×
Varda says it has proven space manufacturing works — now it wants to make it boring | TechCrunch

Varda says it has proven space manufacturing works — now it wants to make it boring | TechCrunch

When Will Bruey talks about the future, the timelines are shorter than most might imagine. The Varda Space Industries CEO predicts that within 10 years, someone could stand at a landing site and watch multiple specialized spacecraft per night zooming toward Earth like shooting stars, each carrying pharmaceuticals manufactured in space. Within 15 to 20 years, he says, it will be cheaper to send a working-class employee to orbit for a month than to keep them on Earth.

The reason Bruey thinks these scenarios are realistic is because he has watched ambitious business projections unfold before, while working as an engineer at SpaceX.

“I remember the first rocket I worked on at SpaceX was flight three of Falcon 9,” he said at TechCrunch’s recent Disrupt event. The partially reusable, two-stage, medium-lift launch vehicle has since completed nearly 600 successful missions. “If someone had told me ‘reusable rockets,’ and ‘[we’ll see as] many [of these] flights as daily flights out of LAX,’ I would have been like, ‘All right, [maybe in] 15 to 20 years,’ and this feels the same level of futuristic.”

Varda has already proven the core concept. In February 2024, after a months-long regulatory odyssey, the company became only the third corporate entity ever to bring something back from orbit – crystals of ritonavir, an HIV medication – joining SpaceX and Boeing in that exclusive club. It has completed a handful of missions since.

The company brings its pharmaceuticals back to Earth inside the W-1 capsule, a small, conical spacecraft about 90 centimeters across, 74 centimeters high, and weighing less than 90 kilograms (roughly the size of a large kitchen trash can). The company this week launched its fifth capsule ever aboard a SpaceX ride-share mission, hosted by a spacecraft bus that provides power, communications, propulsion, and control while in orbit.

So why manufacture crystals in space? In microgravity, the usual forces that interfere with crystal formation on Earth – like sedimentation and gravity pulling on growing crystals – essentially disappear. Varda says that this gives it much more precise control over crystallization, allowing it to create crystals with uniform sizes or even novel polymorphs (different structural arrangements of the same molecule). These improvements can ostensibly translate into real benefits: better stability, greater purity, and longer shelf life for drugs.

The process isn’t quick. Pharmaceutical manufacturing can take weeks or months in orbit. But once it’s complete, the capsule detaches from the spacecraft bus and plunges back through Earth’s atmosphere at over 30,000 kilometers per hour, reaching speeds above Mach 25. A heat shield made of NASA-developed carbon ablator material protects the cargo inside, and a parachute brings it down for a soft landing.

Techcrunch event

San Francisco
|
October 13-15, 2026

The actual business is pretty prosaic, though, Bruey offered. “Forget about space for a second,” he said. “We just have this magic oven . . . where you can create formulations that you otherwise couldn’t.” Added Bruey of what people often get wrong about Varda, the company isn’t “in the space industry; we’re in-space industry,” he said. Space is “just another place to ship to.”

Worth noting: Varda isn’t discovering new drugs or creating new molecules. It’s aiming to expand the menu of what can be done with existing, approved drugs.

This isn’t speculative science, either. Companies like Bristol Myers Squibb and Merck have been running pharmaceutical crystallization experiments on the International Space Station for years, proving the concept works. Varda says it’s just making it commercial by building the infrastructure to do it repeatedly, reliably, and at a scale that might actually matter to the pharmaceutical industry.

As for why now, two things have changed. First, space launches have become bookable and predictable. “Ten years ago, you would have to get a chartered flight. It was like hitchhiking to get to orbit if you were not a primary mission payload,” Bruey explained. “It’s still expensive today, but [it’s dependable, you can book a slot, and we [have] booked launches years in advance.”

Second, end-to-end space service companies like Rocket Lab started producing satellite buses that could be purchased off the shelf. Buying spacecraft from Rocket Lab and integrating its pharmaceutical manufacturing capsules with them is a major unlock.

Still, only the highest-value products make economic sense. That’s why Varda started with pharmaceuticals; a drug that can command thousands of dollars per dose can absorb the transportation costs. 

The “seven domino” theory

When Bruey talks to members of Congress, which he says he does frequently these days, he pitches what he calls the “seven domino theory.” 

Domino one: reusable rockets. Done. Domino two: manufacturing drugs in orbit and returning them. Domino three is the big one: getting a drug into clinical trials. “It’s a big deal because what it means is perpetual launch.”

This is where Varda’s business model diverges fundamentally from every other space company.

Think about how satellite companies work. SiriusXM launches satellites to broadcast radio. DirecTV launches satellites to transmit television. Even Starlink, with its thousands of satellites, is fundamentally building out a constellation – a network that, once complete, doesn’t require constant launches to function. These companies treat launch as a capital investment. They spend money to place hardware in orbit, and then they’re done.

Varda is different. Each drug formulation requires manufacturing runs. Manufacturing runs require launches. More demand for the drugs means more launches.

This matters because it changes the economics for launch providers. Instead of selling a fixed number of launches to build out a constellation, they have a customer with (theoretically) unlimited demand that grows with success. That kind of predictable, scalable demand helps justify the fixed costs of launch infrastructure and drives down per-launch prices.

Domino four triggers the feedback loop: as Varda scales, costs drop, making the next tier of drugs economically viable. More drugs mean more scale, lowering costs again – a cycle Bruey says will “shove launch costs into the ground.”

Varda’s commercial viability remains unproven, and no space-manufactured drugs are currently on pharmacy shelves. But the virtuous cycle Bruey imagines won’t just benefit Varda. Lower launch costs make space accessible for other industries, including semiconductors, fiber optics, and exotic materials – everything that benefits from microgravity but can’t yet justify the expense.

Eventually, Bruey tells his team, launch costs will get so low that it will be cheaper to put an employee in orbit for a month because creating additional automation would cost more.

“I imagine ‘Jane’ goes to space for a month. It’ll be like [heading to] an oil rig. She works at the drug factory for a month, comes back down, and [becomes] the first person ever to go to space and back where she generate[s] more value than the cost to take her there.”

It’s at that moment, Bruey says, when “the invisible hand of the free market economy lifts us off our home planet.”

The near-death experience

The path to those shooting star drug deliveries nearly ended before it began, Bruey told TechCrunch.

Varda launched W-1 in June 2023 aboard a SpaceX Falcon 9 rideshare mission. The pharmaceutical manufacturing process inside the capsule worked as planned, producing crystals of Form III ritonavir, a specific crystalline structure of the drug that’s difficult to create on Earth. The experiments were completed within weeks.

But then the capsule just . . . stayed in orbit. For six months. The problem wasn’t technical, Bruey said; Varda couldn’t get approval to bring its W-1 capsule home.

The Utah Test and Training Range, where Varda wanted to land, exists to “test weapons and train warriors,” as Bruey put it. Space drugs didn’t fall into that category, so Varda wasn’t a priority customer. When higher-priority military missions needed the range, they bumped Varda’s scheduled landing windows. Each bump invalidated the company’s reentry license with the FAA, requiring it to start the approval process over.

“There were 80 people in the office who had spent two and a half years of their lives on this thing, and it’s in orbit, but we’re not sure if it can come home,” Bruey recalled.

The situation looked bad from the outside. To observers, it seemed like Varda had been reckless and launched without proper approvals. But he said in reality, the FAA had authorized Varda to launch without a finalized reentry license because the agency wanted to encourage the nascent commercial reentry industry.

The FAA had authorized Varda to launch without a finalized reentry license, encouraging the nascent commercial reentry industry.

“They encouraged us to proceed with our launch, with the goal being that we would continue to coordinate that license, as well as the use of reentry timing with the range, while we were in orbit,” Bruey explained.

The real problem was that this was the first commercial land reentry ever attempted. There was no established process for the Utah range to coordinate with the FAA. Both entities felt like they were shouldering all the liability.

Varda explored every alternative it could think of. Water landing? The capsule doesn’t float; they’d lose it. Australia? Possible, and they started those conversations. But Bruey says he made a call: no half measures.

“Either you have to push the boundaries of regulation to create this future, or you don’t,” he said. “In order for Varda to be successful, we need to land on land regularly. So we just sucked it up and said, ‘Let’s figure this one out.”

While its first mission remained stranded in orbit, the company continued production on the next capsule. It kept hiring.

In February 2024, eight months after launch, W-1 finally came home. It landed as originally planned at the Utah Test and Training Range, the first commercial spacecraft to land on a military test range and the first to land on U.S. soil under the FAA’s Part 450 licensing framework, introduced by the agency in 2021 to make commercial space operations more flexible.

Now Varda has landing sites in both the U.S. and Australia, and it’s the first company to receive an FAA Part 450 operator license that lets it reenter the U.S. without resubmitting full safety documentation for each flight.

Meanwhile, Varda has a secondary business that emerged from necessity: hypersonic testing.

Very few objects ever travel through the atmosphere at Mach 25. The environment at those speeds is extreme and unique: Temperatures reach thousands of degrees, creating a plasma sheath around a vehicle. The air itself undergoes chemical reactions as molecules are ripped apart and recombine. This environment can’t be replicated on Earth, even in the most advanced wind tunnels.

The Air Force and other defense agencies need to test materials, sensors, navigation systems, and communications equipment in real hypersonic conditions. Traditionally, that would require dedicated test flights that cost upwards of $100 million each and involve significant risk.

Varda offers an alternative. Its W-1 capsules are already reentering at Mach 25. The company can embed sensors, test new thermal protection materials, or validate equipment in the actual flight environment rather than in approximations. The capsule is akin to a wind tunnel, and the reentry is the test.

Varda has already flown experiments for the Air Force Research Laboratory, including an optical emission spectroscopy payload that took in-situ measurements of the shock layer during reentry.

Investors are, big surprise, excited about Varda’s story. The company raised $329 million as of its Series C round this past July, most of it earmarked for building out the company’s pharmaceutical lab in El Segundo. It’s also hiring structural biologists and crystallization scientists to work on more complex molecules, eventually including biologics like monoclonal antibodies, which Bruey says is a $210 billion market.

A whole lot has to go right between then and now for Varda to elbow its way into that business, as well as to make a dent in the business it’s currently targeting. But if Bruey is right, “then” is closer than most people might right now imagine.

Source link
#Varda #proven #space #manufacturing #works #boring #TechCrunch


The price of eggs has become a marker of the state of the economy. But how about the cost of making eggs? Sure, you might be able to do it for next to nothing in your kitchen with nothing but a pan and a burner. But you could also pay $400 for a robot that will do it for you—as long as you don’t want your eggs scrambled, that is.

A new Kickstarter project has finally given us a Juicero for the robotics era, and it’s called E.G.O.R. (that’s Efficient Gastronomic Operational Robot), brought to you by the startup Cheffy. It’s an automated egg maker that promises perfect cracks every time with no shell in the yolk. It has seven different ways that it can cook them up for you, or you can just use it to do the cracking. You can also pair it with a toaster so your eggs and toast finish at the exact right time.

This isn’t something anyone really needs. Cooking an egg over easy takes about two minutes, so calling this a time-saver is a stretch. And if you really don’t want to use the stove, you can buy a basic egg cooker for around $20 almost anywhere. Maybe you’re bad at cracking eggs, but $400 (well, $200 if you get in on the early bird deal) is a lot to pay to not have to pick the occasional shell pieces out.

But hey, the robotics era needed its Juicero moment. To the creator’s credit: it seems like they’re treating this more like a fun novelty (albeit a pretty pricey one) rather than something revolutionary. They also aren’t making you buy proprietary eggs that are required to make the machine work, so they’ve got the Juicero bros beat there.

We’re increasingly seeing robots positioned to do basic tasks. A company called Weave got attention earlier this year with an $8,000 robot that folds clothes. There’s a company in San Francisco offering house cleanings performed by humanoid robots. Last year, 1x made waves by introducing its Neo humanoid robot, which the Wall Street Journal reported “nearly toppled over” while closing a dishwasher and struggled to open the refrigerator.

All of these machines cost an inordinate amount of money to do things that do not cost the average person that much time or money to accomplish. They’re novel, sure. But you don’t have to pay a premium to be a beta tester.

#Robot #Revolution #Juiceroartifical intelligence,automation,Robotics">The Robot Revolution Gets Its Own Juicero
                The price of eggs has become a marker of the state of the economy. But how about the cost of making eggs? Sure, you might be able to do it for next to nothing in your kitchen with nothing but a pan and a burner. But you could also pay 0 for a robot that will do it for you—as long as you don’t want your eggs scrambled, that is. A new Kickstarter project has finally given us a Juicero for the robotics era, and it’s called E.G.O.R. (that’s Efficient Gastronomic Operational Robot), brought to you by the startup Cheffy. It’s an automated egg maker that promises perfect cracks every time with no shell in the yolk. It has seven different ways that it can cook them up for you, or you can just use it to do the cracking. You can also pair it with a toaster so your eggs and toast finish at the exact right time.  NEW: Cheffy just launched a robot that cracks and cooks your eggs. Egor automates breakfast by cooking eggs 7 ways, while Toastr syncs your toast so everything finishes together. -Automatically cracks eggs with no shells or mess-Cooks eggs 7 ways, from sunny-side up to hard… pic.twitter.com/auE3SnfrS3 — Ritwik Pavan (@ritwikpavan) July 30, 2026  This isn’t something anyone really needs. Cooking an egg over easy takes about two minutes, so calling this a time-saver is a stretch. And if you really don’t want to use the stove, you can buy a basic egg cooker for around  almost anywhere. Maybe you’re bad at cracking eggs, but 0 (well, 0 if you get in on the early bird deal) is a lot to pay to not have to pick the occasional shell pieces out.

 But hey, the robotics era needed its Juicero moment. To the creator’s credit: it seems like they’re treating this more like a fun novelty (albeit a pretty pricey one) rather than something revolutionary. They also aren’t making you buy proprietary eggs that are required to make the machine work, so they’ve got the Juicero bros beat there.

 We’re increasingly seeing robots positioned to do basic tasks. A company called Weave got attention earlier this year with an ,000 robot that folds clothes. There’s a company in San Francisco offering house cleanings performed by humanoid robots. Last year, 1x made waves by introducing its Neo humanoid robot, which the Wall Street Journal reported “nearly toppled over” while closing a dishwasher and struggled to open the refrigerator. All of these machines cost an inordinate amount of money to do things that do not cost the average person that much time or money to accomplish. They’re novel, sure. But you don’t have to pay a premium to be a beta tester.      #Robot #Revolution #Juiceroartifical intelligence,automation,Robotics

Kickstarter project has finally given us a Juicero for the robotics era, and it’s called E.G.O.R. (that’s Efficient Gastronomic Operational Robot), brought to you by the startup Cheffy. It’s an automated egg maker that promises perfect cracks every time with no shell in the yolk. It has seven different ways that it can cook them up for you, or you can just use it to do the cracking. You can also pair it with a toaster so your eggs and toast finish at the exact right time.

This isn’t something anyone really needs. Cooking an egg over easy takes about two minutes, so calling this a time-saver is a stretch. And if you really don’t want to use the stove, you can buy a basic egg cooker for around $20 almost anywhere. Maybe you’re bad at cracking eggs, but $400 (well, $200 if you get in on the early bird deal) is a lot to pay to not have to pick the occasional shell pieces out.

But hey, the robotics era needed its Juicero moment. To the creator’s credit: it seems like they’re treating this more like a fun novelty (albeit a pretty pricey one) rather than something revolutionary. They also aren’t making you buy proprietary eggs that are required to make the machine work, so they’ve got the Juicero bros beat there.

We’re increasingly seeing robots positioned to do basic tasks. A company called Weave got attention earlier this year with an $8,000 robot that folds clothes. There’s a company in San Francisco offering house cleanings performed by humanoid robots. Last year, 1x made waves by introducing its Neo humanoid robot, which the Wall Street Journal reported “nearly toppled over” while closing a dishwasher and struggled to open the refrigerator.

All of these machines cost an inordinate amount of money to do things that do not cost the average person that much time or money to accomplish. They’re novel, sure. But you don’t have to pay a premium to be a beta tester.

#Robot #Revolution #Juiceroartifical intelligence,automation,Robotics">The Robot Revolution Gets Its Own JuiceroThe Robot Revolution Gets Its Own Juicero
                The price of eggs has become a marker of the state of the economy. But how about the cost of making eggs? Sure, you might be able to do it for next to nothing in your kitchen with nothing but a pan and a burner. But you could also pay $400 for a robot that will do it for you—as long as you don’t want your eggs scrambled, that is. A new Kickstarter project has finally given us a Juicero for the robotics era, and it’s called E.G.O.R. (that’s Efficient Gastronomic Operational Robot), brought to you by the startup Cheffy. It’s an automated egg maker that promises perfect cracks every time with no shell in the yolk. It has seven different ways that it can cook them up for you, or you can just use it to do the cracking. You can also pair it with a toaster so your eggs and toast finish at the exact right time.  NEW: Cheffy just launched a robot that cracks and cooks your eggs. Egor automates breakfast by cooking eggs 7 ways, while Toastr syncs your toast so everything finishes together. -Automatically cracks eggs with no shells or mess-Cooks eggs 7 ways, from sunny-side up to hard… pic.twitter.com/auE3SnfrS3 — Ritwik Pavan (@ritwikpavan) July 30, 2026  This isn’t something anyone really needs. Cooking an egg over easy takes about two minutes, so calling this a time-saver is a stretch. And if you really don’t want to use the stove, you can buy a basic egg cooker for around $20 almost anywhere. Maybe you’re bad at cracking eggs, but $400 (well, $200 if you get in on the early bird deal) is a lot to pay to not have to pick the occasional shell pieces out.

 But hey, the robotics era needed its Juicero moment. To the creator’s credit: it seems like they’re treating this more like a fun novelty (albeit a pretty pricey one) rather than something revolutionary. They also aren’t making you buy proprietary eggs that are required to make the machine work, so they’ve got the Juicero bros beat there.

 We’re increasingly seeing robots positioned to do basic tasks. A company called Weave got attention earlier this year with an $8,000 robot that folds clothes. There’s a company in San Francisco offering house cleanings performed by humanoid robots. Last year, 1x made waves by introducing its Neo humanoid robot, which the Wall Street Journal reported “nearly toppled over” while closing a dishwasher and struggled to open the refrigerator. All of these machines cost an inordinate amount of money to do things that do not cost the average person that much time or money to accomplish. They’re novel, sure. But you don’t have to pay a premium to be a beta tester.      #Robot #Revolution #Juiceroartifical intelligence,automation,Robotics

The price of eggs has become a marker of the state of the economy. But how about the cost of making eggs? Sure, you might be able to do it for next to nothing in your kitchen with nothing but a pan and a burner. But you could also pay $400 for a robot that will do it for you—as long as you don’t want your eggs scrambled, that is.

A new Kickstarter project has finally given us a Juicero for the robotics era, and it’s called E.G.O.R. (that’s Efficient Gastronomic Operational Robot), brought to you by the startup Cheffy. It’s an automated egg maker that promises perfect cracks every time with no shell in the yolk. It has seven different ways that it can cook them up for you, or you can just use it to do the cracking. You can also pair it with a toaster so your eggs and toast finish at the exact right time.

This isn’t something anyone really needs. Cooking an egg over easy takes about two minutes, so calling this a time-saver is a stretch. And if you really don’t want to use the stove, you can buy a basic egg cooker for around $20 almost anywhere. Maybe you’re bad at cracking eggs, but $400 (well, $200 if you get in on the early bird deal) is a lot to pay to not have to pick the occasional shell pieces out.

But hey, the robotics era needed its Juicero moment. To the creator’s credit: it seems like they’re treating this more like a fun novelty (albeit a pretty pricey one) rather than something revolutionary. They also aren’t making you buy proprietary eggs that are required to make the machine work, so they’ve got the Juicero bros beat there.

We’re increasingly seeing robots positioned to do basic tasks. A company called Weave got attention earlier this year with an $8,000 robot that folds clothes. There’s a company in San Francisco offering house cleanings performed by humanoid robots. Last year, 1x made waves by introducing its Neo humanoid robot, which the Wall Street Journal reported “nearly toppled over” while closing a dishwasher and struggled to open the refrigerator.

All of these machines cost an inordinate amount of money to do things that do not cost the average person that much time or money to accomplish. They’re novel, sure. But you don’t have to pay a premium to be a beta tester.

#Robot #Revolution #Juiceroartifical intelligence,automation,Robotics

The prevailing narrative in Silicon Valley is that Chinese AI models like Kimi K3 and DeepSeek R1 came out of nowhere. But the truth is that even if you never travel to China, there’s an easy way to get a sense of what Chinese AI researchers are up to: log on to X.

In the past year, I’ve seen a steady increase of Chinese artificial intelligence researchers joining X and participating in the wider discourse about AI development and commercialization. Take Moonshot AI, the company behind Kimi K3, as an example. On Thursday, I was able to quickly find around 30 accounts on X run by people who claim to be currently affiliated with Moonshot, including two of the startup’s cofounders, as well as half a dozen former employees and collaborators. And these are not dormant accounts: They talk about Moonshot’s big releases and incremental research papers, make friends with Western researchers, and share details about their personal life and hobbies.

Moonshot is arguably the Chinese AI lab that’s most active on X, but you will easily find executives and technical staff from other leading Chinese startups, like Minimax and Z.ai. Even several employees at DeepSeek, who reportedly are unable to leave China because the government confiscated their passports, regularly post on X about the latest AI news and share job listings to recruit new talent.

“X is the best community” for AI discussions right now, says Meng Fanqing, cofounder of the startup Evolvent AI and a former intern at Moonshot, where he worked on the Kimi K2.5 model. “There are both Chinese and overseas audiences, and the recommendation algorithm works pretty well for showing you content in the same circle that you are in,” he says. Compared with Chinese platforms, where it’s often hard to have serious technical discussions, Chinese researchers like Meng say they would rather spend their time on X.

Meng says he started actively posting on X in mid-2025. At the start of the year, DeepSeek’s R1 model went viral around the world, shining a bright spotlight on China’s AI industry. Many Chinese researchers seized the moment and joined X to start sharing their perspective. “Around that time, all the Chinese researchers are starting to think about branding on an international level,” Meng says.

Chinese AI researchers may also be helping fill a growing gap left by researchers at OpenAI and Anthropic, who have become more reluctant to share their thinking online as their employers have grown into behemoth companies, says Tiezhen Wang, an independent AI analyst and former researcher at Hugging Face. “I guess they feel that their model architecture and how they train the model is kind of a secret, so they’re posting less,” Wang explains. As a result, Chinese researchers are getting more attention from audiences on X who are hungry to make sense of what’s happening in the fast-moving world of AI.

Why Choose X?

One reason X may have become popular among Chinese AI researchers is that there’s not really a comparable social platform in China for high-quality technical discussions about AI and other emerging technologies.

Some people say they used to post on Zhihu, a Quora-like platform that once branded itself as the premier destination for educated professionals to show and tell. But since 2020, Zhihu has increasingly pivoted to hosting fiction content, and expert contributors have gradually fallen off. For example, Deng Yu, a Chinese mathematician who was awarded a Fields Medal this year, publicly wrote on Zhihu in 2018 that he decided to stop posting on the site due to the rise of low-quality content.

#Chinese #Researchers #Finding #Voicemade in china,china,artificial intelligence,x,twitter,rednote,deepseek">Chinese AI Researchers Are Finding Their Voice on XThe prevailing narrative in Silicon Valley is that Chinese AI models like Kimi K3 and DeepSeek R1 came out of nowhere. But the truth is that even if you never travel to China, there’s an easy way to get a sense of what Chinese AI researchers are up to: log on to X.In the past year, I’ve seen a steady increase of Chinese artificial intelligence researchers joining X and participating in the wider discourse about AI development and commercialization. Take Moonshot AI, the company behind Kimi K3, as an example. On Thursday, I was able to quickly find around 30 accounts on X run by people who claim to be currently affiliated with Moonshot, including two of the startup’s cofounders, as well as half a dozen former employees and collaborators. And these are not dormant accounts: They talk about Moonshot’s big releases and incremental research papers, make friends with Western researchers, and share details about their personal life and hobbies.Moonshot is arguably the Chinese AI lab that’s most active on X, but you will easily find executives and technical staff from other leading Chinese startups, like Minimax and Z.ai. Even several employees at DeepSeek, who reportedly are unable to leave China because the government confiscated their passports, regularly post on X about the latest AI news and share job listings to recruit new talent.“X is the best community” for AI discussions right now, says Meng Fanqing, cofounder of the startup Evolvent AI and a former intern at Moonshot, where he worked on the Kimi K2.5 model. “There are both Chinese and overseas audiences, and the recommendation algorithm works pretty well for showing you content in the same circle that you are in,” he says. Compared with Chinese platforms, where it’s often hard to have serious technical discussions, Chinese researchers like Meng say they would rather spend their time on X.Meng says he started actively posting on X in mid-2025. At the start of the year, DeepSeek’s R1 model went viral around the world, shining a bright spotlight on China’s AI industry. Many Chinese researchers seized the moment and joined X to start sharing their perspective. “Around that time, all the Chinese researchers are starting to think about branding on an international level,” Meng says.Chinese AI researchers may also be helping fill a growing gap left by researchers at OpenAI and Anthropic, who have become more reluctant to share their thinking online as their employers have grown into behemoth companies, says Tiezhen Wang, an independent AI analyst and former researcher at Hugging Face. “I guess they feel that their model architecture and how they train the model is kind of a secret, so they’re posting less,” Wang explains. As a result, Chinese researchers are getting more attention from audiences on X who are hungry to make sense of what’s happening in the fast-moving world of AI.Why Choose X?One reason X may have become popular among Chinese AI researchers is that there’s not really a comparable social platform in China for high-quality technical discussions about AI and other emerging technologies.Some people say they used to post on Zhihu, a Quora-like platform that once branded itself as the premier destination for educated professionals to show and tell. But since 2020, Zhihu has increasingly pivoted to hosting fiction content, and expert contributors have gradually fallen off. For example, Deng Yu, a Chinese mathematician who was awarded a Fields Medal this year, publicly wrote on Zhihu in 2018 that he decided to stop posting on the site due to the rise of low-quality content.#Chinese #Researchers #Finding #Voicemade in china,china,artificial intelligence,x,twitter,rednote,deepseek

Chinese AI models like Kimi K3 and DeepSeek R1 came out of nowhere. But the truth is that even if you never travel to China, there’s an easy way to get a sense of what Chinese AI researchers are up to: log on to X.

In the past year, I’ve seen a steady increase of Chinese artificial intelligence researchers joining X and participating in the wider discourse about AI development and commercialization. Take Moonshot AI, the company behind Kimi K3, as an example. On Thursday, I was able to quickly find around 30 accounts on X run by people who claim to be currently affiliated with Moonshot, including two of the startup’s cofounders, as well as half a dozen former employees and collaborators. And these are not dormant accounts: They talk about Moonshot’s big releases and incremental research papers, make friends with Western researchers, and share details about their personal life and hobbies.

Moonshot is arguably the Chinese AI lab that’s most active on X, but you will easily find executives and technical staff from other leading Chinese startups, like Minimax and Z.ai. Even several employees at DeepSeek, who reportedly are unable to leave China because the government confiscated their passports, regularly post on X about the latest AI news and share job listings to recruit new talent.

“X is the best community” for AI discussions right now, says Meng Fanqing, cofounder of the startup Evolvent AI and a former intern at Moonshot, where he worked on the Kimi K2.5 model. “There are both Chinese and overseas audiences, and the recommendation algorithm works pretty well for showing you content in the same circle that you are in,” he says. Compared with Chinese platforms, where it’s often hard to have serious technical discussions, Chinese researchers like Meng say they would rather spend their time on X.

Meng says he started actively posting on X in mid-2025. At the start of the year, DeepSeek’s R1 model went viral around the world, shining a bright spotlight on China’s AI industry. Many Chinese researchers seized the moment and joined X to start sharing their perspective. “Around that time, all the Chinese researchers are starting to think about branding on an international level,” Meng says.

Chinese AI researchers may also be helping fill a growing gap left by researchers at OpenAI and Anthropic, who have become more reluctant to share their thinking online as their employers have grown into behemoth companies, says Tiezhen Wang, an independent AI analyst and former researcher at Hugging Face. “I guess they feel that their model architecture and how they train the model is kind of a secret, so they’re posting less,” Wang explains. As a result, Chinese researchers are getting more attention from audiences on X who are hungry to make sense of what’s happening in the fast-moving world of AI.

Why Choose X?

One reason X may have become popular among Chinese AI researchers is that there’s not really a comparable social platform in China for high-quality technical discussions about AI and other emerging technologies.

Some people say they used to post on Zhihu, a Quora-like platform that once branded itself as the premier destination for educated professionals to show and tell. But since 2020, Zhihu has increasingly pivoted to hosting fiction content, and expert contributors have gradually fallen off. For example, Deng Yu, a Chinese mathematician who was awarded a Fields Medal this year, publicly wrote on Zhihu in 2018 that he decided to stop posting on the site due to the rise of low-quality content.

#Chinese #Researchers #Finding #Voicemade in china,china,artificial intelligence,x,twitter,rednote,deepseek">Chinese AI Researchers Are Finding Their Voice on X

The prevailing narrative in Silicon Valley is that Chinese AI models like Kimi K3 and DeepSeek R1 came out of nowhere. But the truth is that even if you never travel to China, there’s an easy way to get a sense of what Chinese AI researchers are up to: log on to X.

In the past year, I’ve seen a steady increase of Chinese artificial intelligence researchers joining X and participating in the wider discourse about AI development and commercialization. Take Moonshot AI, the company behind Kimi K3, as an example. On Thursday, I was able to quickly find around 30 accounts on X run by people who claim to be currently affiliated with Moonshot, including two of the startup’s cofounders, as well as half a dozen former employees and collaborators. And these are not dormant accounts: They talk about Moonshot’s big releases and incremental research papers, make friends with Western researchers, and share details about their personal life and hobbies.

Moonshot is arguably the Chinese AI lab that’s most active on X, but you will easily find executives and technical staff from other leading Chinese startups, like Minimax and Z.ai. Even several employees at DeepSeek, who reportedly are unable to leave China because the government confiscated their passports, regularly post on X about the latest AI news and share job listings to recruit new talent.

“X is the best community” for AI discussions right now, says Meng Fanqing, cofounder of the startup Evolvent AI and a former intern at Moonshot, where he worked on the Kimi K2.5 model. “There are both Chinese and overseas audiences, and the recommendation algorithm works pretty well for showing you content in the same circle that you are in,” he says. Compared with Chinese platforms, where it’s often hard to have serious technical discussions, Chinese researchers like Meng say they would rather spend their time on X.

Meng says he started actively posting on X in mid-2025. At the start of the year, DeepSeek’s R1 model went viral around the world, shining a bright spotlight on China’s AI industry. Many Chinese researchers seized the moment and joined X to start sharing their perspective. “Around that time, all the Chinese researchers are starting to think about branding on an international level,” Meng says.

Chinese AI researchers may also be helping fill a growing gap left by researchers at OpenAI and Anthropic, who have become more reluctant to share their thinking online as their employers have grown into behemoth companies, says Tiezhen Wang, an independent AI analyst and former researcher at Hugging Face. “I guess they feel that their model architecture and how they train the model is kind of a secret, so they’re posting less,” Wang explains. As a result, Chinese researchers are getting more attention from audiences on X who are hungry to make sense of what’s happening in the fast-moving world of AI.

Why Choose X?

One reason X may have become popular among Chinese AI researchers is that there’s not really a comparable social platform in China for high-quality technical discussions about AI and other emerging technologies.

Some people say they used to post on Zhihu, a Quora-like platform that once branded itself as the premier destination for educated professionals to show and tell. But since 2020, Zhihu has increasingly pivoted to hosting fiction content, and expert contributors have gradually fallen off. For example, Deng Yu, a Chinese mathematician who was awarded a Fields Medal this year, publicly wrote on Zhihu in 2018 that he decided to stop posting on the site due to the rise of low-quality content.

#Chinese #Researchers #Finding #Voicemade in china,china,artificial intelligence,x,twitter,rednote,deepseek

Post Comment