Podcast Atlas
entity Updated 2026-08-24 Topics: Technology, Politics, Culture, Science

硅谷101

E249｜Token经济转点：OpenClaw、Hermes到本地自研的Agent进化之路 adds an agent-economics installment. The show uses 东旭 / Dongxu, 张宏江 / Zhang Hongjiang, Open Claw, Hermes Agent, Slock/Raft, local DeepSeek models, Fable 5, and GLM 5.2 to examine the shift from Token Maxxing to Token Efficient Agent Workflow, while connecting agent engineering to AI Inference Cost Structure, Model Routing Cost Control, Local Agent Execution, Multi-Agent Collaboration, Agent Native Software, and Human Agency Under AI.

E248｜一个“催发货”AI要跑通260步，和阿里瓴羊彭新宇聊聊中国式FDE adds a China-enterprise AI deployment installment. The show interviews 彭新宇 of 瓴羊 to examine Chinese-Style FDE / 中国式 FDE, Enterprise Growth Agent / 企业级增长 Agent, the 260-step “催发货” workflow, data foundations, and outcome-based agent deployment in customer service, marketing, sales, and operations.

E247｜对话盛颖：xAI，Infra的浪漫，SGLang，开源，平权与“甄嬛传” adds an AI-infrastructure founder installment. The show uses 盛颖, SGLang, xAI, Redix ARK, LM-SYS, and LM Arena to connect Radix Attention, Day-Zero Model Support, Agent RL, AI Infrastructure As Product, and Open-Source AI Democratization.

E246｜何谓蒸馏？聊聊硅谷如何看中国开放模型逼近前沿 adds an open-model and infrastructure installment. The show uses 王铁镇, Keith Zhai, Moonshot AI, Kimi K3, and OpenRouter to examine Model Distillation / 模型蒸馏, Scaling Efficiency, Open-Weight Commercial Licensing, Closed Model API Moat Pressure, Model Sovereignty / 模型主权, Open Model Safety Governance, and Agent Inference Workload.

E245｜藏在大模型背后的新闻人：GPT们的回复是这样写出来的 adds a media-labor and model-behavior installment. The show uses Face, 东尼 / Tony, and Bianca to explain Content Engineering, AI Answer Evaluation, AI Interaction Internationalization, AI Trainer Labor, and Consensus-Trained Art Boundary as ways that journalists, editors, and creative workers shape how ChatGPT, Gemini, and other assistants reply.

E227｜美国医疗市场AI争夺战：巨头押注，创业公司能赢吗？ adds a U.S. healthcare AI installment. The show uses 张璐 / Zhang Lu, 周叶冰 / Zhou Yebing, OpenEvidence, ChatGPT Health, ChatGPT for Healthcare, and Claude for Healthcare to examine Physician Administrative Burden, Healthcare AI Infrastructure, Medical Billing and Coding Automation, HIPAA-Constrained Medical AI, Evidence-Grounded Medical RAG, Vertical Medical Small Models, and doctor-led Human Judgment Under AI.

E226｜聊聊DeepMind创始人哈萨比斯：一个科学家与失控的AI竞赛 adds an AI founder biography and technology-history installment. The show uses 周建功 / Zhou Jiangong, Sebastian Mallaby, Demis Hassabis, and DeepMind to connect Reinforcement Learning AGI Path, AlphaGo, AlphaFold, DeepMind Acquisition Choice, Google DeepMind, Gemini, and Scientific Ideal vs AI Arms Race.

E239｜SpaceX要让太空算力从科幻走向现实，但它划算吗？ adds a space and AI-infrastructure installment. The show uses Louis Hong / 洪力德, 刘彬晏 / Liu Binyan, SpaceX, Starship, Starlink, and Starbase to examine whether orbital compute can move from science-fiction narrative into a real business through Orbital Data Center Economics, Orbital Data Center Thermal Management, Orbital Compute Governance, and Moon-Mars Strategy Split.

E237｜央视和FIFA谈判纷争背后，体育赛事转播权的博弈与生意 adds a sports-media-rights installment. The show uses 中央广播电视总台 / China Central Television, FIFA, 联想 / Lenovo, 朱晓东, and 张斌 to examine Sports Media Rights, Sports Rights Business Model, Sports Rights Sublicensing, Sports Rights Bubble, and Sports Broadcast Public Access around the 2026 FIFA World Cup rights negotiation.

E238｜聊聊Harness时代AI-First的组织架构：从信任人到信任AI adds an AI organization and agent-infrastructure installment. The show interviews Peter, 陈凯, and Clark of Creo to examine Harness Engineering, AI-First Organization, agent-driven development, AI-driven QA, go-to-market bottlenecks, and the shift from trusting people by role to trusting governed AI systems.

硅谷101 is the podcast/show context for E244｜端到端vs上下分层：机器人路径之争，正在转向？, E229｜从手工作坊到全球第一：中国动力电池逆袭史, E243｜特朗普“缓刑”红牌之外，美国资本如何硬控全球足坛, E241｜跑鞋技术迭代史：马拉松跑进2小时，靠人还是靠鞋？, E242｜最快半年AI跑通自进化？与陈天桥首席科学家聊聊硅谷模型必争之地, E240｜OpenAI联手PE砸下40亿美元，聊聊硅谷最火新职位FDE, and E225｜SaaS业数千亿市值蒸发：AI如何变革组织架构？. In E244, it uses Han Zheng / 韩正 and 速度科技 / Sudu Technology to compare end-to-end robot policies with Layered Robot Architecture, Sim2Real, Structured 3D Robot Data, and Open-World Robot Manipulation. In E229, it uses 何倩然 / He Qianran, 杨璐 / Yang Lu, and 张从志 / Zhang Congzhi to examine China’s power-battery rise through Ten Cities Thousand Vehicles, Power Battery Whitelist, CATL / 宁德时代, BYD, ATL / Amperex Technology Limited, Northvolt, Battery Manufacturing Know-How, Power Battery Industry Chain, and next-generation routes such as Sodium-Ion Battery Storage. In E243, the show extends its sports-business branch from running shoes into football finance by using Premier League, Manchester United, Arsenal FC, Liverpool FC, Chelsea FC, and UEFA to examine American Sports Capital In European Football, Football Club Financial Engineering, and Football Commercialization Fan Conflict. In E241, the show moves into sports and consumer products by using 盖德, Nike, Adidas, Nike Vaporfly, and Adidas Adios Pro Evo 3 to examine Running Shoe Technology, Marathon Performance Ecosystem, and Performance Footwear Market. In E242, the show frames Recursive Self-Improvement as a current Silicon Valley model-company battleground and uses Apodex to examine how Deep Research, coding, AI Verification, and Research Taste might become a self-improving AI system. In E240, it uses Cresta, Invisible Technologies, Jove, and Oliver to examine why Forward Deployed Engineer work, Forward Deployed Product Manager specialization, AI Workflow Triage, and Private Equity AI Transformation are becoming central to enterprise AI adoption. In E225, it frames enterprise agents as a challenge to SaaS seats and organization design through Bairong Intelligence, Digital Employees, and Result As A Service.

E230｜1万亿收入预期背后：英伟达的巅峰与软肋 adds a semiconductor and AI-infrastructure business installment. The show uses 张璐 / Zhang Lu, 肖志斌 / Xiao Zhibin, Mark Ren, and Alex to examine Nvidia’s $1 trillion order narrative across Blackwell, Vera Rubin, inference demand, chip design, Advanced Packaging, High Bandwidth Memory, GPU Cloud Operations, Neo Cloud, and land and power.

E228｜谷歌TPU能撼动英伟达吗？前TPU工程师首次揭秘 adds a TPU-focused semiconductor installment. The show uses former Google TPU engineer Henry to explain TPU versus GPU, XLA, JAX, TPU Pods, Ironwood, High Bandwidth Memory, Advanced Packaging, Broadcom, Anthropic’s TPU fit, and why Nvidia pressure is real but conditional.

E231｜从B2B到A2A：Agent新基建，如何让“一人企业”做全球生意？ adds a B2B agent infrastructure installment. The show interviews 张阔 / Zhang Kuo on Alibaba, Axio, B2B to A2A, Agentic B2B Sourcing, and why small-business agents need trade data, sourcing workflows, verification, long memory, and model-responsive organization rather than only a general assistant.

E232｜餐饮出海有新招儿吗？从鼎泰丰与贡茶拿下美国初代销冠聊起 adds a restaurant and tea-drink globalization installment. The show uses Din Tai Fung / 鼎泰丰, Gong Cha / 共茶, Bibigo, Thailand, Haidilao / 海底捞, HeyTea / 喜茶, Moli Nai Bai / 茉莉奶白, Chagee / 霸王茶姬, Wallace / 华莱士, and Jollibee to examine Restaurant Cultural Legibility, U.S. Restaurant Real Estate Constraint, and Culture-Led Food Adoption in the United States.

E233｜硅谷右翼的权力网如何形成？聊聊彼得·蒂尔的思想启蒙拼图 adds a Silicon Valley politics and ideology installment. The show uses Peter Thiel, Stanford Review, Rene Girard, Carl Schmitt, Leo Strauss, Palantir, Founders Fund, Thiel Fellowship, Right-Wing Media Infrastructure, and JD Vance to examine Technology Right Power Network, Mimetic Competition Avoidance, Exception-State Surveillance Technology, Straussian Kingmaker Politics, and Political Talent Pipeline.

E234｜未来实拍电影还存在吗？与导演陆川聊聊AI给影视人的恐惧与自由 adds an AI film and voice installment. The show uses 陆川 / Lu Chuan, 黄英 / Huang Ying, KESA, StoryStorm, ByteDance, Midjourney, and 《南京！南京！》 / Nanjing! Nanjing! to examine AI Video Production Workflow, Industrial-Grade Film Models, Live-Action Film Under AI, AI Dubbing, and AI Voice Cloning Rights.

E235｜20年内CAR-T治愈癌症？与刘诚博士聊聊癌症治疗的底层哲学 adds a biotech and oncology installment. The show uses 刘诚, Eureka Therapeutics, Legend Biotech / 南京传奇, and Johnson & Johnson / 强生 to examine CAR-T Cell Therapy, Cancer Immune Recognition Problem, Ex Vivo CAR-T Manufacturing, In Vivo CAR-T, Allogeneic CAR-T, Solid Tumor CAR-T Constraints, Tumor Microenvironment, Cytokine Release Syndrome, and China Cell Therapy Regulatory Dual Track.

E236｜99%的作业都是AI写的：当代名校生眼里，大学还剩下什么？ adds an AI and higher-education installment. The show interviews Alfred 林童雨, Kelento 侯泰宇, and Jack 饶街五 to examine AI Default Learning Environment, Degree As Trust Credential, AI University Assessment Reform, AI As Tutor, AI Shortcut Risk, AI Use Pacing, and Human Judgment Under AI as university value shifts from routine knowledge output toward judgment, verification, trust, and problem choice.

Connections