Updated · 2 episodes · 2 shows · 2 source notes
Poke Robotics
Overview
Poke Robotics, or 破壳机器人, is the robotics startup founded by Xu Huazhe and discussed in 166: 许华哲再次具身创业:不想错过最大的西瓜 and 当具身智能走到十字路口|对谈苏度、蚂蚁灵波、自变量、破壳:四种一线判断. The company is tracked as a household/general-service robotics route aimed at Physical AGI rather than a narrow industrial machine.
Current Profile
Poke’s current profile combines a household-robot Physical AGI thesis with a specific capability boundary around complex contact. Xu credits Astra-style general models with semantic and spatial strength, but treats complex physical contact, reliable manipulation, and deployment economics as the remaining hard problems that a robotics company must solve.
Key Characteristics
- Pursues general household robotics as a route toward Physical AGI.
- Treats active use and repeated value as more meaningful than shipments or demos.
- Uses product safety boundaries to exclude high-risk direct body-care tasks in early stages.
- Sees general models as useful for semantic and spatial reasoning but limited on complex physical contact.
- Distinguishes short-term post-training deployments from the longer ambition of foundation-model-style general robotics.
Evidence
- Founding evidence: 166: 许华哲再次具身创业:不想错过最大的西瓜 says Xu founded Poke Robotics after leaving Xinghaitu to pursue general household robots.
- Product evidence: 166: 许华哲再次具身创业:不想错过最大的西瓜 frames the first stage as building the body, models, and product definition together, likely through a wheeled dual-arm form before full bipedal humanoids.
- Safety evidence: 166: 许华哲再次具身创业:不想错过最大的西瓜 says early products should avoid feeding, infant handling, massage, and turning people over.
- Astra evidence: 当具身智能走到十字路口|对谈苏度、蚂蚁灵波、自变量、破壳:四种一线判断 records Poke testing Astra and finding strong semantic/spatial generalization but weaker complex physical contact.
- Commercialization evidence: 当具身智能走到十字路口|对谈苏度、蚂蚁灵波、自变量、破壳:四种一线判断 records Xu separating post-training deployments by deployment hours and site replication from the longer foundation-model success-rate climb.
Qualifications
Poke’s household and Physical AGI route is still a founder thesis, not established customer-scale proof. The Astra test discussion is useful but source-scoped; it does not settle general-model capability across standardized robotics tasks.
What Changed
- Converted the page to the synthesis-v1 entity schema.
- Added Poke’s Astra-response evidence and complex-contact boundary.
- Added the distinction between short-term post-training deployments and long-term general robot models.
Relationships
- Xu Huazhe - founder.
- Physical AGI, AI Native Robotics, and Unified Robot Models - technical route.
- Home Service Robots, Consumer Robotics Full Stack, and Household Robot Data Flywheel - product and deployment context.
- General Model Robot Boundary - Astra-triggered boundary around semantic/spatial versus physical-contact capability.
- Robot Deployment Success Economics - commercialization standard Poke helps define in the new source.
Sources
2 source notes across 2 shows
- 166: 许华哲再次具身创业:不想错过最大的西瓜 晚点聊 LateTalk
- 当具身智能走到十字路口|对谈苏度、蚂蚁灵波、自变量、破壳:四种一线判断 十字路口Crossing