close the gate between the Upside Down and the real world, or will die heroically closing the gate and saving her friends and loved ones.Others believe Will Byers, who was the first to venture into the Upside Down and appears to still be connected to it as well as the series’ villain Vecna, will ultimately die in the finale.Of course, these are just fan theories. Hopefully, all the kids end up just fine and there's a big, happy ending!Stranger Things Season 5 will pick up after the epic events of Season 4, in which the kids learned about the evil Vecna, who ended the season by opening a hellish portal between the town of Hawkins and the Upside Down.The fifth season will be released in three parts: The first four episodes will hit Netflix on Nov. 26, three episodes will begin streaming on Dec. 25 and the series finale will air on Dec. 31.Sitcom Moments That Were Surprisingly DarkSitcoms such as The Simpsons and The Golden Girls are often seen as light-hearted comedies, but these darker TV moments offer a different, deeper perspective.Gallery Credit: Ryan ReichardGet our free mobile appREAD MORE: TV Shows Everyone Loves That Are Actually BadChild Stars Who Quit ActingStacker compiled this list of 25 child actors who quit show business, pulling from historical news coverage to include everyone from Mary-Kate and Ashley Olsen to Carrie Henn, who played the little girl in Aliens.Gallery Credit: Sophia June

We're expecting a ton of AAA and indie game announcements during Summer Game Fest 2025, but the Southeast Asian Games Showcase will highlight the best games coming out of the region. The event will offer some exclusive reveals, new trailers and update announcements on more than 45 titles from Southeast Asian game developers. You can catch all the news through the YouTube livestream, which will start Saturday, June 7 at 3PM ET. The showcase teased some updates with Polychroma Games' Until Then, a narrative adventure game set in a fictional Filipino city, Bit Egg's Lost & Found Co., a quirky hidden object game, and kopiforge's SEDAP! A Culinary Adventure that's equal parts co-op cooking and combat.  Right before the Southeast Asian Games Showcase, you'll be able to catch the Latin American Games Showcase at 2PM ET and the Women-Led Games Showcase at 1PM ET, both on June 7. For the rest of the Summer Game Fest 2025, tune into Engadget's coverage, where we'll have in-person reports on some of the biggest gaming announcements expected during the event.This article originally appeared on Engadget at https://www.engadget.com/gaming/how-to-watch-the-southeast-asian-games-showcase-at-summer-game-fest-2025-181518526.html?src=rss

Are we on the cusp of truly affordable Teslas? A YouTuber flew a drone over Tesla's Fremont test track and saw what may be its new cheap model racing around it.As spotted by Teslarati, new flyover footage from an account that calls itself "Met God in Wildnerness" — which along with frequent Fremont flyovers also publishes what appear to be Christian missives in Chinese — there's a good chance that the disguised vehicle could be a prototype for Tesla's purported affordable model."A disguised Tesla car testing on test track," the video's caption reads. "We could be seeing the new low cost model."Though the blog describes the vehicle as resembling a "compact Model Y," the black-and-white machine looks to our eyes like an elongated Volkswagen Beetle. In the video, it's seen stopping and going on the track, which is located off to the side of Tesla's manufacturing plant in Fremont, California, followed by a disguised Cybertruck.In its writeup, Teslarati notes that although there's a possibility the car is the new performance version of the Model Y, its compact size suggests it could be a prototype for the long-teased affordable model.As InsideEVs pinpointed back in April, Tesla admitted in its abysmal first-quarter earnings report for 2025 that switching over production lines to make room for the new Performance Model Y, which resulted in "several weeks of lost production" on the upgrade, was also paramount as it seeks to produce cheaper versions of its cars."During the switchover, we also prepared our factories for the launch of new models later this year," the report's fine print read. "Given economic uncertainty resulting from changing trade policy, more affordable options are as critical as ever."When discussing that earnings report in an investor call, CFO Vaibhav Taneja said that production of the cheaper model — which may utilize design aspects of Model 3 and Model Y — is "planned for June."As with most of the company's timelines, we're taking that one with several grains of salt — especially because CEO Elon Musk was mocking the idea of a $25,000 Tesla, which he'd been promised for years, less than a year ago.Right now, we don't know any specifics about what will constitute an "affordable" Tesla or when it will hit production lines or dealerships — but that flyover video could be our first glimpse of what's to come.More on Tesla: Tesla Can't Find Legal Places to Store All Its Unsold CybertrucksShare This Article

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.Samsung announced the new Galaxy S25 Series back in January, a three-phone lineup with a fourth variation that was teased a month later, which you can preorder right now with a $50 Amazon gift card before its May 30 release. Out of those four, the Galaxy S25 is the most basic model with the lowest price, and currently, it's hitting a new record low price of $624.99 (originally $799.99), according to price-tracking tools. SAMSUNG Galaxy S25 128GB AI Smartphone, Unlocked Android, AI Camera, Fast Processor, ProScaler Display, Long Battery Li $624.99 at Amazon $799.99 Save $175.00 Get Deal Get Deal $624.99 at Amazon $799.99 Save $175.00 The S25 has some upgrades from the S24 lineup: The best features in the OS are shared by all four phones, so you can still enjoy the best features from One UI 7, including exclusive media and notifications features from the cheapest option. Two of the best are Now Bar and Now Brief, which show you brief alerts or notifications personalized to your feed and activities.The Galaxy S25 runs on the Snapdragon 8 Elite for Galaxy, has three rear lenses, including a 50MP wide-angle lens, and a 6.2-inch and 120Hz OLED display. The RAM is a generous 12GB, with the storage options starting at 128 GB. Currently, the mint color is the cheapest available. PCMag gave the Galaxy S25 an "excellent" review for its useful AI features, a bright screen, good battery life, and excellent build quality. You can expect about 14 hours of juice from this phone. Keep in mind it does not support the S Pen stylus. Given its small size of 5.78 by 2.78 by 0.28 inches, it's a great option for those who prefer smaller phones.

If you were wondering whether Mark Hamill might play Luke Skywalker once again as a Force ghost, think again. The legendary actor has ruled out a return to the world famous sci-fi franchise, saying it’s now time for Star Wars to move past legacy characters and focus on the future.In an interview with ComicBook.com to promote his new movie, The Life of Chuck, Hamill was asked about a potential return as Luke’s Force ghost in upcoming Rey-led Star Wars movie, New Jedi Order. His answer was loud and clear: no chance. After all, Luke left his robes behind in 2017's The Last Jedi.Mark Hamill first played Luke Skywalker in 1977's Star Wars: Episode IV - A New Hope.“I am so grateful to George [Lucas] for letting me be a part of that back in the day, the humble days when George called Star Wars ‘the most expensive low-budget movie ever made,’” Hamill replied.“We never expected it to become a permanent franchise and a part of pop culture like that. But my deal is, I had my time. I’m appreciative of that, but I really think they should focus on the future and all the new characters.“And by the way, when I disappeared in [The Last Jedi], I left my robes behind. And there’s no way I’m gonna appear as a naked Force ghost.”PlayRey, however, is set to return to the world of Star Wars in the Sharmeen Obaid-Chinoy-directed sequel to the universally panned Star Wars: Episode 9 - The Rise of Skywalker. It will tell the story of Rey as she looks to rebuild the Jedi Order roughly 15 years after the events of that film.We know next to nothing about New Jedi Order. In January 2024, speaking to AlloCiné, Ridley teased a little more about her new film: "Once I knew what the story was and everything, I knew that it was something I really wanted to do," she said. “I think it's a really fantastic exploration of the Star Wars world. It's a really cool way of taking the story on in a bit of a different direction."In the shorter term, The Mandalorian and Grogu is due out 2026. Star Wars: Starfighter, Shawn Levy's Star Wars movie starring Ryan Gosling, is due out in 2027.Every Upcoming Star Wars Movie and TV ShowAt 2023's Star Wars Celebration, Lucasfilm announced three new Star Wars feature films: a Dave Filoni-directed New Republic film set in his Mando-verse, a Dawn of the Jedi movie led by James Mangold, and the aforementioned New Jedi Order.There is still so much that can and will change when it comes to upcoming Star Wars movies and TV shows. But what we can be sure of is Mark Hamill playing Force Luke won’t be a part of them.Photo by Sunset Boulevard/Corbis via Getty Images.Wesley is Director, News at IGN. Find him on Twitter at @wyp100. You can reach Wesley at wesley_yinpoole@ign.com or confidentially at wyp100@proton.me.

The hugely popular first-person shooter game Valorant is getting some long-awaited upgrades as the game approaches its 5-year anniversary. In the latest dev update, the Valorant team announced the long-anticipated replay feature. They’ll function about the same as replays in other online shooters like Counter-Strike 2, allowing players to revisit their previous matches and watch them from the perspective of any player in the game or from a free-roaming camera. Replays will initially launch on PC with patch 11.06 in September and will come to consoles later in the year. After initially developing and updating the game in Unreal Engine 4 for more than ten years, the dev team announced a port to Unreal Engine 5 with patch 11.02 around the end of July. The Valorant devs say this will bring higher frame rates and faster future patch downloads without changing the current gameplay feel. In their video update, the team teased that moving to Unreal 5 will enable a whole host of possibilities in the future. Competitive players will take particular note of new anti-smurfing initiatives. Smurfing is the practice of higher-ranked players using lower-ranked accounts to gain a competitive advantage by playing against players far below their skill level. In an attempt to further curb this behavior, Valorant will be asking suspicious accounts to complete multi-factor authentication and is exploring requiring it on all competitive accounts. These guardrails will launch later this year with more details to come. Additionally, users who wish to report other players for smurfing can now select a specific "Rank/Matchmaking Abuse" category to help the better identify those accounts. The update also showcased some gameplay balancing measures involving character abilities, teased a new competitive map and highlighted the team’s vision for the future of Valorant in esports.This article originally appeared on Engadget at https://www.engadget.com/gaming/pc/valorant-is-finally-adding-replays-and-upgrading-to-unreal-engine-5-173931808.html?src=rss

Playdate Season Two is here, bringing with it two new games for the quirky yellow handheld every week until July 3. And if the first two titles are any indication of what this season will be like, it's sure to be a great one. Season Two kicked off on May 29 with the arcade action game Fulcrum Defender — from the studio behind FTL: Faster Than Light and Into the Breach — and the delightfully chill Dig! Dig! Dino!. The two games couldn't be more different from each other, but they're both bangers in their own right. Panic also released Blippo+, which can only be described as a fever dream of cable TV, with the first drop of Season Two, and it is amazingly bizarre. Fulcrum Defender Subset Games "Survive for 10min!" sounded almost like a threat when I first started reading through Fulcrum Defender's How To Play guide. Between all the on-screen information you need to pay attention to, the many different types of enemies that'll be attacking and the various weapon upgrades you can earn over the course of a run, there's a lot to take in, and I braced myself for a tense and complicated playing experience. But, while that may be closer to the case on Hard Mode, I found that Fulcrum Defender wasn't all that punishing of a shooter on Normal Mode. It's a challenge, for sure, but one with a surprisingly achievable goal that I was able to enjoy without losing my mind. At least, not until crossing the 10-minute mark. After that, all hell breaks loose. In Fulcrum Defender, you're positioned at the center of a circular arena and have to fend off a continuous swarm of enemies. Your shield will take damage any time an enemy collides with it, and once enough have breached that zone, it's game over. To avoid that, you need to shoot them down one by one, using the crank to aim your weapon and the D-pad to shoot. Some enemies can be taken out in one shot, but others — distinguished by their filled-in appearance — require multiple shots. Over time, you'll earn weapon upgrades to build out a more powerful defense system, with options like large, guided projectiles and a flail that can knock out several enemies in one sweep. It's unexpectedly addicting. The music is beautiful and calming, giving the whole thing a pleasant atmosphere despite the fact that you're surrounded by enemies at any given moment and trying not to die. Once I realized it was absolutely possible to survive 10 minutes and even go beyond that, I got sucked into the loop of trying over and over to beat my high scores. I'd love to see a global leaderboard for this game at some point, because I just know I'd be floored by how long some players will be able to last. If you liked this one and want to know a little more about the making of it, be sure to check out our interview with Jay Ma, the co-founder of Fulcrum Defender developer Subset Games. Dig! Dig! Dino! Dom2D & Fáyer I can't think of anything I'd rather be doing right now than pretending to be a paleontologist and casually digging for bones. No thoughts, just dig. That's exactly what Dig! Dig! Dino! has going on, and it's awesome. You're working as part of a crew (made up entirely of anthropomorphic animals) at the site of some really unusual dinosaur fossils, and it's your job to dig up new bones and artifacts. Once you've got the entire skeleton of a particular dinosaur, you can scan it in the lab to reveal what it was like when it was alive. That information, coupled with the peculiar artifacts scattered around the site, paints a picture of some pretty strange activities that went on there long ago. For example, some of these dinosaurs seem to have had crystals growing out of their bodies, and it looks like they were warned about the asteroid extinction event. Fishy! The gameplay is extremely low stakes — this is one for when you just want to zone out playing something that'll keep your hands busy. You're equipped with a shovel, a drill and a radar gadget for detecting items beneath the surface, and have no time-sensitive goals to hit. You only have so much energy, though, which will be consumed with each use of your tools. When you run out, the round is over. But you can visit each site as many times as you need to in order to find all of the dinosaur pieces hidden there, so it can be a really casual undertaking if you want it to be. It's a really nice time, with a fun story to tie it all together. You'll get a solid few hours of playtime out of this, too, and the simplicity of it all means you can put it down and come back to it later without having to rack your memory to figure out where you left off. I loved this one. Blippo+ Panic What can one even say about Blippo+? This bizarre "1-bit television" experience came as a bonus with the first Season Two games, and it is something. Panic first teased it back in December 2024 as a Steam title, but here it is for the Playdate now, complete with a roster of channels playing hallucinatory programs and Femtofax, an interactive message board of sorts where you can find affirmations, neighborhood drama, chatter among amateur astronomers and more. Panic describes it as being "comparable to an old episode of The Twilight Zone," but it's more like an old episode of The Twilight Zone if it were made by Tim & Eric and aired after midnight on Adult Swim. I think I am obsessed with it? I'm really interested to see where this goes. I would totally park my Playdate in a dock (but not the Stereo Dock </3) on my desk and leave Blippo+ running in the background all day if it has enough fresh material to sustain it. The song playing alongside the endlessly scrolling Blippo+ TV guide screen is already stuck in my head, and I don't hate it. The program guide with this week's schedule is online, if you're curious about what's going on right now.This article originally appeared on Engadget at https://www.engadget.com/gaming/playdate-season-2-review-fulcrum-defender-dig-dig-dino-and-blippo-140036697.html?src=rss