Podcast Atlas
entity Updated 2026-07-08 Tags: Podcast, Media, Business

张小珺Jùn｜商业访谈录

张小珺Jùn｜商业访谈录 is the podcast/show context for 130. 张月光创业两年首次访谈：妙鸭不是AI Native产品、流程到上下文设计、One Way Door和乙女游戏, 131. 印奇出任阶跃星辰董事长的访谈：聪明人的诱惑、取舍、超长链路残酷淘汰赛、阶跃函数和超多元方程, 132. 对星海图创始人高继扬的3小时访谈：鲶鱼、曾国藩、Waymo与Momenta的两面、一只狼与许华哲的离开, 133. 对谢赛宁的7小时马拉松访谈：世界模型、逃出硅谷、AMI Labs、两次拒绝Ilya、杨立昆、李飞飞和42, 134. 【数据的综述】和谢晨聊，新时代的石油、历史、版图、数据金字塔、定价与Recipe, 145. 口述SpaceX开发史：和前高管洪力德聊，马斯克用人观、最大IPO、太空与AI、人类文明扩张前奏？, 140. 对姚顺宇的4小时访谈：请允许我小疯一下！在Anthropic和Gemini训模型、技术预测、英雄主义已过去, 135. 和自然选择创始人Tristan聊，Elys、赛博分身、灵魂、Context的获取与流动和AI社交网络, 137. 对洪乐潼的4小时访谈：AI for Math、把数学变成Lean、数学天书中的证明、直觉、被创造与被发现的, 138. 对罗福莉3.5小时访谈：AI范式已然巨变！OpenClaw、Agent范式很吃后训练、卡的分配、组织平权, and 139. 【Agent的综述】和苏煜聊Agent技术史、OpenClaw Moment、边界的消弭和社会的辐射. The ingested episodes cover 张月光’s product transition from 妙鸭 to Docky, Yin Qi’s appointment as chair of StepFun, Gao Jiyang’s Xinghaitu robotics path, Xie Saining’s AMI Labs world-model startup story, 谢晨’s 光轮智能 data/simulation view of robotics, Louis Hong / 洪力德’s SpaceX manufacturing and space-infrastructure account, Yao Shunyu / 姚顺宇’s frontier-model training perspective from Anthropic and Google DeepMind, Tristan’s Natural Selection / 自然选择 product thesis around EVE, Elys, Cyber Avatars, and AI Social Networks, Hong Letong / 洪乐潼’s Axiom thesis around AI For Math, Lean Theorem Prover, Auto-Formalization, and Formal Verification, Luo Fuli / 罗福莉’s Xiaomi model-team thesis around Open Claw, Open Cloud, Agent Post-Training, Agent RL, Memo VR, and flatter AI Organization Design, and Su Yu / 苏煜’s agent-history thesis around Language Agent, OpenClaw Moment, Continual Learning, World Models, and Universal Digital Agent.

Episode 141 adds Freda / Friday’s investor lens through Token Maxxing, AI Economic Diffusion, and Human Connection Under AI. It connects the show’s agent and model-company thread to market metrics, software revaluation, hyperscaler capex, public/private AI valuation, organization redesign, and the emotional experience of trying to keep up with accelerating tools.

Episode 143 adds He Xiaopeng / 何小鹏 and XPeng / 小鹏汽车 through Physical AI, Stitched AI Architecture, XPeng Iron, XPeng GX, and Humanoid Robot Commercialization. It connects the show’s embodied-AI thread to autonomous-driving route changes, robot full-stack manufacturing, physical-world data costs, vehicle terminals, and CEO-led organization change.

Episode 144 adds Yang Meng / 杨萌 and Anker Innovations / 安克创新 through Consumer Electronics Lifecycle, Third Type Company, Creator Culture, In-Memory Computing For Edge AI, On-Device Model Hierarchy, True Smart Home, Household Security Robots, and Enterprise Prearranged Agents. It connects the show’s embodied-AI and organization-design thread to a consumer-electronics operator’s path from Amazon-channel products toward edge-side models, smart-home autonomy, security robots, and AI-native internal workflows.

Episode 142 adds Dai Yusen / 戴雨森 through Agent Harness, Model Harness Co-Evolution, Agent Marketplace, AI Investment Metrics, and AI Economic Diffusion. It connects the show’s agent and model-company thread to an investor’s view of why Claude Code, Codex, Manus, and Open Cloud matter as product/runtime layers, why terminal AI returns remain unresolved, and why the next major AI company may not resemble ByteDance.

Episode 136 adds a quarterly frontier-model synthesis through AGI Three Acts and Model As Operating System. It connects the show’s existing agent/coding thread to a stronger claim that Claude Code and Codex mark AGI’s second act, that Anthropic, OpenAI, Google, Meta, and xAI are competing for platform-level model infrastructure, and that Intelligence Devaluation may hit knowledge work faster than institutions expect.

Connections