
何嘉斌
Key Views & Dialogues
Vol.185 Industry Watch 34 | Where Are the Non-Consensus Opportunities in AI Companion Robots?
AI companion hardware is entering a deliverable third wave, with Ropet using a “100% fun, 0% useful” weak-robot design to build attachment and usage before adding stronger intelligence. Nearly 10,000 deliveries, global all-age demand and two low-budget validations show early commercial traction, while proprietary behavioral data, hardware iteration and the path from 1 TOPS to 10 TOPS remain decisive tests.
View summary (7)
李丰与何嘉斌把AI陪伴硬件定位为本轮技术周期的“第三波”:2023年看大模型,2024年主要看具身智能机器人(同时也有人讨论Agent),今天则要看谁能交付产品、创造需求并赚到钱。 Once foundational technology has absorbed decades of accumulated “vital energy” and made a major leap, its growth rate tends to flatten; the grandest software-and-hardware stories are also usually the hardest to commercialize, so opportunities spread into vertical applications. “It may not be the most AI-heavy product, but it must involve relatively new technology, scenarios, applications and ways to make money.”
Ropet的非共识不是把大模型塞进毛绒玩具,而是先用可爱、沉默的“弱机器人”获得开机率,再用闭源的行为数据长出智能。 Its camera continuously captures facial expressions, touch, smiles and gestures within roughly half a meter and a field of view of about 120 degrees; if 100,000 users eventually maintain high daily usage, this dataset would exceed the one-on-one pet-interaction samples available from ordinary home surveillance. 何嘉斌’s sequence is explicit: “First you have the hardware, then the behavioral data matched to that hardware, and only then can you call it intelligence.”
产品被定义为“百分百好玩,零分有用”,核心商业选择是避开工具硬件的配置与性价比竞争。 Ropet does not speak human language directly; it uses eye contact, movement and vocal tone to mimic the perception of a pet or a child aged zero to three. Strong conversational ability would make it feel like “a highly intelligent agent stuffed into a plush toy,” or like a two- or three-year-old suddenly arguing with its parents, undermining the sense of being a pet and the trust that comes with it. The team wants to trigger not the dopamine decay that follows novelty, but the urge to care, oxytocin and the satisfaction of “being needed.”
目前接近一万台的交付已经推翻了最初单一的用户画像,显示陪伴产品可能是全球、全年龄市场。 The team initially targeted urban women aged 24 to 30 who lived alone, but more than 50% of its core users in Japan are women aged 50 to 70, while many North American buyers purchase Ropet as a gift for children; roughly half of users still fit the original urban-woman profile. 何嘉斌 therefore calls the company a “future pet company”: desktop, bag-mounted, in-car and roaming home formats could all belong to the same category of “silicon-based pets.”
项目的早期商业验证来自极低预算下两次连续正反馈,而不是先靠融资堆出来的规模。 After four months of keeping the six-person, debt-laden company alive by borrowing from old shareholders, the team spent less than RMB10,000 on a booth at a Shanghai IP licensing fair, where 5 prototypes became the hottest new product. Old shareholders then continued to support the company, while individual shareholders raised roughly RMB10M more. At CES, the founder visited the US for the first time and hired an interpreter for about RMB1,500 a day; the product generated 6B views globally in backend statistics over 6-7 days, making the commercial value of something that could be touched, pre-ordered and delivered visible.
中国消费电子供应链构成了AI硬件最实在的结构优势,但首代产品仍必须承受漫长而昂贵的迭代。 Ropet went through 3-4 major hardware-selection changes from its first prototype to nearly 10,000 delivered units, with each change potentially causing at least a 2-month delay. To create genuine eye contact alone, the team insisted on 2 screens positioned at roughly a 15-degree angle and supporting dual-screen display; it is still working through servo noise, heat dissipation under the plush covering and quiet-fan design. “Be bolder”: building a roughly 50-point framework from mature components first, then iterating quickly through Shenzhen’s supply chain, is what makes mass production at least a year later possible.
智能升级是一条带条件的信任曲线,而不是首代产品把能力拉满。 Today’s 1 TOPS of edge compute must simulate vision, hearing, touch and gravity sensing at the same time. 何嘉斌 imagines that if the device can move to 10 TOPS at the same cost a year from now, while the company accumulates in-the-wild data from 20,000-30,000 units, the pet may grow from “3 years old” to “10 years old” and begin language interaction in an appropriate way. 李翔 contrasts this with the decline in Character.AI engagement and the stepwise progression of autonomous driving from L2 and L3 toward L4 and L5: consumer AI needs to “ship with low expectations first, then climb the AI ladder gradually.”
50. Six Months After CES Went Viral: Ropet on Weathering the Storm, Copycats, and the Meaning of Companionship
Ropet converted CES attention into financing, 3,000 August deliveries, nearly 40 employees, and Japanese crowdfunding above ¥50M. It rejects conversation, using gaze, touch, personality formation, and 11 edge models on 1 TOPS. JD.com entry around October or Singles’ Day will test durable Chinese demand after production delays from heat, noise, and charging.
View summary (8)
- Ropet went from prototype to CES breakout, then faced copycats and a drawn-out fundraising process. It closed one financing round this year, with two closings planned, delivered 3,000 units in August, and expanded the team from 7 people in its early days to nearly 40. Its Japanese crowdfunding campaign reached ¥26M (about RMB1.3M) on day one and has now exceeded ¥50M, targeting ¥70M-¥80M and potentially ¥100M. Founder 何嘉斌 called it a Japanese product of the year and a phenomenon for an overseas, non-functional product. The $299 price point was deliberate: “Price can select users,” filtering for the core customer base.
- The most important product decision was to categorically avoid conversation: Ropet tested Doubao, Gemini, and ChatGPT, but none met its internal experience standard. 何嘉斌’s logic is that once the product speaks with omniscience, “the ecological niche changes”—from a small creature that needs care into a problem-solving tool. One Kickstarter user even said: “If you make it talk, I will never trust this company again.” With AI conversation, “one mistake is enough to lose trust.”
- “De-modeling” is explicitly positioned as an ecological defense against the giants. Faced with the potential pressure of an OpenAI-Mattel partnership around Barbie, 何嘉斌’s answer is that a startup whose edge rests entirely on a foundation model “has no chance of competing.” Ropet’s non-conversational value—gaze, touch, and personality formation—could instead become irreplaceable. Its edge device runs about 11 vertical small models on 1 TOPS of compute; for now, in-situ data is processed as abstract behavioral data, with volunteer-authorized raw data potentially used to train a cloud model once the installed base reaches roughly 20,000-40,000 units.
- The benchmark is not an AI toy but Japan’s LOVOT: “95% daily active use over 1,000 days,” with users continuing to pay a monthly fee. This is an “oxytocin product,” not a “dopamine product.” Dopamine products rely on constant novelty and stimulation; oxytocin products come from “devoting your own energy to another life.” 何嘉斌 believes China’s AI-toy sales, despite rising 600% sequentially in 1H 2025, still reflect trial and gifting rather than durable demand. Ropet will not fire the first shot domestically; it plans to enter China with JD.com around October or Singles’ Day.
- The brutality of Chinese competition is documented in full. 何嘉斌’s previous venture, Luka, was disassembled and its BOM analyzed; competitors said, “We can make it for one-third of your price.” White-label products can gradually eat away at the low end, while enforcement is expensive. The only answer is to “build enough detail into the product experience—every detail is a moat,” even if that means paying several times more for packaging with a better hand feel.
- The founder’s personal ledger is unusually exposed. When he took over as CEO, most of the company’s previous financing had already been spent and it carried RMB700K in debt. He assumed the legal representative role and the liabilities, borrowed from legacy shareholders, and paid himself RMB300K a month to cover payroll. Early support came from friends and supply-chain partners; one backer invested $50K, followed later by $1.5M in additional financing. After CES, he received multiple term sheets but would not sign unlimited joint-and-several liability: he had earned only RMB1M-plus to RMB2M-plus in a decade of employment, while a Shunyi apartment he bought fell from roughly RMB4M to RMB1.8M. During CES he slept only 1-2 hours a night for 6 straight days, passed out on the flight home, and said his hair “went white overnight.”
- The contrarian product philosophy is “slow technology”; his university thesis was titled “How to Make a Phone Last a Lifetime.” “The best things in nature are slow things—the setting sun, morning dew, flowers blooming and falling.” Technology, too, needs “things that happen slowly.” Ropet is a cyber plant: emotional data nourishes its personality, and “life is memory”; digital memories can be passed on to the next generation.
- 卫诗婕 believes China will increasingly resemble Japan’s lifestyle and that the companionship market could become a Japan-like market; 何嘉斌 argues China’s population can support a market of similar scale, while Japan is the better place to refine the product. The discussion noted that China now has fewer children under 4 than pets, and 1st-tier cities have 6 million pets. Women account for 70%-80% of crowdfunding users, and 何嘉斌 wants “women” to occupy a section of the map of technology equity.