Nathan Labenz × Keerthana Gopalakrishnan

Gemini Robotics 2 advances whole-body control and cross-embodiment transfer, but Keerthana still calls robotics its “GPT-2 moment”: one-shot demonstrations do not yet prove reliable generalization across bodies and changing scenes.Google’s three-model hierarchy separates embodied reasoning, action, and local execution, pointing to controlled factories, warehouses, and professional servicing as early wedges while latency, hand economics, safety, and compounding handoff errors remain unresolved.

Henry Yin × 程曼祺

个人助理正从聊天入口转向主动执行的操作层：Dots月费100/200/500美元，Muse靠免费版本和社交平台分发。GPT-6 Astra在RoboDojo中平均成功率超22%，领先GPT-5.5与π0.5，但算力短缺和安全事故仍是风险。

李翔 × 李丰

中美元首会谈将贸易竞争措施暂缓安排延至次年1月10日，覆盖黑五至元旦购物季，细则待后续机制明确。AI叙事转向安全与可控性，云端模型或先服务高付费编程；中国资金转向固收，而数据中心高融资成本与延期交付仍是美国金融风险线索。

卫诗婕 × 熊友军

人形机器人运动会正从速度竞赛转向自主化与具身智能验证场，天工 Omni 从彩排9秒44迭代至8秒64，“无脸跑”则是在肩关节和功耗约束下自我学习出的最快解。商业化仍受7×24小时运行、甚至99%成功率的工业门槛约束，工业场景将先提供收入与数据，但长程规划、错误反思、数据规模和人形终局仍是待验证变量。

Peter Diamandis × Salim Ismail × Dave Blundin × Alexander Wissner-Gross × Palmer Luckey

Anduril’s $2.2 million wildfire XPRIZE win, including $1 million from Lockheed, shows its autonomy workflow can transfer to emergencies where a 10-minute versus hour-long response requires different tools.Luckey’s specialized, verifiable systems and self-funded product model challenge a market where five companies once captured roughly 80% of major defense acquisition programs, while Anduril’s choice to use rather than build humanoids keeps platform scope and procurement scale in focus.

Anastasis Germanidis × swyx × Vibhu

Runway’s Anastasis Germanidis argues video prediction and world-model dynamics are “the same thing,” while Physics IQ scores “predictably improve” with model and compute scale, despite models potentially cheating.Real-time generation is “inevitable”: Runway’s 24-FPS characters model demonstrates serving-cost and UX advantages, but autoregressive error, simulation fidelity, and Western competitive lag constrain pixel-native and robotics ambitions.

Koji × 徐梦迪

具身智能已出现Scaling Law早期信号：预训练数据从十万小时增至一百万小时后，未见测试集表现改善，但loss仍不等于success rate。徐梦迪将当前阶段比作GPT-1，更看好世界模型与ICL，因为任务难以靠穷举扩展。未来三到五年的关键催化剂是成功率在未见任务上随数据和模型规模上升；最大风险仍是用户可见进步能否启动数据闭环。

Jared Isaacman

NASA is shifting capital from consensus-driven programs toward explicit national objectives, targeting Artemis 3 in summer 2027 and Artemis 4 in 2028.Repeated lunar missions could create demand for landers, rovers, and resource experiments, while public funding concentrates on fission power and nuclear-electric propulsion.SR1 Freedom’s 2028 target, uncertain lunar returns, and China’s ambitions make execution and launch capacity key watchpoints.

Koji × 王家伟

具身智能正从单次演示转向 OOD 适应、steerability 与 context 理解，覆盖即时反馈和长期记忆。深普以 UMI 数据闭环推进真机验证，超过95%轨迹可执行；年底模型的泛化和评测、scaling law 仍待验证，延迟与数据成本构成约束。

卫诗婕 × 郑俊聪

祥峰中国四年不投DeepSeek、智谱、Kimi等大模型，转向AI for science与物理AI，新基金至少50%投具身智能。宇树以成本降一个梯级打开应用场景，量产一致性成为壁垒；中国供应链按周迭代与香港、内地上市，重塑Long China的创新与退出逻辑。