Innovators
3 Curated Dialogues
陈哲
Key Views & Dialogues
184: Chen Zhe’s “Embodied AI Quarterly 26Q3”: After Astra’s Shock, Will General Models Eat Robotics?
GPT-6 Astra directly outputs end-effector poses that inverse dynamics converts into motor signals; on Galaxy General’s 10 Robot Dojo tasks, supervising π0.5 raised success from 26% to 48%. That challenges the “physical AI is bigger than digital AI” valuation premise, though VLA may retain a role and Astra remains too slow, with low absolute success, for real-world deployment.
View summary (8)
- The quarter’s biggest shock was GPT-6 Astra, which showed that general models can enter physical AI: frontier labs are inclined to solve AGI first and treat physical AI as a subproblem. Astra does more than break tasks down; it can directly output end-effector poses, which are converted into motor signals through inverse dynamics—a step previously thought to belong to embodied models. Open-source replications showed a significantly higher success rate than Claude; Manqi also cited a separate Robot Dojo comparison in which Astra outperformed some leading embodied models. “If you haven’t seen the miracle, you simply won’t believe it.”
- Chen Zhe sees Astra’s impact in 3 layers: directly generating policies, supervising and correcting VLA, and—most “hair-raising” of all—acting as a robotics researcher. In Galaxy General’s case, Astra directly controlled 10 Robot Dojo tasks with a 26% success rate; supervising π0.5’s outputs raised that to 48%. The third-layer vision is to have Astra build simulation environments in Blender, generate objects and their properties, and debug motion-control policies: “In the physical world, AI can train AI.” Chen believes physical RSI is coming and is a direction to watch in Q4; he calls it “the latest fundraising concept.”
- The valuation story for embodied-model companies is under pressure: the assumption that “physical AI is bigger than digital AI,” which supports high valuations for Physical Intelligence and Generalist, may not hold. Astra suggests that “general models are a more formidable force, and physical AI may just be a subset of general models.” VLA may still have a role, but not necessarily be delivered by standalone VLA companies; general models may also distill into on-device models. At the same time, the market could expand and deployment accelerate, so this is not necessarily bad news for every startup.
- Chen expects every leading general-model company to make moves in robotics: there are rumors around Kimi and Zhipu, Liang Wenfeng has talked about physical AGI, and teams such as ByteDance Seed and Tencent RoboticsX may also pivot toward a generative-AI approach. The compute gap remains vast: leading embodied-AI companies may have GPUs in the thousands, versus hundreds of thousands or even millions at top general-model companies—a gap of 2 to 3 orders of magnitude. A friend at a Frontier Lab also cautioned that conclusions drawn from its data volumes and compute “do not apply to the vast majority of startups.”
- World Labs and AMD have signed an $8.2B acquisition agreement structured as an all-stock deal, but the transaction has not closed. Chen sees it as a reference point for exit discussions across the sector; there are also rumors that Anthropic may acquire Physical Intelligence. China has “basically no decent acquisition examples”; one leading embodied-AI founder said the company had raised RMB3B in cash but was in no rush to spend it, and wanted to wait for the bubble to burst before buying teams cheaply.
- Unitree’s IPO surged and then plunged—from a market cap of more than RMB440B at the open to more than RMB170B—but Chen says he “doesn’t feel much,” while Astra has made him more confident in a RMB100B valuation. Revenue in the first half was RMB1.152B, up 48% year over year. He drew a parallel with Nvidia: “No matter how successful OpenAI gets, Nvidia still benefits.” Meanwhile, regulators are tightening scrutiny of embodied-AI companies waiting to list. Chen warns that doing things contrary to the industry’s development just to generate revenue gets the priorities backward—like forcing OpenAI to monetize through a harness in the GPT-1 era.
- The embodied ICL capabilities shown by GEN-1.5 and S1 could cut the post-training needed for deployment to minutes of demonstration video. Two surprising findings: out-of-distribution tasks may be learned from a single demonstration, and long sequences of combined skills—such as packing and moving objects—can also be taught through video demonstrations. A growing industry view is that architecture may not be decisive: given data of the same quality and scale, VLA and WAM could both produce equally good policy models.
- At the end of July, the FCC announced restrictions on some advanced robots manufactured overseas, covering robots weighing more than 2 kg with mobility, perception, connectivity, and control capabilities. Chen says the share of core components that must be sourced domestically “should be” 65%; companies including Boston Dynamics, which had previously supported restrictions, have also appealed or raised objections, though the transcript does not clearly identify Figure as an appellant. The near-term impact is limited: already-certified Unitree models such as the G1, R1, and H2 can continue to be sold without major hardware or software changes. Over time, the policy could push companies to build factories in the US and bring suppliers overseas, but China’s supply-chain advantage, built over many years, will be “hard to change through short-term restrictive policies.”
170: With 陈哲’s “Embodied Quarterly 26Q2”: The World-Model Gale Keeps Blowing, and Those Who Don’t Want Labels
World models are moving from concept to product and paradigm. Nvidia’s Cosmos 3, released on June 1, is the quarter’s benchmark: the market’…
View summary (8)
- World models are moving from concept to product and paradigm. Nvidia’s Cosmos 3, released on June 1, is the quarter’s benchmark: the market’s first fully open-source Omni Model, using a Mixture of Transformers architecture to “stitch” autoregressive reasoning and diffusion generation into a unified model, with super, nano, and edge versions released at once for deployment. 陈哲’s view: “Rather than saying world models are overturning VLA, they are bringing new capabilities and ideas to existing SOTA models.” Pi 0.7’s connection to a lightweight world model and Generalist Gen One’s refusal to be categorized as either VLA or world model are both signs of convergence.
- Honor’s clean sweep of the Yizhuang Marathon podium was a preview of the big manufacturers’ entry. All 3 of its autonomous-navigation “Lightning” robots finished in roughly 50 minutes, versus Tiangong’s 2 hours and 40 minutes to win the remotely controlled race in 2025—a more than 3x improvement in a year. Honor kept the motors from overheating with race-specific motors and precision liquid cooling, spending “significantly more than the other 2 teams.” The host’s question was whether a body-maker’s ability to IPO this year could become a resource-sufficiency line; 陈哲’s answer was that Xiaomi, XPeng, Li Auto, and other major end-device manufacturers are already entering seriously, and humanoid robotics is “evolving from single-technology entrepreneurship into a system of systems engineering and systematic execution.”
- Figure’s 100-plus-hour livestream sorting 130,000 parcels marked the humanoid robot industry’s “zero-to-one” moment. At roughly 1 parcel every 3 seconds, flipping and flattening soft packages—a motion that was “basically unsolvable with the technology stack of 4 or 5 years ago”—the robots went beyond the pick-and-place limits of suction-cup systems. The teleoperation debate is beside the point: “I believe every humanoid robot goes through a long period of teleoperation before entering a deployment scenario like this,” while remote intervention is already the norm in robotaxis. 星动纪元’s comparable deployments with China Post and SF Express are “in no way behind” those of US companies.
- Dexterous hands are becoming the next competitive battleground, strengthening the case for direct drive. At ICRA Vienna, 五G released a second-generation direct-drive hand with 20 degrees of freedom, much better backdrivability and thermal management, and roughly half the volume of ShaPa’s hand—positioning it as “Unitree in dexterous hands.” “Whoever can provide reliable, stable, inexpensive dexterous hands faster will become the industry’s de facto standard faster.” Domestic major manufacturers’ move to Tesla’s tendon-drive route is more about organizational risk management: “If Tesla turns out to be right, and you try a new direction, who takes responsibility for that mistake?” Optimus Gen 3’s hand continues to iterate, while leading high-DOF hands ship only a few thousand units.
- Data is the next positioning battle. Collection paradigms have moved rapidly over 2 years from ALOHA real-robot teleoperation to UMI without a body, egocentric video, and full-body motion capture; “data we start collecting today could translate into breakthroughs in models 3 or 6 months from now.” Dexterous-hand data is highly dependent on hand geometry and sensor choices, making standardization by third-party data vendors difficult. Unlike lidar point clouds, which can run through a single pipeline, “dexterous hands and humanoid bodies will most likely be locked in a long-term contest and long-term coexistence.”
- Gen One was the quarter’s most discussed model advance. Generalist collected 500,000 hours of UMI-style, body-free data and trained an end-to-end model from scratch without a pretrained model, lifting success rates on complex long-horizon tasks from the low 60% range to 99% and claiming to have “found the key to scaling.” Peter Florence explicitly rejected both the VLA and world-model labels; the confidence behind that route reflects a US venture ecosystem that “rewards the first person to do something.”
- The endgame question is why embodied models would not belong to general-purpose model companies. “What reason do we really have to believe that, many years from now, the model for embodied AI will not be provided by Anthropic, OpenAI, or Google?” Cosmos 3’s 10K-GPU-scale training is already the ceiling for embodied-AI research compute, yet not an especially large budget for leading LLM companies. OpenAI’s high-profile robotics restart draws from the Sora team, whose internal positioning has always been world models, with dexterous hands likely to be a major investment area. 陈哲 expects embodied models to end up like language models: “high-quality open source plus top-tier closed source, with no room for other models to survive.”
- Unitree’s IPO approval could become the industry’s valuation anchor, while the next 6 months may be the last window for major manufacturers to enter. “If you do not start entering this market in 2026, by the time humanoid robots are broadly deployed in 2027 or 2028, you will have no position.” The China-specific paradox is that investors tolerate neither 10 years without commercialization nor a company that is not full-stack: “Every embodied-brain company is announcing that it is building a body.” Yet the companies with the largest market caps are precisely not full-stack: Nvidia is not, and Anthropic may not be either if it goes public.
E178｜The Smart-Glasses Battle at CES: Product Substance Beats AI
China’s consumer-hardware edge now spans premium pricing and iteration speeds several times to 10x faster, while patent portfolios and cross-licensing reduce the drag from iRobot’s older patents. At least 20 smart-glasses vendors showed physical products, but AI remains mostly assistive; unless glasses escape phone dependence, share may accrue to phone ecosystems rather than standalone brands.
View summary (7)
- The key question at this CES was not whether a product had AI, but whether AI redefined the product—and whether the hardware fundamentals held up. PLAUD uses OpenAI API to turn a roughly $100-plus card-sized recorder into a summarization tool, making it a product that would not exist without AI; but 泓君 ultimately went back to a recorder from 10 years ago because PLAUD kept dying at crucial moments. The conclusion is simple: “It is a consumer product first, and only then a smart device.”
- China’s consumer-hardware advantage has expanded beyond mature supply chains into product innovation, brand premiums and iteration speeds several times to 10x faster. The shift is clearest in robot vacuums: leading players such as Roborock and Dreame are experimenting with robotic arms, while their patent portfolios and cross-licensing arrangements have weakened the constraints imposed by iRobot’s older patents. 陈哲 believes that once innovation speed reaches an order-of-magnitude gap, overseas manufacturers’ “loss of innovative momentum is almost inevitable.”
- Robot-vacuum arms are currently better understood as an entry point for differentiated R&D than as an established consumer need. Current payloads are around 300–400 grams, making even a heavy slipper difficult to pick up; reliable, fast pick-and-place remains a core robotics challenge, and 陈哲 believes another 2 or 3 product generations may be needed. 泓君 argues that consumers currently care more about automatic water connection, battery life and complete cleaning. The long-term value lies in exploring the possible form factors for $100s-to-$1,000 home-service robots, not in immediately driving sales by “picking up socks.”
- The AI-glasses field is crowded, but it has yet to prove suitable for standalone startups or venture capital. At least 20 companies with physical products exhibited at CES, most of them Chinese, while similar supply chains quickly produced Meta Ray-Ban copycats; AI remains mostly an assistive feature, and AR still lacks a sufficiently compelling use case. 泓君 believes the harder-to-copy advantages may instead be comfort, fit across face shapes, fashion design, SKU breadth and retail networks, while a single electronic feature “is not enough to support commercial success.”
- If smart glasses need a phone for connectivity, compute and AI, their “most likely destiny is to become an accessory to the smartphone.” TWS offers a precedent: Apple, Samsung, Huawei, Xiaomi, OPPO and vivo should capture the overwhelming majority of the true-wireless-earbuds market, squeezing independent brands and traditional audio makers. Unless glasses can break free from the phone and become a new computing platform, the eventual share is more likely to accrue to phone ecosystems than to today’s large crop of standalone brands.
- The winning edge in companion robots is not superior intelligence, but the emotional value created jointly by product design and “emotional intelligence.” Products such as Looi and LOVOT suggest that users may want pet-like responses, cuteness and precisely calibrated movement—not a talking smart speaker; saying “it does not need to be intelligent” does not mean AI is irrelevant, because voice, vision, identity and emotion recognition still need to support natural feedback. The problem is that demand is highly fragmented, and there is no clear winning formula yet.
- The more certain industrial opportunities may lie in foundational toolchains and organizations that genuinely know how to build hardware, rather than in the hottest consumer concepts of the year. Nvidia released the Cosmos physical-AI foundation model and is using tools such as Isaac Sim to help customers train robot models, creating a feedback loop that supports GPU demand; DJI-linked companies and Bambu Lab show that product strength comes from first-principles engineering culture, talent development and continuous iteration. CES can point to the hot themes of the next 1 or 2 years, but it cannot answer “who is worth investing in,” because “many companies will disappear.”