Apple Intelligence Hits WWDC as U.S. Payrolls Gain 172,000

Apple Intelligence 亮相 WWDC，美非农新增就业 17.2 万 苹果在 WWDC 面临转向随机性 AI 集成的战略转折，而韧性十足的美国就业增长则削弱了关于 AI 即将导致劳动力市场崩溃的叙事预测。 主持人们深入剖析了苹果 WWDC 主题演讲，评估了 Siri 的大语言模型重构、操作系统性能升级与注重隐私的私有云计算（Private Cloud Compute），同时探讨了围绕智能手机过度使用展开的更广泛社会议题。他们还分析了积极的美国劳动力市场数据及市场波动，反驳了有关 AI 即将引发就业崩溃的论调。最后，本期节目探讨了 X 平台上引发热议的 VC 惊悚故事，讨论了分期出资交易结构、创始人路演博弈以及投资人的高压行为。 苹果 WWDC 以“All Systems Glow”为主题拉开序幕，市场预期集中在日常 iOS 体验中引入成熟的实用 AI 功能与重塑 Siri，而非推出激进的新硬件突破。 (0:02) 开发者对 Mac Mini 作为本地 AI 智能体服务器的需求持续走高；同时外界也对苹果 Private Cloud Compute 的基础设施构成究竟是自建数据中心还是 GCP 等白标云服务提出了疑问。 (4:18) 美国 5 月非农就业岗位增加 17.2 万个，失业率维持在 4.3%，展现出劳动力韧性并反驳了 AI 即将引发就业崩溃的悲观论调；但持续的通胀与能源价格压力仍可能促使美联储转向加息。 (14:47) Greg Eisenberg 发起的社交媒体话题引爆了风险投资圈黑料，创始人公开分享投资人路演打瞌睡及红杉资本两阶段分期注资结构（Tranche Investing）等争议，引发对真实估值与公关宣传的广泛讨论。 (19:28)

The Diet TBPN hosts argue that grounded practical realities are puncturing tech industry hype, seen in Apple prioritizing daily software utility over speculative breakthroughs and robust employment figures defying immediate AI-driven labor disruption. Apple Integrates Practical AI While Cloud Infrastructure Questions Linger (0:02) At WWDC, Apple oriented its "All Systems Glow" keynote around pragmatic software updates rather than major hardware reveals, aiming to integrate proven conversational AI features into core OS workflows and revive Siri via deep LLM integration. The hosts note this requires a major cultural shift for Apple's PR department, which must adapt from tightly controlled, deterministic software outputs to managing stochastic models prone to viral hallucinations. Simultaneously, questions surround Apple's "Private Cloud Compute." While Apple emphasizes high security for off-device processing, the hosts question whether inference for over one billion users relies on proprietary data centers, clustered Mac Minis, or white-labeled Google Cloud Platform instances using custom Gemini fine-tunes. Reported rate limits and potential subscription tiers indicate that local on-device processing cannot handle the full workload, which will eventually draw attention to Apple's energy consumption and corporate ESG disclosures. Utility Outweighs Hype as Smartphone Saturation Faces Scrutiny (7:06) The hosts highlight societal pushback against mobile saturation, citing research highlighted by journalist Derek Thompson attributing up to 30% of the recent drop in fertility rates below replacement levels to smartphone overuse. They note device makers remain conspicuously quiet about these social consequences, referencing the Prime Minister of Denmark's remark that she would rather let her children smoke cigarettes than use an iPhone. From a product perspective, everyday consumers place greater value on core performance gains—such as a 30% faster lock screen unlock or quicker app launch times—than on abstract AI features. Referencing reporter Mark Gurman, the hosts emphasize that reliability in battery life, display quality, and speed drives adoption, warning that competing purely on spec metrics like FLOPS risks commoditizing hardware and eroding premium margins. Resilient Job Data Defies AI Labor Disruption Predictions (14:47) Despite persistent predictions of catastrophic labor displacement from artificial intelligence, official economic data continues to show employment resilience. The U.S. Labor Department reported a seasonally adjusted 172,000 nonfarm payroll jobs added in May, keeping the unemployment rate flat at 4.3% across three consecutive months of expansion, supported by hiring in travel, tourism, and World Cup preparations. This labor stability unfolded alongside severe financial volatility, including a 4.2% single-day plunge in the Nasdaq that recovered 1.5% the following Monday. With inflation remaining above the Federal Reserve's 2% target and energy disruptions near the Strait of Hormuz driving up oil prices, the hosts suggest the Fed may contemplate interest rate hikes rather than cuts, presenting persistent valuation headwinds for high-growth tech firms. Founder Grievances Reveal Structural Mechanics in Venture Fundraising (19:28) A viral social media thread started by Greg Eisenberg prompted founders to air pitch and boardroom grievances, detailing experiences ranging from sleeping investors to aggressive negotiation tactics. The hosts distinguish between early-stage pitch discourtesy, which can be dismissed across dozens of meetings, and difficult board members who pose genuine operational and governance hazards to a company. The conversation also addressed Mercor CEO Brendan Foody's criticism of Sequoia's two-tranche funding rounds, which Foody initially termed a scam before acknowledging as standard venture practice. In these setups, split valuation tiers—such as investing across $500 million and $1 billion caps—yield an intermediate blended share price while letting founders announce the higher headline valuation. The hosts argue that misleading public impressions stem from founders chasing PR milestones rather than the established tranche mechanics used by venture firms. Across consumer operating systems, macroeconomic forecasts, and startup deal mechanics, the hosts observe that verifiable technical execution, practical utility, and standard financial rigor are overtaking speculative promotional narratives.

本期 TBPN 节目探讨了苹果 WWDC 围绕实用人工智能与系统基础优化的转向，同时对比了美国强劲就业数据与科技股震荡下的宏观现实，并剖析了近期网络热议的风投怪象与交易结构争议。 WWDC预期转向与系统级AI集成 (0:02) 本届 WWDC 以“All Systems Glow”为主题拉开帷幕，苹果首席执行官 Tim Cook 强调了系统级变革。与此前外界追捧 Apple Vision Pro 等硬件突破不同，当前用户的期望更为务实，仅希望苹果将 ChatGPT、Gemini、Claude、Grok 等已被市场验证的 AI 功能深度融入日常操作系统流程中。节目主持团队指出，苹果早先通过广告牌力推“Apple Intelligence”拉高了外界预期，如今转向更为务实的策略，重点放在盘活已被边缘化多年的 Siri 按钮上。然而，将非确定性的大语言模型引入核心系统，意味着苹果一贯注重精确输出的公关与品牌文化将面临 AI 幻觉和内容不可控的严峻考验。 私有云算力疑云与生态权限博弈 (4:18) 围绕苹果公布的“私有云计算”（Private Cloud Compute），节目深入讨论了其在十亿级活跃 iPhone 用户规模下的实际落地可行性。分析人士质疑苹果究竟是在自建高度保密的数据中心、将 Mac Mini 组网运算，还是借用了贴牌的 Google Cloud Platform 实例；同时，线上用户提到的调用频率限制与潜在订阅模式，暗示终端设备无法完全独立承载推理任务，庞大的能耗也势必会在未来的企业 ESG 披露中显现。此外，生态开放性也是核心焦点：一方面，开发者正大量将 Mac Mini 用作 OpenClaw 等本地 AI 代理的服务器；另一方面，ChatGPT 或 Claude 等第三方应用若想深入调用 iMessage 或完整相册，仍受制于苹果在隐私保护与便捷授权之间的严格限制。 基础体验价值与手机成瘾争议 (7:06) 针对设备体验，节目援引彭博社记者 Mark Gurman 的观点指出，普通消费者最为在意的往往是电池续航、屏幕素质、相机成效以及日常响应速度，而非空泛的 AI 承诺；锁屏解锁提速 30% 这类基础优化，在日常使用中往往比前沿功能更具实际价值。如果科技产品放弃品牌情感溢价，单纯陷入参数与算力的数字内卷，将不可避免地面临硬件商品化和利润受挫的风险。与此同时，智能手机对社会的深层冲击持续引发讨论。记者 Derek Thompson 引用的研究表明，智能手机过度饱和与社交媒体沉迷可能导致全球生育率下降中多达 30% 的低于替代率缺口，甚至有欧洲政要直言宁愿让孩子吸烟也不愿其沉迷 iPhone，这与硬件厂商对此类社会议题的长期回避形成了鲜明反差。 宏观就业韧性与科技估值分歧 (14:47) 在宏观经济层面，美国劳工部数据显示 5 月非农就业人口经季节调整后增加 17.2 万人，失业率维持在 4.3%，实现了连续第三个月的就业扩张。这一数据直接反驳了所谓“AI 导致劳动力市场迅速崩溃”的极端悲观论调，旅游业及大型赛事筹备有力拉动了招工需求。然而资本市场呈现出明显分化：尽管纳斯达克指数在经历单日下跌 4.2% 的一年来最大跌幅后迅速反弹 1.5%，但霍尔木兹海峡受阻引发的油价波动推升了通胀担忧，使美联储降息前景受阻甚至可能面临加息压力，进而对依赖远期现金流折现的高成长科技企业估值形成持续压制。 风投乱象争议与分批投资机制 (19:28) 社交网络上由创始人发起的风投吐槽引发广泛关注，曝光了诸如合伙人在路演中入睡、要求开除团队成员作为出资前提、或在副驾驶座用肚子顶着电脑审阅 PPT 等失范行为。节目分析认为，路演阶段的不专业行为虽然令人生厌，但董事会层面的恶意干预才是初创企业的真正生存威胁。针对 Mercor 首席执行官 Brendan Foody 指控红杉资本通过“两批次投资”（如先后以 5 亿美元和 10 亿美元估值注资）制造“混合估值”涉嫌误导的言论，主持团队解释指出，分批投资早在红杉投资 YouTube 时期就已是行业惯例，核心责任在于创始人是否在对外宣传中故意隐瞒混合价格以夸大估值，而非风投机制本身的问题。 本期节目梳理了科技产业在技术落地与资本运作中的多重现实差距：从苹果在用户体验与云端算力瓶颈间的平衡，到宏观紧缩预期对高估值科技泡沫的挤压，再到创投圈潜规则的公开展露。在基础算力成本、真实用户体验与商业诚信的审视下，行业正被迫从激进的叙事炒作转向务实的价值落地。

## Background 本期内容由科技商业访谈节目 TBPN 于 2026 年 6 月 23 日发布在 YouTube 平台。该视频属于《Diet TBPN》精炼系列，旨在以 30 分钟的高效节奏呈现当日两小时直播的精华探讨，涵盖 WWDC 开幕与苹果最新动作、宏观就业与科技股走势，以及近期在社交网络刷屏的硅谷风投奇葩内幕故事。 ## TL;DR 节目围绕苹果 WWDC 开幕展开，深入分析了 Apple Intelligence 走向实用主义的生态整合、Siri 改版以及对“私有云”算力与隐私的现实考量，指出苹果正在告别确定性时代的公关包袱并回归续航、相机等基础体验。宏观层面，主播对照 5 月强韧的非农就业数据与纳指剧烈波动，反击了部分 AI 领袖鼓吹的大规模失业末日论，并探讨了后零利率时代的投机现象。此外，节目重点拆解了社交媒体上热议的风投奇闻，从合伙人 Pitch 打瞌睡、软银的心理施压战术，到红杉分批投资引发的混合估值争议，剖析了早期融资的残酷真相与行业博弈规则。 ## Atmosphere & style 整场节目充盈着硅谷圈内人特有的松弛感与辛辣幽默，主播与导播之间的接梗抛梗行云流水，在极客深度与街头调侃间切换自如。从嘲讽苹果发布会的“便当盒参数”到重现孙正义“10分钟时限”和车内路演的荒诞名场面，现场爆笑不断且戏剧张力拉满。听众能真切感受到一种身处创投一线茶水间的快节奏“脱口秀”氛围，既有对宏观经济与大模型底层的硬核解构，又充满对创投圈光怪陆离生态的犀利吐槽。 ## Bulletin - [00:02](https://www.youtube.com/watch?v=fGnb656Syn8&t=2s) 苹果 WWDC 正式发布 Apple Intelligence，产品策略从追求划时代 AGI 转向务实整合 ChatGPT 与 Gemini 等成熟体验，依托 Siri 按键完成轻量化系统嵌入。 - [12:00](https://www.youtube.com/watch?v=fGnb656Syn8&t=720s) 针对苹果宣称的私有云架构，面对全球超 10 亿活跃设备生态，模型推理面临巨大的云端算力消耗、速率限制与 ESG 碳排放压力。 - [15:20](https://www.youtube.com/watch?v=fGnb656Syn8&t=920s) 美国 5 月新增非农就业 17.2 万人且失业率维持在 4.3%，连续三个月稳步增长，有力反驳了 AI 导致人类大规模失业的极端悲观论调。 - [24:29](https://www.youtube.com/watch?v=fGnb656Syn8&t=1469s) 社交媒体掀起风投吐槽热潮，针对红杉被指控‘两期分批注资’压低实际混合估值的做法，分析指出该交易结构本身合法且历史悠久，核心法律风险仅在于创始人是否在后续融资中隐瞒条款进行虚假陈述。 ## Key takeaways - [00:02](https://www.youtube.com/watch?v=fGnb656Syn8&t=2s) 苹果 AI 落地策略回归务实：消费端更看重成熟大模型在系统内的日常嵌入与轻便交互，而非遥不可及的 AGI 宏大概念。 - [02:00](https://www.youtube.com/watch?v=fGnb656Syn8&t=120s) 非确定性输出考验企业文化：苹果习惯于交付高度确定性的系统功能，大语言模型不可避免的偶发翻车与幻觉将持续考验其公关团队的承受力。 - [04:20](https://www.youtube.com/watch?v=fGnb656Syn8&t=260s) 开发者生态张力浮现：Mac mini 引发本地算力与 Agent 部署热潮，但第三方 AI 应用能否深度触碰系统隐私数据仍面临严格的权限博弈。 - [06:57](https://www.youtube.com/watch?v=fGnb656Syn8&t=417s) 智能手机屏幕沉迷与社会学争议：围绕数码依赖可能压低生育率的讨论升温，消费电子巨头未来迟早需要在公众层面正视‘认知退化’的社会学问责。 - [08:54](https://www.youtube.com/watch?v=fGnb656Syn8&t=534s) 硬件参数比拼与利润挤压：当发布会陷入芯片跑分和规格网格图的比拼时，产品往往逐步沦为大宗商品属性，面临品牌溢价削弱与利润率压缩风险。 - [10:48](https://www.youtube.com/watch?v=fGnb656Syn8&t=648s) 隐私防线升级持续冲击移动广告：苹果在系统层面对用户数据和追踪接口的进一步收紧，直接增加了广告技术生态的变现难度。 - [12:00](https://www.youtube.com/watch?v=fGnb656Syn8&t=720s) 十亿级设备的推理算力谜团：苹果宣称的‘私有云’本质上仍需庞大数据中心支持，全端侧运行的不现实直接反映在系统速率限制与高昂能耗上。 - [15:20](https://www.youtube.com/watch?v=fGnb656Syn8&t=920s) 实体经济韧性证伪就业末日论：强劲的非农招聘与稳定的失业率证明，实体劳动力需求依然旺盛，AI 短期内迅速吞噬海量岗位的说法不攻自破。 - [17:51](https://www.youtube.com/watch?v=fGnb656Syn8&t=1071s) 高利率环境下的投机狂欢：通胀顽固使降息预期降温，但即便在后 ZIRP 时代，市场的投机狂热依然活跃，同时赋予了美联储在未来危机时降息护航的政策空间。 - [19:28](https://www.youtube.com/watch?v=fGnb656Syn8&t=1168s) 辨析 Pitch 糟糕体验与董事会实质控制：风投合伙人打瞌睡或不读材料仅属于双向选择时的体验瑕疵，真正具有毁灭性风险的是进入董事会后强行干涉治理与人事换血。 - [19:28](https://www.youtube.com/watch?v=fGnb656Syn8&t=1168s) 创投博弈遵循长期重复博弈规则：硅谷正和博弈生态决定了投资人通常追求体面收场，极少愿意在创业失败时彻底与创始人撕破脸。 - [24:29](https://www.youtube.com/watch?v=fGnb656Syn8&t=1469s) 分批注资财技的边界与责任：分期注资与混合折价是全行业通行了 25 年的成熟结构，若引发欺诈争议，主要责任在于创始人对外宣传时是否隐瞒混合估值真相。 - [27:07](https://www.youtube.com/watch?v=fGnb656Syn8&t=1627s) 路演生理机制与 30 分钟危险期：实验表明老年男性在低刺激环境中入睡中位时间为 36.9 分钟，创始人面对昏暗会议室与舒适人体工学椅时应严格控制宣讲时长直奔核心。 - [28:42](https://www.youtube.com/watch?v=fGnb656Syn8&t=1722s) 软银极端路演名场面揭示的商业生存逻辑：无论面对抽烟咳嗽的心理威压还是 10 分钟极限时限，创始人必须具备在极端高压下迅速直击商业核心的抗压能力。 ## Detailed walkthrough ### [00:02](https://www.youtube.com/watch?v=fGnb656Syn8&t=2s) WWDC 正式开幕：软件转型的合理预期与现实落地 历经近两年的铺垫与发酵，苹果的“Apple Intelligence”终于迎来实质性交付。一年一度的苹果全球开发者大会（WWDC）于周一正式开幕，并将持续至周五，首席执行官 Tim Cook 刚刚完成了揭幕主题演讲。今年大会的主题定为“All systems glow”（系统泛光 / 全面焕发），既呼应了 Siri 呼出时环绕屏幕的光晕设计，也带有狂欢派对般的意味。作为一场纯粹聚焦于系统软件（iOS、macOS 等）而非 iPhone 硬件的大会，外界此前对这套全新软件系统的预期处于一个非常奇妙的状态：一方面，大众确实期待苹果能带来一场真正的软件蜕变，交出一份让人津津乐道的高分答卷；但另一方面，公众的预期并没有像当初面对 Apple Vision Pro 那样膨胀到苛求前所未见的划时代突破。 人们目前对苹果 AI 的真实诉求其实非常朴素：只需将 ChatGPT、Gemini 和 Claude 身上那些已经验证成熟的最佳实践融合进系统即可。正如社交平台 X 内置的 Grok 一样，用户刷到感兴趣的推文时随手问一句“Hey Grok, 这是真的吗？”，就能免去手动复制粘贴到独立大模型 App 的繁琐；又如 Google 搜索上线的 AI Overviews，或是财务软件 Ramp 内置的智能对话框——用户直接输入“我们上个月到底在亚马逊花了多少钱”，系统便能精准调出账单。这种轻量化的日常嵌入并不是什么革命性的通用人工智能（AGI），却切实改善了交互体验。 ```mermaid flowchart TD accTitle: 苹果 AI 落地与生态整合路径 accDescr: 展示苹果从以往激进概念走向务实整合的系统落地架构与面临的权衡 subgraph UserLayer["用户端交互预期"] direction LR U1["一键唤醒 Siri"] U2["日常功能无缝嵌入"] U3["轻量化问答与摘要"] end subgraph AppleEcosystem["系统整合核心"] direction TB A1["Apple Intelligence 框架"] A2["Siri App Intents 架构"] A3["本地与云端模型路由"] end subgraph ExternalAndRisk["外部协同与现实挑战"] direction TB E1["第三方大模型接入<br>(ChatGPT / Claude 等)"] R1["非确定性输出与幻觉"] R2["公关阵痛与隐私权限博弈"] end UserLayer --> AppleEcosystem AppleEcosystem --> ExternalAndRisk ``` --- ### [02:00](https://www.youtube.com/watch?v=fGnb656Syn8&t=120s) 务实的产品策略：告别确定性时代的公关阵痛与幻觉挑战 苹果此次选择让市场期待自然发酵，被证明是一次明智的公关降温。回顾一两年前，外界铺天盖地打出各种关于 Apple Intelligence 与底层技术的广告，甚至内部也在自我加戏造势（让人忍不住自嘲“我们是不是要被淘汰了？”），这种过热预期往往是在给自己“挖坑”。如今，苹果找到了正确的合作伙伴与产品演进路径：既然大众已经熟稔各类大语言模型工具，只要将它们接入合理的操作入口即可，而这个入口首选便是过去两年多被严重削弱（nerfed）的 Siri 按键。纵观历史，Siri 始终未能享受到苹果其他明星产品那样的拥趸与热度，如今只要能让用户在熟悉的界面里一键调出可靠的模型能力，便足以赢得用户的认可。 借鉴 Google AI Overviews 的落地历程可以看出，推广大语言模型虽非造火箭般不可逾越，却极其考验系统韧性。即便是顶尖的 DeepMind 团队打造出的前沿模型，在处理日常提示词时依然可能闹出笑话——比如用户输入“disregard”（忽略），模型不会给出词义解释，反而自我代入地回复“好的，我再也不这样做了”。这种病毒式的翻车瞬间在苹果系统上线后同样无可避免（例如生成出让人啼笑皆非的离谱短信摘要），这必然会让习惯于交付“确定性产物”的苹果公关团队陷入持续的心惊肉跳。不过，这种由随机性（stochastic）算法引发的非确定性笑话并不会真正重创用户留存或产品使用率；相反，只要遵循整个 AI 领域摸索了一两年的最佳实践，苹果便能稳步渡过这一文化与产品转型的深水区。 - 大模型在消费端更看重场景便利性，而非遥不可及的 AGI 宏大叙事。 - 苹果长久以来的软件基调是“确定性输出”，非确定性大模型的偶发幻觉需要其企业文化和公关部门做出重大心理调适。 - 诙谐的文本摘要失误虽易在社交媒体引发调侃，但因其自带幽默感与基础实用性，很少造成用户流失。 --- ### [04:20](https://www.youtube.com/watch?v=fGnb656Syn8&t=260s) 开发者生态博弈：本地算力狂潮、Vibe Coding 与隐私权限深水区 在基础模型能力之外，业界密切注视着苹果在开放生态方面的站位。首先是硬件领域的意外红利：凭借优异的能效比与统一内存架构，Mac mini 近期在本地 AI 部署及开源智能体（如各种本地 Agent 框架）社区中掀起购买狂潮，甚至卖到断货。苹果是在接下来的系统更新中积极拥抱这股开源潮流、为开发者工具释放更多便利，还是会借保护隐私之名筑起高墙收紧限制，构成了目前的一大战略张力。 其次，针对近期在开发者社区大火的“Vibe Coding”（氛围编码 / 由 AI 辅助快速生成代码）模式，App Store 未来将采取何种审核姿态同样备受关注。面对这类趋势，苹果往往不会在第一时间公开发表看法，而是习惯于暗中解决机制问题，等到商业闭环完全跑通后再大张旗鼓地宣传——就像当年对待气候变化与碳中和目标一样，先在园区铺设光伏板、改造供应链，待到实现净零排放目标后才全面对外发声。 最后，第三方 AI 应用（如 ChatGPT、Claude、Gemini）在 iOS 系统内部的权限深度成为最大看点： - **数据调用链路**：第三方模型是否能够触及用户的 iMessage 短信流水？若用户同意授权，系统究竟是一次性永久放行，还是每次调取都弹出二次确认弹窗？弹窗过频将直接破坏交互体验。 - **权限授予心理**：如同目前许多社交软件动辄索要相册全量持久访问权限一样，大量用户往往出于信任或图省事而选择直接全盘放行；AI 时代的权限索取势必引发更深层次的博弈。 - **系统架构依托**：第三方应用的功能渗透是主要依赖现有的 Siri App Intents 架构，还是能够接入更底层的 iOS 核心 API。 --- ### [06:57](https://www.youtube.com/watch?v=fGnb656Syn8&t=417s) 延伸思考：智能手机的“数码毒害”争议与人口趋势大讨论 苹果对敏感议题的滞后回应风格，让人联想到消费电子行业未来可能面临的一项潜在社会学问责：智能手机与屏幕依赖是否正在对全球生育率产生下行压力。知名分析撰稿人 Derek Thompson 近期援引相关研究指出，自己此前对此持怀疑态度，但最新数据让他开始相信，智能手机的泛滥与社交孤立可能为近期多国生育率跌破世代更替水平（低于 2.0）贡献了高达 30% 的原因。这种关乎人类族群延续的宏大命题，迟早会迫使各大科技硬件与社交媒体巨头在开放式访谈中作出正面回应，而不能仅仅由那些热衷于畅聊此类哲思的 AI 公司创始人独自高谈阔论。 欧洲政界对此已有激化言论，主播提到有欧洲领导人甚至直言“宁可让自己的孩子抽香烟，也不愿让他们过早使用 iPhone”。对此主播展开了理性驳斥：抽烟显然是直接吸入物理毒素，一两口便有实质危害；而多看一眼屏幕不可能直接造成生理上的绝育，二者性质完全不同。但这也折射出大众在数码时代对“认知退化”（brain rot）的深层焦虑——攻克癌症的医学方案或许已在转角处，但治理数码沉迷的心智解药依然悬而未决。在节目准备转入更多 WWDC 现场连线与各界即时反馈前，主播还插播了赞助商 Console 的资讯（Console 致力于通过企业级 AI Agent 实现 IT、HR 和财务支持流程中 70% 的自动化，秒级解决员工的系统权限申请与密码重置需求）。 ### [08:54](https://www.youtube.com/watch?v=fGnb656Syn8&t=534s) 基础体验与规格参数：苹果的“便当盒”哲学 节目开场由一段日常吐槽切入：原本设置好夜间自动更新的 iPhone，到了第二天早晨却弹出一句仿佛在说“兄弟，我昨晚没搞定，感觉不对，气场不合（Vibe was off）”的提示。主持人指出这种尴尬几乎人人都经历过，通常与电池充电量未达标有关，而这些显而易见的技术瑕疵（Low-hanging fruit）理应在 WWDC 上得到彻底解决。虽然发布会上抛出了一连串微小性能提升的数字——比如锁屏启动速度提升 30%、特定应用开启加快 30%——但在他们与彭博社记者马克·古尔曼（Mark Gurman）交流时，古尔曼明确表示，对大众用户而言，AI 功能往往显得过于抽象，大家真正渴望的其实是扎扎实实的基础体验：更持久的电池续航、出色的相机、细腻明亮的屏幕以及系统运行的绝对流畅。眼下正是苹果回归现实、埋头干脏活累活（Time to chop wood）的时刻。 随后两人调侃起苹果发布会上的经典视觉元素——“便当盒（Bento box）”网格图。这一展示各项芯片核心数、GPU 算力、浮点运算（Flops）、相机像素与存储容量的信息图，本质上就是苹果独创的“模型卡（Model card）”。他们进一步探讨了商业逻辑的残酷演变：当一款产品不再凭借纯粹的品牌溢价来驱动销售，而是开始陷入逐项比拼性能与规格的境地时，往往意味着利润率面临压缩（Margin compression）。这就像汽车行业中的对比，如果是为了续航、零百加速、马力、价格和座椅布局去精挑细选，产品就已经沦为大宗商品属性；而人们购买法拉利（Ferrari）只因为它是法拉利，根本无需细问参数。相比之下，卢西德（Lucid）等新兴电动车不得不面对“花这个价钱买 2.5 秒零百加速其实不算划算”的参数审判，这显然失去了超豪华品牌的特权地位。 - 用户对更新失败的共鸣多源于充电阈值与电量要求等基础体验瑕疵。 - 强调用户更关注电池寿命、相机和屏幕等基本盘，而非抽象的 AI 概念。 - 发布会网格图被戏称为硬件版“模型卡”，标志着从品牌溢价向参数内卷与利润挤压的过渡。 --- ### [10:48](https://www.youtube.com/watch?v=fGnb656Syn8&t=648s) 恶搞公关稿与 WWDC：苹果的“隐私防线”与广告之争 话题转向社交平台上的知名科技恶搞博主萨姆·亨利·戈尔德（Sam Henri Gold）。戈尔德仿照苹果官方新闻室（Apple Newsroom）的口吻炮制了一篇虚构新闻稿，宣称“苹果宣布马克·古尔曼‘死亡’”。恶搞声明一本正经地写道：苹果宣布代号为“One More Thing”的多年期行动已圆满结束，该行动旨在彻底终结古尔曼对苹果未发布新品信息的持续泄露。古尔曼连续 16 年准确提前爆料苹果核心机密，如今终于被“中和”，该任务“按期、在预算内且无任何波折地完成”。虽然现实中苹果不至于如此极端，但常年精准剧透的古尔曼确实一直是苹果管理层的眼中钉。与此同时，正主古尔曼正在彭博官网上如火如荼地进行 WWDC 实时图文直播。 移动广告与营销专家埃里克·索弗特（Eric Seufert）同样高度关注本届发布会。古尔曼此前预告，除了聚焦 AI、Siri 以及各大操作系统的底层性能质量飞跃之外，隐私（Privacy）与安全（Safety）功能将是当天的另外两大核心。对于靠广告技术变现的从业者而言，这无异于警钟再次敲响——苹果进一步收紧隐私政策、屏蔽底层用户数据，向来都会直接重创移动广告追踪与商业化变现的精准度。 - 恶搞新闻稿讽刺了苹果保密部门与古尔曼长达 16 年的猫鼠游戏。 - 古尔曼预测 WWDC 的核心聚焦于 AI/Siri、系统性能、隐私保护与安全功能。 - 隐私功能升级通常意味着进一步屏蔽数据接口，对移动广告变现形成直接冲击。 --- ### [12:00](https://www.youtube.com/watch?v=fGnb656Syn8&t=720s) 私有云还是公有云？十亿用户的推理算力谜团 随着 WWDC 直播的推进，导播泰勒（Tyler）与主持人们展开了关于苹果 AI 架构的深度剖析。苹果在展示各项 AI 功能时，反复强调这些操作运行在“私有云（Private Cloud）”之上，力证其极高的安全性与隔离性。然而，这一说法引发了现场的一连串质疑：所谓“私有云”究竟意味着什么？是苹果完全自研的基础大模型，还是依托与谷歌 Gemini 协议实现的白标（White-label）与微调（Fine-tune）版本？虽然苹果屏幕上确实闪现了 Gemini 的字样，让苹果能利用自身品牌包装前沿大模型，但这背后牵扯出了巨大的算力疑问。 全球拥有超过 10 亿 iPhone 活跃用户，如果每个人整天都在频繁呼出 Siri 提出各种复杂需求，其所产生的模型推理（Inference）体量将是天文数字。如果算力来自云端，苹果难道秘密建造了足以匹敌前沿规模的巨型数据中心？抑或是把谷歌云（GCP）的一角包装成了所谓的“私有云”？还是说机房里只是用成千上万台 Mac mini 拼接而成的网络？即便苹果在资本支出（CapEx）财报和 SEC 申报中隐藏了机房硬件投入，如此庞大的能耗也必然会在碳排放和 ESG 报告中露出马脚。观众弹幕提及苹果设置了速率限制（Rate limits）甚至可能推出订阅制方案，这从侧面反推了全端侧运算的不现实：端侧运行本不存在速率限制与云端成本，苹果宣称的“私有云”本质上仍是脱离设备的远程数据中心，即便苹果早已拥有存储海量 iCloud 照片与办公数据的服务器底座，面对生成式 AI 推理的能耗黑洞，其云基础设施的可持续性依然存疑。 ```mermaid flowchart TD accTitle: Apple AI Inference Query Routing accDescr: High-level architectural flowchart of an iPhone AI query splitting between on-device processing and private cloud inference. subgraph Client["客户端 (iPhone / 10亿+ 用户)"] UserReq["用户触发 Siri / 复杂 AI 任务"] DeviceCheck{"本地芯片是否具备算力？"} end subgraph Edge["本地推理 (On-Device)"] LocalModel["端侧小模型"] LocalResp["极速私密响应 (无速率限制)"] end subgraph CloudInfra["苹果私有云 (Private Cloud)"] DataCenter["专用数据中心 / GCP 隔离集群"] GeminiBase["Gemini 底座微调 / 白标大模型"] CloudCost["算力瓶颈: 速率限制、订阅收费、ESG 能耗"] end UserReq --> DeviceCheck DeviceCheck -- "轻量基础任务" --> LocalModel --> LocalResp DeviceCheck -- "复杂推理需求" --> DataCenter DataCenter --> GeminiBase --> CloudCost ``` - 面对 10 亿级设备生态，调用前沿大模型所产生的推理压力难以单纯依托端侧硬件消化。 - 速率限制与潜在订阅制的迹象印证了重度依赖云端基础设施的现实。 - 巨大的计算能耗使外界对苹果能否在保持 ESG 碳中和指标的同时支撑 AI 算力产生怀疑。 --- ### [14:40](https://www.youtube.com/watch?v=fGnb656Syn8&t=880s) 液体玻璃退场与设计争议 在插入网络安全赞助商 CrowdStrike 的口播后，主持人补充了关于苹果系统 UI 调整的最新动态。设计评论人简（Jane）指出，苹果似乎正在其主打的“液体玻璃（Liquid Glass）”拟物拟真设计上做出妥协，为了提升实际界面的可用性（Usability）而选择主动后撤，并放出了对比截图作为佐证。 主持人则给出了不同的第一感受：新设计的 Apple Maps 图标依然保留了微缩的液体玻璃光泽质感，观感十分惊艳；不过社区论坛上关于全新 macOS 界面色彩的抱怨声不绝于耳，不少测试用户抗议新系统的对比度失调，导致界面要么刺眼过亮，要么暗淡不清。对于苹果这样以工业设计和人机工程学见长的巨头，在设计新语言与可读性实用体验之间的拉锯从未停止。 --- ### [15:20](https://www.youtube.com/watch?v=fGnb656Syn8&t=920s) 强劲非农对决纳指暴跌：AI 失业末日论被推翻？ 视角随后切换至上周五剧烈震荡的宏观金融市场。美国劳工部公布的最新非农就业报告显示，5 月份经季节调整后的新增非农就业人口达 17.2 万人，失业率稳稳停留在 4.3% 的水平。《华尔街日报》以“美国招聘势头正劲，就业连续三个月增长”作为头条报道。虽然增幅较上月略有放缓，但连续三个月实现数十万规模的稳步净增，证明劳动力市场韧性十足。与此同时，金融市场却遭遇大幅清算，上周五纳斯达克指数单日重挫 4.2%，创下一年多以来的单日最大跌幅，引来市场对“泡沫破裂”的哀嚎。不过到了周一，纳指迅速反弹 1.5%。主持人调侃道，当前市场并非重演 1999 年狂欢或 2000 年泡沫顶峰，而是进入了类似“2003 年”的泡沫出清后稳步重构期。 面对强韧的就业数据，主持人痛快反击了鼓吹“AI 毁灭就业”的末日论者（Black-pilled AI leaders）。部分 AI 行业高管不断煽动恐慌，扬言 AI 将迅速吞噬 10%、20%、50% 甚至 100% 的人类岗位。然而现实是，即便伴随着世界杯旅游等相关服务业的季节性繁荣，强悍的美国劳工群体与实体经济依然把失业率牢牢钉在低位，“AI 引发的工作岗位世界末日至少在 5 月份宣告破产”。尽管通胀隐忧依然挥之不去——特别是霍尔木兹海峡（Strait of Hormuz）的封锁引发了国际油价与汽油成本飙升，使得降通胀进程面临美联储所不乐见的物价压力——但无论如何，劳动力供需两旺所折射出的宏观实体经济，远比悲观主义者的预测来得健康。 - 美国劳工部公布 5 月经季调后新增非农就业 17.2 万人，失业率维持在 4.3%，招聘实现连续三个月走高。 - 纳斯达克经历单日重挫 4.2% 的一年来最惨下跌后，周一迅速回血反弹 1.5%。 - 面对 AI 领袖关于大规模失业的悲观断言，实体经济就业增长依然展现出极强韧性。 - 地缘政治导致霍尔木兹海峡局势紧张，油价上扬令通胀再度承压，美联储利率政策前路依然复杂。 ### [17:51](https://www.youtube.com/watch?v=fGnb656Syn8&t=1071s) 利率前景、抗通胀与后 ZIRP 时代的投机狂潮 通胀率依然显著高于美联储 2% 的目标水平，而且这种状态已经持续了相当长时间。这种情况让市场对降息的预期变得愈发渺茫，甚至有重返加息区间的可能。对于盈利预测已经排到未来十年的科技公司而言，高利率显然不是什么好消息。不过，如果一定要找些心理慰藉（"copium"），高利率唯一的积极面在于：一旦经济真的出现下行拐点，美联储的工具箱里至少还有子弹可用。当前经济处于过热运转且利率处于高位，如果未来股市暴跌或失业率飙升，美联储尚有充裕空间将利率一路下调至 3%、2%、1% 乃至 0；而在新冠疫情初期，利率原本就已经触底，突发大规模失业时只能依靠政府疯狂撒钱刺激、给全美居民邮寄支票，当时美联储在货币政策端几乎无计可施。 更让人感慨的是，在当前的利率高度下，市场中的投机氛围居然还能如此狂热。那些在零利率时代（ZIRP era）成长起来的人，曾以为当基准利率骤升 3% 宣告 ZIRP 终结后，市场将彻底告别非理性的泡沫与狂欢，科技股再也不可能重现 2000 年互联网泡沫的盛况。当时甚至有人做出了“求求上帝，再给我们一个泡沫吧”的汽车保险杠贴纸。如今看来，上帝确实回应了这一祈求，市场自身的狂热投机甚至反过来成为了进一步加息的绝佳论据。 ### [19:28](https://www.youtube.com/watch?v=fGnb656Syn8&t=1168s) 创投黑料热潮：Pitch 吐槽与董事会博弈的本质差异 Greg Isenberg 在社交平台上捅了马蜂窝，引发全网创始人集体翻旧账，曝光自己遭遇的风投（VC）奇葩经历，不少顶级投资机构和知名投资人无辜“躺枪”。Greg Isenberg 自曝曾在一家排名前三的顶级风投机构办公室 Pitch 一笔 1,500 万美元的 A 轮融资，现场坐了 12 个人，结果其中一位普通合伙人（GP）竟当场睡着，整整昏睡了 30 多分钟，全程无人敢提，大家只能硬着头皮继续讲下去。与此同时，Cloudflare 首席执行官 Matthew Prince 也爆料称，维诺德·科斯拉（Vinod Khosla）在考量是否领投其 C 轮融资时提出条件：要求 Matthew 必须解雇几名刚借故去洗手间的核心参会团队成员。 ```mermaid flowchart TD accTitle: VC Horror Stories Spectrum accDescr: 分类对比 Pitch 阶段的失礼行为与进入董事会后的实质控制风险 subgraph PitchStage["Pitch 融资阶段（低实质危害）"] P1["合伙人会议中睡着"] P2["不读 BP / 态度粗鲁傲慢"] P3["创始人可自由 Pass 投资人"] end subgraph BoardStage["Board 董事会阶段（高实质风险）"] B1["强行插手团队人事换血"] B2["治理结构与投票权争夺"] B3["法律与控制权层面的深度套牢"] end PitchStage -.->|"仅属于体验糟糕"| Decision{"融资双向选择"} BoardStage ==>|"真正具有毁灭性"| Disaster["创始人长期噩梦"] ``` 针对风投合伙人打瞌睡的指责，主播们打趣道：如果是创始人在办公室累瘫睡着，大家会称赞他是拼搏的英雄，比如埃隆·马斯克睡在工厂车间毫无违和感，但 VC 睡着就会被痛批不敬业。不过客观而言，硅谷长期处于高增长和正和博弈的生态中，相对体面的行为举止通常才是纳什均衡。哪怕一家创业公司走向失败，投资人也不会彻底拉黑创始人，因为后者很可能还会创办下一家划时代的企业；投资人往往会配合完成并不利于 VC 的并购整合，帮创始人兜底薪水、介绍新工作或担任推荐人。创投本质上是一场无限重复博弈，投资人绝不想永远树敌。 主播犀利地指出，必须将“融资 Pitch 的糟糕体验”与“董事会层面的黑天鹅”严格区分开来： - **Pitch 阶段的委屈不值一提**：创始人为了融一轮资可能要见 50 家机构，遇到一两个没读材料、态度粗鲁甚至中途爽约的投资人再正常不过，看不顺眼直接 pass 即可。 - **投资人的风格各有千秋**：有的投资人直言不讳：“我只负责出钱和帮你融下一轮，平时吃个饭当朋友，绝不插手日常业务”，这种风格反倒极受创始人欢迎；有的机构主打深度赋能 GTM（走向市场）或品牌营销，关键在于是否兑现承诺、言行一致。 - **创始人的核心本职**：作为创始人，日常最核心的工作之一就是阶段性出让股权换取现金，理应提前做好背景调查、梳理竞品投资、并且在 Pitch 时有能力牢牢抓住对方的注意力。相比经营企业过程中遇到的真正险阻，投资人在台上打瞌睡根本算不上什么真正的灾难。 ### [24:29](https://www.youtube.com/watch?v=fGnb656Syn8&t=1469s) “红杉骗局”还是常规财技？分批投资与混合估值的真相 AI 招聘平台 Mercor 的 CEO 兼往期节目嘉宾 Brendan Foody 也在社交平台上公开发难，火力全开炮轰 YC 和红杉资本（Sequoia Capital）。他指责红杉资本近半年来在多笔交易中采用“两期分批注资（two tranches）”的手法，称其为“红杉骗局（Sequoia scam）”： ```mermaid flowchart LR accTitle: Sequoia Tranche Valuation Mechanics accDescr: 红杉分批注资模式下对外宣称估值与实际混合估值的机制对比 subgraph Transaction["红杉分批投资注资"] T1["第一期注资: 5 亿美元估值"] T2["第二期注资: 10 亿美元估值"] end Transaction --> Blended["实际混合估值 (Blended Price)<br/>远低于 10 亿美元"] Blended --> Distortion["公关与向后轮天使融资宣称:<br/>'以 10 亿美元估值完成融资'"] ``` Brendan 痛陈，在这种结构下，红杉分别以 5 亿美元和 10 亿美元估值分批注入资金，但对外公关、向员工沟通期权价值以及向后续天使投资人推销时，所有人都在吹嘘这家公司拿到了 10 亿美元估值，刻意隐藏了前段的低价，导致红杉的实际混合价格（blended price）远低于对外宣称数字的 50%。 主播们对此展开了深度剖析： - **交易结构合法且极其常见**：分批注资本身不仅完全合法，而且在很多特定场景下对投融双方都有实际商业合理性。只要追溯到红杉早期对 YouTube 的投资备忘录就会发现，结构化分步投资已经存在了至少 25 年，对冲基金与跨界基金（crossover funds）更是玩得极为熟练。 - **真正的法律雷区在于虚假陈述**：结构化本身没有问题，但如果创始人在明知混合估值极低的情况下，对外向后续天使投资人谎称红杉是全额以 10 亿美元单价投资且隐瞒条款结构，就可能构成严重的证券欺诈（securities fraud）。这种欺瞒责任在于操盘宣传的创始人，而非 VC 本身。 - **信息不对称带来的道德责任**：普通员工在接受期权激励时，以及大多数财经媒体，往往缺乏辨识复杂投资条款与混合折价的专业能力，容易被虚高的账面估值误导。 - **舆论反转的戏剧性**：在发出激进指责并在网上引爆数千点赞后，Brendan 仅仅过了半小时便在原帖下方追加回复，承认“平心而论，这种结构在全行业所有顶级投资机构中都属于普遍操作”，让这场充满火药味的“揭黑”瞬间显得有些雷声大雨点小。 ### [26:34](https://www.youtube.com/watch?v=fGnb656Syn8&t=1594s) 特拉维斯·卡兰尼克的硬核融资往事：雷克萨斯副驾上的路演 节目录制现场继续盘点风投圈的奇葩遭遇，主持人念起了 Uber 联合创始人特拉维斯·卡兰尼克（Travis Kalanick）回忆 2001 年融资时发的一段自述。当时正值互联网泡沫破裂的极寒时期，卡兰尼克在一名 VC 合伙人正准备逃出办公室前将对方当场堵截。最终，他硬是坐进了对方停在车位上的雷克萨斯（Lexus）副驾驶座上强行展开推介。过程中，那位合伙人一把抓过他的笔记本电脑，直接架在自己顶着方向盘的大肚子上，飞快地自行翻看幻灯片。卡兰尼克感慨道："2001 fundraising hit different"（2001年的融资体验确实不同凡响）。 演播室的主持人们被这段充满画面感的往事逗乐，纷纷好奇这个故事的结局究竟如何——那位投资人最后到底有没有当场掏出支票簿投资？抑或是被卡兰尼克的执着打动，感叹一句“好吧特拉维斯，你赢了”。 ### [27:07](https://www.youtube.com/watch?v=fGnb656Syn8&t=1627s) 风投瞌睡实验与路演生存法则：30分钟警戒线与 Herman Miller 诱惑 随后，Tyler 拿出一项真实的科学研究来论证 VC 会议中“投资人容易睡着”的生理机制。该研究选取了一群年龄在 66 岁至 83 岁之间的老年男性，将他们置于光线略显昏暗且缺乏刺激的环境中，测试他们多久会进入睡眠。测试得出的入睡时间中位数仅为 36.9 分钟。主持人们随即调侃道，这描述的完全就是硅谷风投暗淡闭塞的会议室，如果再递给他们一份 Cloudflare 详尽沉闷的 S-1 招股书，催眠效果堪称立竿见影。 基于这一数据，主持人们总结出了一条给创业者的实用法则：常规路演通常被定为一小时，即使开局顺利，一旦会议时长漂移突破 30 分钟，就正式进入了投资人随时可能睡着的“危险区（danger zone）”。尤其是许多硅谷顶级办公室普遍配备极度舒适的 Herman Miller 人体工学椅，大家笑称这种椅子“本质上就是为了在办公室打盹而设计的，根本没人能在上面干正经活”，其舒适度进一步放大了困意。针对这种情况，主持人们打趣提议，如果创业项目比较枯燥，创始人面对年长 VC 时要么必须严格把时间压在半小时内，要么就得自备一个刺耳的“气笛喇叭（air horn）”，随时把快要睡着的投资人震醒。 - **催眠实验数据**：受试群体为 66 至 83 岁男性，在低刺激弱光环境中，入睡时间中位数仅为 36.9 分钟。 - **路演时间分水岭**：早期交流通常预留 30 分钟，一旦拖长至 40 分钟以上便极易诱发 VC 困意。 - **环境催化因素**：硅谷风投办公室昏暗的环境搭配 Herman Miller 人体工学椅，构成了顶级的打盹温床。 ### [28:42](https://www.youtube.com/watch?v=fGnb656Syn8&t=1722s) 软银路演名场面：Rajeev Misra 的心理战与孙正义的“十分钟时限” Tyler 还分享了当年团队为 Divvy 筹资时与软银（SoftBank）交手的传奇故事，称其过程堪称“绝对的电影级名场面（absolute cinema）”。他们首先前往软银位于红木城（Redwood City）的总部向高级合伙人拉吉夫·米斯拉（Rajeev Misra）推介。会面上，对方全程施展了令人叹为观止的心理威压手段：嘴里嚼着尼古丁袋（Zyns）、手里抽着电子烟，甚至故意发出巨大的咳嗽声来打乱创业团队的陈述节奏，期间还有助理不断走上前俯身在 Rajeev 耳边低声耳语。主持人们惊呼这简直是教科书级别的“权力施压（power move）”，戏称这种神神秘秘的耳语助手简直是后奇点时代才有的特殊职位。 紧接着，团队又飞了整整 20 个小时前往东京拜见软银掌舵人孙正义（Masayoshi Son）。经历长途跋涉后，孙正义在会议开场抛出的第一句话却是：“你们有 10 分钟（You have 10 minutes）。” 尽管开局极其紧迫高压，主持人也承认，这种极度压缩时间的方式确实是逼迫创业者迅速剥离废话、直奔商业核心的最快手段。 节目最后，主持人在欢快的氛围中收尾，呼吁观众在 Apple Podcasts 和 Spotify 上打出五星好评，订阅官网每日 Newsletter，并相约下一个工作日上午 11 点准时直播再会。 ## Notable quotes > [00:02](https://www.youtube.com/watch?v=fGnb656Syn8&t=2s) All systems glow > [08:54](https://www.youtube.com/watch?v=fGnb656Syn8&t=534s) 兄弟，我昨晚没搞定，感觉不对，气场不合（Vibe was off） > [08:54](https://www.youtube.com/watch?v=fGnb656Syn8&t=534s) Time to chop wood — 马克·古尔曼 > [10:48](https://www.youtube.com/watch?v=fGnb656Syn8&t=648s) One More Thing — 萨姆·亨利·戈尔德 > [06:57](https://www.youtube.com/watch?v=fGnb656Syn8&t=417s) 宁可让自己的孩子抽香烟，也不愿让他们过早使用 iPhone > [17:51](https://www.youtube.com/watch?v=fGnb656Syn8&t=1071s) 求求上帝，再给我们一个泡沫吧 > [19:28](https://www.youtube.com/watch?v=fGnb656Syn8&t=1168s) 我只负责出钱和帮你融下一轮，平时吃个饭当朋友，绝不插手日常业务 > [26:34](https://www.youtube.com/watch?v=fGnb656Syn8&t=1594s) 2001 fundraising hit different — 特拉维斯·卡兰尼克 > [27:07](https://www.youtube.com/watch?v=fGnb656Syn8&t=1627s) 本质上就是为了在办公室打盹而设计的，根本没人能在上面干正经活 > [28:42](https://www.youtube.com/watch?v=fGnb656Syn8&t=1722s) 你们有 10 分钟（You have 10 minutes） — 孙正义 ## Mentioned TBPN, John Coogan, Jordi Hays, Tyler, Tim Cook, Apple, Apple Intelligence, WWDC, Siri, ChatGPT, Gemini, Claude, Grok, Google, Google DeepMind, Google Cloud Platform (GCP), Ramp, Mac mini, Derek Thompson, Console, Mark Gurman, Lucid, Ferrari, Sam Henri Gold, Eric Seufert, CrowdStrike, Greg Isenberg, Matthew Prince, Cloudflare, Vinod Khosla, Brendan Foody, Mercor, Y Combinator (YC), Sequoia Capital, YouTube, Travis Kalanick, Uber, Herman Miller, Divvy, SoftBank, Rajeev Misra, Masayoshi Son ## Video description Sign up for TBPN’s daily newsletter at TBPN.com Diet TBPN delivers the best of today’s TBPN episode in 30 minutes. TBPN is a live tech talk show hosted by John Coogan and Jordi Hays, streaming weekdays 11–2 PT on X and YouTube, with each episode posted to podcast platforms right after. Described by The New York Times as “Silicon Valley’s newest obsession,” the show has recently featured Mark Zuckerberg, Sam Altman, Mark Cuban, and Satya Nadella. TBPN is made possible by: Ramp - https://ramp.com Public - https://public.com Cisco - https://www.cisco.com Console - https://www.console.com CrowdStrike - https://www.crowdstrike.com Figma - https://www.figma.com MongoDB - https://www.mongodb.com NYSE - https://www.nyse.com Railway - https://railway.com Shopify - https://www.shopify.com/ Follow TBPN: https://TBPN.com https://x.com/tbpn https://open.spotify.com/show/2L6WMqY3GUPCGBD0dX6p00?si=674252d53acf4231 https://podcasts.apple.com/us/podcast/technology-brothers/id1772360235 https://www.youtube.com/@TBPNLive [Source](https://www.youtube.com/watch?v=fGnb656Syn8)

[0:02] All systems glow [8:54] 兄弟，我昨晚没搞定，感觉不对，气场不合（Vibe was off） [8:54] 马克·古尔曼: Time to chop wood [10:48] 萨姆·亨利·戈尔德: One More Thing [6:57] 宁可让自己的孩子抽香烟，也不愿让他们过早使用 iPhone [17:51] 求求上帝，再给我们一个泡沫吧 [19:28] 我只负责出钱和帮你融下一轮，平时吃个饭当朋友，绝不插手日常业务 [26:34] 特拉维斯·卡兰尼克: 2001 fundraising hit different [27:07] 本质上就是为了在办公室打盹而设计的，根本没人能在上面干正经活 [28:42] 孙正义: 你们有 10 分钟（You have 10 minutes）

[7:06] Host: She specifically said I would rather let my kids smoke cigarettes than use an iPhone, which is crazy... maybe the cure for cancer is right around the corner, but the cure for brain rot is not. [19:28] Investor (quoted by Host): I give you money, and then I'll help you raise more money. That's the only thing I'm going to do, and I'm going to be your friend.

[0:02] Two WWDC's ago, we were talking about [0:03] Apple Intelligence, and it's all finally [0:06] coming together. The package is finally [0:09] uh being delivered. I think the response [0:10] has been really good. Some of the guys [0:12] on the team have been watching WWDC [0:13] already. We have a bunch of folks [0:14] calling in to discuss WWDC throughout [0:17] the week, but let's give you the high [0:20] level first, go through some of the [0:21] immediate reactions, and of course we'll [0:22] be covering that throughout this week. [0:24] So, Apple's annual worldwide developers [0:27] conference started today, and it runs [0:29] through Friday. Tim Cook just kicked it [0:31] off with an opening keynote. The theme [0:33] for this year's conference is all [0:36] systems glow. All systems glow. Little [0:39] party. We're finally getting answers [0:41] about what the next version of Siri will [0:43] look like. I think expectations in in a [0:45] weird place. Like they're high. [0:47] Everyone's expecting like this next [0:49] version of iOS, which is what they're [0:51] demoing. That's the main thing of the [0:53] software version. This isn't an iPhone [0:55] event. Uh this isn't a hardware event. [0:57] This is WWDC about the software. [0:59] Everyone's expecting that the software [1:00] will go through an actual [1:02] transformation, and the next version of [1:04] Apple software will be good, and people [1:07] will be talking about how good it is [1:08] because they will deliver on a bunch of [1:09] things. At the same time, expectations [1:11] aren't so high like going into Apple [1:14] Vision Pro, where people are are [1:16] expecting a breakthrough that's no one [1:18] uh no other company has ever done [1:20] before. All people are asking for is [1:23] implement the best practices from [1:25] ChatGPT, Gemini, Claude, like the stuff [1:27] we know and love. Even Grok has like [1:30] nicely inter [1:31] interfaced into X, where like people [1:34] say, "Hey, Grok, is this real?" And it [1:35] just pulls it up for you. I find myself [1:37] on tweets just going, "Oh, yeah. Like I [1:39] don't want to copy this text out and [1:40] into another LLM app. I'm happy just [1:43] asking Grok real quick for an extra [1:45] detail or one more fact." Google search [1:47] over use like There's a been a lot of AI [1:49] diffusion into products that's been [1:51] good. Even Ramp, like Ramp has like chat [1:53] interface, where you can just say like, [1:54] "Hey, how much did we actually spend on [1:55] Amazon last month?" And it'll just pull [1:57] it up for you. And it's not like this [1:59] revolutionary AGI. [2:00] >> the thing. Here's the thing. I think [2:02] they did a I think they did a good job [2:04] letting the hype build or or the [2:06] interest build organically, right? You [2:08] rewind [2:09] >> Yeah. [2:09] >> a year or two ago, they were running [2:11] billboards for Apple intelligence, for [2:13] Chad Mojo, all this stuff. So, really [2:15] setting themselves up for failure. [2:16] >> stuff was overhyped. And I think this [2:19] >> And they were they were a part of a They [2:20] were They were They were hyping it [2:22] themselves. [2:23] >> Of course. Of course. [2:24] >> They were like, "Are we coded?" [2:26] >> Yeah, but it seems like they have the [2:27] right partnerships and and the right [2:29] like product strategy at this point. [2:31] People are familiar with these tools, so [2:32] just putting them in different places [2:34] seems reasonable. Models are good now, [2:36] so make them available at the click of a [2:38] button, ideally the Siri button, which [2:40] has been completely nerfed for the last [2:42] 2 years. [2:44] Actually, more than that because Siri [2:46] has has never really seen the adoption [2:48] or love product love that many other [2:49] products have seen. And users will be [2:51] happy. So, Google Google AI search [2:53] overviews are a good example of proof [2:55] that rolling out LLMs, it can be [2:57] nerve-racking, but it's not rocket [3:00] science. Like, you spit the text out [3:02] where people expect the answer. There [3:04] will be funny viral hallucinations. [3:06] Like, right now, even with all the crazy [3:08] Google IO news, amazing 5.3.5 [3:12] foundation models, like they're doing [3:14] really good stuff. Deep Mind's really [3:15] good great team. Like, you still say [3:17] "disregard" and instead of just giving [3:19] you the definition, it says like, "Okay, [3:21] got it. I won't I won't do that [3:23] anymore." And it's it gets confused. So, [3:25] there will be those like viral jokey wow [3:28] it failed it it flopped. That's going to [3:30] happen to Apple. And that's not what [3:32] Apple likes to deal with. [3:34] >> Not at all. [3:34] >> But, I don't think any of that will show [3:36] up in the user metrics. I think that, [3:39] you know, churn and usage will be [3:41] unaffected from the viral moment of [3:42] like, "Oh, Apple like the the Apple [3:46] text summaries, they're hilarious and [3:47] they often hallucinate and get things [3:49] wrong." [3:50] >> You still keep them on, though? [3:51] >> I left them on. I don't mind them. I and [3:53] I think a lot of times they're sort of [3:54] useful and they're always delightful cuz [3:56] they're funny cuz they're so like [3:58] hallucinatory. But your PR team will [4:01] have many heart attacks because you're [4:03] not in the world of deterministic [4:04] outputs anymore. So that's going to be a [4:06] big cultural shift for Apple I think in [4:08] this [4:09] non-deterministic stochastic AI era. But [4:12] I think that they can get through it [4:13] just by running best practices that have [4:15] been established for a year or 2 years [4:18] in the rest of the AI world. The other [4:20] big questions that everyone's asking [4:21] around open ecosystems. So will Apple [4:24] lean into the open claw Mac mini boom at [4:27] all? That would be interesting to see. [4:29] Say, you know, in in the next version of [4:32] the software that goes on the Mac mini, [4:34] hey, we're going to embrace open claw. [4:35] We're going to embrace AI agents. We're [4:37] going to do things that make those tools [4:40] more effective in our ecosystem. You [4:42] already know and love the hardware. [4:43] You're buying it non-stop. It's out of [4:45] stock. But we could do more to lean into [4:47] that community or we could do less. We [4:49] could say, hey, we're going to shut it [4:50] down. We're worried about privacy. These [4:52] are the two tensions that they're going [4:53] to have to deal with. Will there be a [4:55] pivot around vibe coding apps in the iOS [4:57] App Store? You asked John Gruber from [4:59] Daring Fireball this when he joined the [5:01] show on May 29th. Fantastic interview. [5:04] Had a ton of fun with the Gruber nator. [5:06] But it's a big question and it's [5:08] something that Apple is they don't [5:11] really have to respond yet. They're [5:12] pretty quiet when things happen. They [5:15] usually go and solve the problem and [5:17] >> Oh, they really don't want to talk about [5:18] it. [5:19] >> Yeah, but eventually they start. Like [5:22] they didn't want to talk about climate [5:23] change, but then they did a bunch of [5:24] things to get to like net zero and [5:26] eco-friendly buildings and solar panels [5:29] on the roof and stuff. And then once [5:30] they did, they were really noisy about [5:32] it cuz they were like, we are carbon [5:33] neutral or we will be by a certain date. [5:35] >> once they figure out how to make money [5:37] on it, then they'll start. [5:39] >> should. It's [5:39] >> Which which yeah, I fully I will say I [5:42] fully support. [5:43] >> Exactly. Will native iOS apps from other [5:46] AI labs have more access to iPhone [5:49] functionality. What's the pathway for [5:51] ChatGPT interfacing, Claude, Gemini? If [5:54] you're using those apps, how how how [5:57] many how many hooks are there? Will they [5:58] be able to siphon in your text messages [6:00] if you click, "Yeah, I want to share my [6:03] text messages, my iMessage with my app [6:06] of choice." Or will you need to will you [6:08] need to say yes every single time, which [6:10] will be a huge burden to having that [6:12] integration happen? Uh there's a lot of [6:14] there's a lot of social media apps that [6:16] say, "Hey, we want access to your whole [6:17] camera roll. All of it forever." And [6:20] that's quite [6:21] somewhat intimidating. Uh you get a [6:23] number of different options. You can [6:24] say, "I want a temporary access. Just [6:27] take this photo. You can't see my whole [6:29] library." Now, people trust many people. [6:32] May- maybe they just don't know, but [6:33] they they feel like they trust a lot of [6:35] these maps. So, they just say, "Yeah, [6:36] yeah, take the whole camera roll." Sort [6:37] of a crazy thing to do that they can [6:38] just download your entire camera roll if [6:41] you press that button. But people have. [6:43] And so, what will the what will the AI [6:45] version of that be? And how deep will it [6:47] be? And how much will it build on top of [6:49] the uh Siri app intents functionality [6:52] versus other APIs that are deeper in the [6:55] iOS ecosystem. There's one last thing, [6:57] which is we I mean, we talked about how [7:00] Apple doesn't want to address eco stuff [7:02] until it's like, "Okay, we got it [7:04] solved." And I'm wondering if in the [7:06] coming years there's going to be [7:08] pressure for some sort of response on [7:11] the whole like phones are reducing the [7:13] fertility rate stats. Because I don't [7:15] know if you saw, but there was another [7:16] research paper that was posted and Derek [7:18] Thompson basically uh zoomed out and [7:21] said that I wasn't convinced and now he [7:24] he is convinced that it's like up to 30% [7:26] of the reason for the recent decline [7:29] below two. And I don't know where you [7:31] sit on it. Below the replacement rate. [7:32] Not good if you're a fan of humanity and [7:36] having a high population. But uh that is [7:39] something that you could see bubbling up [7:43] to being something that big tech [7:45] companies, social media companies, [7:47] device manufacturers have to answer to [7:50] when they're in like free form podcasts [7:52] basically. But, you know that these [7:54] companies [7:54] >> Yeah, it's interesting. It's [7:55] interesting. As opposed to the AI CEOs [7:57] that are like, "Oh, you want to talk [7:58] about it? Telling everyone? Absolutely. [7:59] I'd love to talk about it for an hour." [8:01] >> the PM of Denmark or what what country [8:03] was saying [8:04] >> No, she specifically said I would I [8:05] would rather let my kids smoke [8:06] cigarettes than use an iPhone, which is [8:09] crazy. [8:13] Um but, I don't know, maybe there's [8:14] something there. Maybe the cure for [8:16] cancer is right around the corner, but [8:18] the cure for brain rot is not. [8:19] >> Yeah, and it's such a different debate [8:21] because clearly a single drag on a [8:24] heater [8:25] is no good, right? You're ingesting [8:27] poison. [8:29] A single look at a screen, right, is not [8:31] what is, you know, reducing uh the the [8:34] fertility rate. Uh so, [8:37] anyways. [8:39] >> It's interesting. [8:40] >> Uh anyway, let's go to some reactions [8:42] from WWDC, but first I'm going to tell [8:43] you about Console. Console builds AI [8:45] agents that automate 70% [laughter] of [8:47] IT, HR, and finance support, giving [8:49] employees instant resolution for access [8:51] requests and password resets. [8:52] >> iPhone, I'm going to update your [8:54] software tonight while you sleep. Next [8:57] morning, iPhone says, "I couldn't do it, [8:59] bro. Just didn't feel right. Vibe was [9:01] off." [9:02] >> This is something that [9:04] I didn't know was so common, but I think [9:06] everyone's been in this position, which [9:08] is like a funny thing. It has to do with [9:10] how how much the battery is charged, but [9:12] lots of low-hanging fruit hopefully [9:14] resolved in WWDC. I did see a whole [9:16] bunch of stats about [9:18] just just little performance gains, 30% [9:21] faster opening of the lock screen, 30% [9:24] faster on opening this app, and just [9:27] little optimizations that I think will [9:29] go a lot further when we talked to Mark [9:30] Gurman. He talks about how like the AI [9:33] features are too abstract. People want [9:35] battery life, cameras, beautiful [9:37] screens, fast, like they want the basics [9:40] most of the time. And so, I think this [9:43] is uh time to chop wood. [9:44] >> be funny if they they effectively threw [9:46] up a model card and they're comparing [9:48] it's like [9:48] >> They actually sort of did that. The [9:49] bento box is the original model card. [9:51] >> It's 2% better on [9:53] >> That's sort of what they do. [9:55] You know the bento box, right? The bento [9:57] box with like how many cameras, [9:59] megapixels, flops, and how powerful the [10:02] GPU is, how many cores are over here, [10:04] how much storage it has. Like that [10:06] graphic is their model card. We we we [10:08] were riffing on this earlier like when a [10:10] product ceases to be sold on brand and [10:13] and is instead sold on performance, uh [10:16] that's usually [10:17] >> compression. [10:18] >> Usually margin compression. Like you [10:19] comp it to cars. Like if you're purely [10:21] buying on like range and price and speed [10:25] and uh horsepower and seating [10:27] arrangements, that feels much more [10:29] commoditized than you got to have a [10:30] Ferrari because it's a Ferrari. Don't [10:32] ask about the specs. And I think for for [10:34] the Lucid, this is the first time that [10:35] people were talking about like, "Oh, 0 [10:37] to 60 in 2.5 [10:39] for that price isn't actually that [10:41] good." It's like that's not the [10:42] conversation you should be having about [10:44] Ferrari. You should just be like, "It's [10:45] a Ferrari. Next question. Like do you [10:47] want one or not?" Hilarious post from [10:49] Sam Henry Gold here. Of course in jest, [10:52] but they put up a Apple press release in [10:55] the newsroom. Apple announces the death [10:57] of Mark Gurman. [10:58] >> It is done. [10:59] >> It is done. [11:00] Apple today announced the completion of [11:02] operation one more thing, a multi-year [11:04] initiative to permanently end the [11:05] unauthorized disclosure of Apple's [11:07] pre-release product information by [11:08] Bloomberg Intelligence reporter Mark [11:10] Gurman. Gurman, who had for 16 [11:12] consecutive years obtained and published [11:14] accurate details about Apple's products [11:16] before their announcement, has been [11:18] neutralized. Operation one more thing [11:19] was completed on schedule, on budget, [11:21] and without complication. They're not [11:23] that aggressive towards Mark Gurman, but [11:25] he is probably a thorn in their side. [11:26] And here he is. He has a live blog up, [11:29] live blog for WWDC. Go check it out at [11:31] bloomberg.com. Eric Seufert is leaning [11:33] in. Why is he leaning in? Because Mark [11:35] Gurman says something about WWDC. He [11:37] says, "In addition to the focus on AI, [11:39] Siri, and major quality and performance [11:41] improvements across the operating [11:43] systems, I'd expect two other focal [11:46] points today, privacy and safety [11:49] features." I imagine Seufert is leaning [11:51] in because privacy, safety features, not [11:54] great for ad monetization usually. [11:56] You're hiding more information. Tyler, I [11:58] what what's been your reaction [11:59] generally? Take me through whatever. [12:00] >> Yeah, I've been watching the live [12:02] stream. It's it's still going on, so [12:03] there's there's still releasing stuff. [12:04] But I I would say just on that point, [12:05] they've been like every single time [12:07] >> or have you been studying? [12:07] >> I was studying. [12:08] >> Mhm. [12:09] >> Like every single every time they talk [12:10] about a new feature with AI, they always [12:13] say like this is on a private cloud. [12:15] This is like not public. This is like [12:17] extremely secure. So they're they're [12:18] really pushing on that point, I think. [12:20] >> What does private cloud mean? [12:23] >> I don't know. They're talking about like [12:24] their own foundation models, I think, so [12:26] they're trying to emphasize that. [12:27] >> So [12:27] >> They said the word Gemini. [12:29] >> Uh I don't know if they said it, but it [12:31] was on screen. [12:32] >> So so it's [12:33] >> It seems like they have the ability with [12:35] their deal with Gemini to basically [12:38] white label, fine-tune, mid-train, do [12:40] whatever they need to, and then resell [12:43] that with their own branding. That's [12:44] great deal. My question is like who's [12:47] inferencing that? There's a billion [12:49] iPhone users. [12:51] And if they're all pushing the Siri [12:52] button all day long, and they're [12:54] anywhere near the frontier, that's a [12:57] significant amount of inference. Has [12:59] Apple built some sort of secret data [13:00] center that can serve that? Are they [13:02] saying private cloud and really it's [13:04] like a corner of GCP? Is it a bunch of [13:06] Mac minis wired together? Like what did [13:09] they do to actually build that private [13:11] cloud? Because even if they did it, even [13:13] if they did it and didn't show up in the [13:14] capex numbers, didn't show up in the in [13:17] the any of the [13:18] SEC filings, like it would show up in [13:21] the emissions data and the and the ESG [13:23] numbers. Because unless they have some [13:26] crazy solar nuclear powered facility, [13:29] where is that inference coming from? I [13:31] imagine some of it can be done on [13:32] device. That's exciting. [13:33] >> says it's on device. [13:35] >> Can it all be done on device? Like [13:37] >> Tyler [13:38] >> I don't know. [13:39] >> Goroxen says they got rate limits and a [13:41] subscription plan. [13:43] >> Huh. [13:44] >> Did you see that? [13:45] >> Yeah, I'll look for [13:46] >> I mean, you don't have rate limits or [13:47] subscription plan if it's on device [13:49] because why would you? Uh and then also [13:52] you would never say private cloud if [13:54] you're doing it on device. You would [13:56] just say it's on device. Of course it's [13:57] private. So, private cloud cloud means [13:59] not on device. It's in the cloud. And [14:01] so, I obviously Apple has a lot of data [14:04] centers. They have a lot of cloud [14:05] capacity across their productivity [14:09] suite. Where are they Where are the [14:10] photos stored in iPhotos? Like obviously [14:14] those are stored in the cloud. They have [14:15] a lot of storage. Like everyone gets 2 [14:17] terabytes when they get a phone. So, [14:18] they clearly have a lot of data center [14:20] capacity and they've done a lot to make [14:22] that ESG compliant. But I'm very curious [14:24] about where that goes over time. Anyway, [14:27] we should move on to the the the news [14:30] from Friday. The stocks are already back [14:32] up. But first I'm going to tell you [14:33] about CrowdStrike. Your business is AI. [14:35] Their business is securing it. [14:37] CrowdStrike secures AI and stops [14:38] breaches. [14:39] >> So, uh one more thing. [14:41] >> Oh, sure. [14:41] >> Apparently [14:42] >> thing, huh? [14:43] >> One more thing. [14:45] >> more take from Jordan about Apple. [14:47] >> Jane says Apple concedes on liquid glass [14:50] design compromising for usability and [14:53] shares a couple screenshots here. Uh and [14:56] so [14:57] >> Wait, so they're going deeper in liquid [14:59] glass or pulling back from it? [15:00] >> Pulling back. [15:01] >> I I saw the new [15:03] the new Apple Maps icon and it looked [15:05] pretty good. It had a little feature of [15:07] liquid glass in there. I thought it [15:08] looked cool. I did hear people [15:09] complaining about the new Mac operating [15:12] system being like too too bright or too [15:15] dark or something. Like some contrast [15:16] issue. I haven't noticed it, but it was [15:19] something I saw people complaining [15:20] about. Anyway, Friday was a big day in [15:21] the market. We were off, but uh the news [15:24] kept moving, so we're covering it today. [15:26] Uh the Labor Department reported that [15:29] the US added a seasonally adjusted [15:30] 172,000 [15:32] jobs, and that the unemployment rate [15:35] remained unchanged at 4.3%. So, on the [15:37] cover of the Wall Street Journal, US [15:39] hiring gathers steam, third straight [15:41] monthly increase. Uh slight decrease [15:44] from the previous month, which was [15:45] slight down from the previous month, but [15:47] still in adding territory in the [15:49] hundreds of thousands. So, uh the bubble [15:52] popped. The bubble popped. Friday was [15:54] the worst day for the Nasdaq in more [15:56] than a year, 4.2% down. It's over, but [15:59] good news is that today we're back. [16:01] We're up 1.5% now. Uh it's officially [16:03] 2003. People ask what year it is. [16:07] >> It's No, no, it's not re-inflating. [16:08] We're building back. The bubble popped. [16:09] It's 2003 now. [16:11] It's not '99. It's not 2000. It's 2003. [16:15] We're well past the bubble popping. [16:16] >> Okay. Got it. [16:17] >> Right? So, uh but actually, quick [16:19] explanation of what happened. So, uh the [16:21] US labor market is really picking up, [16:23] 172,000 seasonally adjusted jobs added [16:25] in May. [16:26] >> Yes, a third month in a row. Decent [16:27] amount. [16:28] >> Decent amount. [16:28] >> Are for the World Cup? [16:31] >> Yes, and a lot of travel, and uh and and [16:34] workers related to the World Cup stuff [16:35] that's going on, tourism. This is [16:37] terrible news for all those black-pilled [16:38] AI leaders that have been praying, [16:40] manifesting job losses. Despite their [16:42] Herculean efforts, they can't get the [16:45] unemployment rate to go up at all. It's [16:47] crazy. They've been saying 10%, 20%, [16:50] 50%, 100% of all jobs are going away. [16:53] Good luck. Yeah, you're going to have to [16:55] work harder because the US economy is [16:58] undefeated, and the American worker is [17:00] undefeated, as evidenced by this latest [17:03] jobs report. No. [17:04] Um [17:05] the AI job apocalypse is canceled, at [17:06] least for the month of May. We will see [17:09] where things go, of course, of course. [17:10] But it is good news. We want hiring. We [17:12] want jobs to be abundant in our society. [17:16] Uh and so, in general, it's good news. I [17:17] don't I I believe the jobs report. I [17:19] don't think the numbers are going to be [17:21] massively revised down. I think that uh [17:23] they're they're they're generally [17:25] accurate and track with the ADP numbers [17:28] and a lot of other numbers. So, um [17:29] >> [snorts] [17:30] >> I think the jobs are are really being [17:32] added. They're not in all the most [17:33] critical industries. There's a lot of [17:35] nuance there. How long will it go on? [17:36] But in general, the economy is healthy. [17:38] Uh but inflation is rising. Clo- The [17:41] closing of the Strait of Hormuz has [17:42] spiked the cost of gas and overall [17:44] prices have been increasing more quickly [17:46] than the Fed would like for some time [17:47] even before the Strait of Hormuz. [17:49] Inflation was running a little hot. [17:51] >> Yeah, well above the 2% target [17:53] >> Yes. [17:53] >> for basically [17:56] as long as I've been an adult. [17:57] >> Yeah. And so, this makes the likelihood [18:00] of a rate cut more unlikely. In fact, it [18:03] looks like we might be in rate hike [18:04] territory soon, which is of course not [18:07] good for tech companies that have [18:08] earning forecasts that stretch out into [18:10] the next decade. So, the silver lining [18:12] in high rates, if you want some copium, [18:15] is that at least the Fed has something [18:17] useful in the tool chest in case the [18:19] economy does slow down, you know? Like [18:21] we're running hot, we have high rates. [18:24] At least there's room to cut to 3%, 2%, [18:28] 1%, 0 if the market's selling off, if [18:33] the unemployment rate's going up. You [18:34] have something in the tank, whereas [18:36] during COVID, like the rates were [18:37] already so low, there was a lot of [18:39] unemployment all of a sudden, and it was [18:40] just like stimulus, spend a bunch of [18:42] money, [18:43] uh you know, send mail everyone a check. [18:45] There wasn't that much that the Fed [18:46] could do. [18:47] >> Yeah, there was a time that we couldn't [18:49] imagine this level of speculation in the [18:51] markets at something like a at at where [18:54] rates are right now, right? [18:56] Uh [18:56] >> Yeah. [18:57] >> People that were sort of born in the [18:59] ZIRP era. [19:00] >> yeah. [19:00] >> Uh with with all the speculation right [19:02] now is by itself an argument to raise [19:05] rates [19:05] >> Yeah, yeah, yeah. Totally. Well, yeah, [19:07] once the rates spiked uh [19:09] off of like uh like that 3% jump, like [19:12] the end of ZIRP, it was like, okay, like [19:14] there will never be froth again, [19:15] certainly. [19:16] >> [laughter] [19:17] >> Certainly, like tech stocks will never [19:19] see the year 2000. [19:20] >> Blake, made bumper stickers that just [19:22] said, "Please God, just one more [19:24] bubble." [19:25] >> And God delivered. Okay, so. [19:28] The other story, VC horror stories. [19:30] This, uh, this was kicked off by Greg [19:33] Eisenberg. I didn't realize that he was [19:34] the one that started this whole thing. [19:36] He kicked the hornet's nest and everyone [19:37] came out of the woodwork with their VC [19:39] horror stories. He said, uh, he had a [19:41] big discussion about founders' bad [19:43] experience with venture capital and a [19:44] number of high-profile firms, VCs, [19:46] caught strays. Mercor CEO, Brendan [19:49] Foody, detailed what he calls the [19:50] Sequoia scam, uh, which is something [19:52] interesting we should actually dig into. [19:54] Uh, Cloudflare CEO, Matthew Prince, [19:56] accused Vinod Khosla of offering to [19:59] invest in his series C only if he would [20:01] fire a few of of the people at the pitch [20:03] meeting who had just left the room [20:04] momentarily to go to the bathroom. See, [20:07] there was a separate So, Greg Eisenberg, [20:09] I was I was really crossing wires here [20:11] because Greg Eisenberg is the one who [20:13] says, uh, he says, "I was once pitching [20:17] in a boardroom at a top three VC firm [20:19] for a $15 million series A." That's [20:21] pretty easy to narrow down. But anyway, [20:24] he says, "12 people in the meeting, one [20:25] of the GPs fully fell asleep because [20:27] some of the top like, you know, FF [20:30] doesn't have 12 GPs. [20:31] >> [laughter] [20:32] >> They don't really do partner meetings [20:33] like this. So, uh, for $15 million [20:35] series A, so even if you put them in the [20:36] top three, you got can't catch a stray [20:38] here. [20:39] Anyway, one of the GPs fully fell [20:41] asleep, out cold for 30-plus minutes. [20:44] Nobody acknowledged it, everyone kept [20:45] going. Then separately, Matthew Prince [20:47] >> Okay, founder or or an operator falls [20:52] asleep in the office cuz they're so [20:54] tired cuz they're grinding so hard, and [20:56] they're a hero. [20:56] >> Yeah, this is what Packy said. When Elon [20:59] falls asleep in the in the factory, it's [21:01] no big deal. But when a VC falls asleep, [21:03] it's, uh, it's the end of the world. [21:05] >> We're aware of some static in the [21:07] ultradome. [21:08] >> Yeah, I think we got a new new uh new [21:10] mic line, but well I'll try not to touch [21:12] it and we'll see what happens. So [21:13] there's lots of chatter about the VC [21:14] horror stories. Certainly certainly a [21:16] discussion worth having. You know, you [21:18] got to keep them in check. But uh [21:20] Silicon Valley has always been so high [21:22] growth and positive sum that relatively [21:24] good behavior is usually the [21:26] equilibrium. Uh it's pretty rare that [21:29] you get a really a really bad VC. Uh [21:31] because even when a startup fails, [21:33] investors can't write off the founder [21:36] entirely because they might start the [21:38] next generational company. And so they [21:40] might wind up giving on a [21:43] on a non-VC friendly sort of acquire [21:45] because they're like and help or like [21:48] help them land a paycheck somewhere, get [21:50] a job, serve as a good reference, maybe [21:52] even fund the next product in the next [21:54] company because they're just like this [21:56] is an iterated game and we don't want to [21:59] have you on our bad side forever. Now, [22:01] that doesn't mean don't like they can't [22:03] pass. And so there's basically this wide [22:05] gap that I'm seeing in these discussions [22:07] where there's VC pitch horror stories [22:10] and then VC board horror stories. And I [22:14] have much less sympathy for the former. [22:16] Like I don't really care about VC pitch [22:18] horror stories all that much because you [22:20] can just pass. [22:20] >> cuz a lot of founders will have 50 [22:23] meetings for a financing. You would [22:25] expect at least a couple to just be [22:27] terrible, right? The person didn't know [22:30] who you were, didn't read the materials [22:32] ahead of time, was rude. [22:34] >> And you might want that person [22:36] >> up. Didn't show up. [22:37] >> You might want a checked out VC who's [22:39] just going to let you cook. And they're [22:40] like, yeah, I see this purely [22:42] financially. My my my team crunched the [22:44] numbers. I'm in, but like [22:46] don't count on me to value add. Like I'm [22:48] not going to be in the weeds with you [22:50] every day. And then there's a different [22:52] there's a different VC who's like I'm [22:53] going to be in the office. [22:54] >> The [22:56] best you, oh I'm going to be grinding [22:58] for you every day. I'm going to be [22:59] helping you get all your key hires. [23:01] They'll Yeah, they'll just tell you. [23:02] Like one of my favorite VCs in the world [23:05] just says like, "I give you money [23:07] >> Yep. [23:07] >> and then I'll help you raise more money. [23:09] And that's the only thing I'm going to [23:10] do [23:10] >> And that's fine. [23:11] >> and I'm going to be your friend. We'll [23:12] get dinner now and then, but that's what [23:14] you can expect from me. [23:15] >> Yep. [23:15] >> And that's exactly what he gives [23:16] founders. And so everyone is like, "This [23:18] guy's great." [23:19] >> Yeah. And then there's other There's [23:20] other firms that say, uh "We'll help you [23:22] with go to market." and they will. [23:24] There's other people that say, "We will [23:25] help you with marketing." and they [23:26] don't. And you just want to be [23:28] transparent and accurate there. Uh and [23:30] so like as a founder your job is to sell [23:33] equity from time to time. Your job's to [23:35] find buyers for that equity investors. [23:37] >> actually only do that. [23:38] >> That's true. They don't actually talk to [23:39] the product occasionally. [23:41] >> [laughter] [23:41] >> Uh so your job's to find investors who [23:43] want to purchase that that equity with [23:45] cash VCs. You need to reference check [23:47] them beforehand. Make sure they're fit [23:48] through Think through their competitive [23:50] investments. Keep them entertained and [23:52] awake during the pitch. VCs shouldn't be [23:54] disrespectful. Like literally falling [23:56] asleep is. Uh but this is the far This [23:59] is far from the worst thing that [24:00] regularly happens in the course of [24:01] growing a business. [24:01] >> We pitched for Figma's seed round in [24:04] 2013. Most folks didn't get it, but [24:06] everyone I met was super nice to me. [24:08] >> [laughter] [24:08] >> Uh yeah, it's so funny when you when you [24:10] compare when you compare VCs to the [24:12] other like I said the other kinds of [24:14] calls you have. VCs are probably like [24:17] generally way nicer, right? Like a a [24:19] customer is less likely to like be [24:22] overly friendly if they're not [24:23] interested in what you're selling, [24:24] right? They're like, "Yeah, this is like [24:26] doesn't really seem like a you know, [24:28] doesn't really seem like a fit." [24:29] >> Yeah. [24:29] >> Brendan from our core former guest and [24:32] friend of the show uh is has a bone to [24:35] pick with uh everyone. [24:37] >> Yeah. [24:38] >> Uh he's coming after YC. He's coming [24:40] after Sequoia. [24:41] >> He is. [24:41] >> Uh he says, "In the last 6 months I've [24:43] seen half a dozen rounds where Sequoia [24:45] invests in two tranches. Everyone [24:47] pretends they only did the higher [24:48] valuation. Founders misrepresent this to [24:50] their employees and then shop it to [24:52] angels, too. [24:53] Sequoia's blended price is completely [24:55] deceptive. Less than 50% of the one [24:57] projected to the market. [24:59] >> Yeah, this has been happening. They'll [25:00] invest at half a billion and a billion, [25:03] two tranches, and then the founder will [25:05] go out and say we raised 100 million at [25:07] a billion when really Sequoia got in at [25:09] the blended price, right? Um now, that's [25:13] not illegal. There's nothing wrong with [25:15] that, and that can make sense for both [25:17] sides for a variety of reasons. You [25:19] don't want to misrepresent that though. [25:21] If you go and misrepresent that to [25:23] another investor and you don't tell them [25:25] the actual structure of the deal, that [25:26] can be securities fraud. So, that is a [25:29] major risk. [25:30] >> But, that's not on the VC. Like, the VC [25:32] should not go to another investor and [25:33] say, "Oh yeah, we just did uh the full [25:35] deal at a billion." They should give [25:37] that context. But, that doesn't seem [25:38] like this [25:39] >> it's on the founder in the same way. [25:41] >> unique thing. This is like classic like [25:42] all the crossover funds were doing this. [25:44] Tons of funds have done this. This I [25:45] mean, [25:46] also this isn't some secret. Like, you [25:48] go back to the original Sequoia YouTube [25:50] investment memo, and it's tranches and [25:52] structured. Like, they've been doing [25:54] this for 25 years, and it's just like [25:57] nobody read nobody read the manual or [25:59] something because like if you actually [26:00] study and know ball, like you would know [26:03] that structured investments exist. At [26:05] the same time, I don't put it on every [26:07] employee who's getting stock options and [26:09] thinking about the heat on a company and [26:10] every journalist to understand that [26:12] structure. So, there is there is an [26:14] impetus and a responsibility. [26:16] >> Yeah, very aggressive and unnecessary to [26:19] call this [26:20] a quote Sequoia scam cuz Brendan also [26:23] replied to his own tweet saying uh [26:24] >> just 30 minutes ago after the post had [26:26] gone viral. In fairness to Sequoia, this [26:28] is common practice in the industry [26:29] across all top firms. [26:31] >> 1,000 likes [26:32] >> Brutal. [26:33] Brutal. [26:34] >> Travis Kalanick said in 2001, "I [26:36] intercepted a partner to VC who was [26:38] trying to escape his office before our [26:39] meeting was supposed to start. I ended [26:41] up pitching him in his parked Lexus [26:43] uh from the passenger seat. At one [26:44] point, he grabbed my laptop, placed it [26:46] on his large belly, which was pressed [26:48] against the steering wheel, and rapidly [26:50] flipped through the [26:52] slides himself. 2001 fundraising hit [26:55] different. [26:56] >> I want to know how the story ended. [26:57] >> did did he invest or not? [26:59] >> pull out his checkbook? [27:00] >> Cuz maybe he was like, "You got me, [27:01] Travis." You know, hey. We actually had [27:04] a plan to test this theory. Tyler, do [27:06] you want to take us through it? [27:07] >> Yeah, well, so I I was just looking at [27:08] this. I found a study. Basically, it was [27:10] you they took a bunch of older men, 66 [27:13] to 83. [27:14] >> This actually happened. [27:14] >> They they put them in a room where it [27:16] was kind of a slightly dim room. [27:18] >> Okay. [27:18] >> There's not a lot of like stimulating [27:20] >> These are offices. They're very shady. [27:22] >> just sat there and they tested how long [27:23] >> them Cloudflare's S-1. [27:25] >> [laughter] [27:26] >> And then they basically tested to see [27:28] how long would it take them to fall [27:29] asleep. And so the median was 36.9 [27:33] minutes. [27:34] >> I feel like Okay, so [27:35] >> So, if you ever like I mean, and here's [27:36] the thing. [27:36] >> A pitch [27:37] >> A pitch [27:37] >> Usually an hour. [27:38] >> Well, sometimes I mean, if it's an early [27:40] pitch, early conversation, maybe it's 30 [27:42] minutes. [27:42] >> Maybe. [27:42] >> If it's going well, it starts to drift [27:44] over, but that's the danger zone. [27:46] >> Danger zone. [27:46] >> [laughter] [27:47] >> So, this is the lesson. This is the [27:49] lesson for founders. If you have a [27:51] boring business If you have a boring [27:53] business and you're pitching a VC who's [27:55] an older gentleman, you got to keep the [27:57] pitch meeting to less than 30 minutes. [27:59] You got to bring the air horn. [28:00] >> [laughter] [28:01] >> Especially so I I think in in that [28:03] story, it also said the older gentleman [28:05] was in a Herman Miller chair, which we [28:07] know are quite comfortable. [28:08] >> Those are quite popular in Silicon [28:09] Valley, too. [28:10] >> compounds the effect. [28:11] >> Okay. [28:12] >> Yeah, I mean, that's an elite [28:14] >> sleep spot That's an elite spot for a [28:16] nap. [28:16] >> chair was arguably designed for office [28:18] naps. No one's getting like real work [28:20] done [28:21] >> I think so. [28:23] I think so. So, [28:25] uh yeah, I I this one's tough because [28:27] it's easy to just jump right on in [28:29] Matthew Prince's camp. I'm in Vinod [28:31] Khosla's camp. Also, Vinod didn't fall [28:33] asleep, so it's all moot, [28:35] [clears throat] but [28:36] we know we know the secret. It's the air [28:39] horn. Bring the air horn. No one's [28:40] falling asleep. Problem solved. [28:42] >> Tyler [28:44] had a good story [28:46] when when they were building Divvy they [28:48] pitched Rajeev Misra in the SoftBank [28:51] Redwood City HQ. [28:53] And then Mas in Tokyo soon after it was [28:55] absolute cinema. Poppin Zins smoking a [28:58] vape loud coughing to try to throw us [29:02] off. [29:02] >> Wait, what? Intentional coughing? [29:04] >> Assistants whispering in Rajeev's ear. [29:07] >> That's a power play. [29:08] >> That's that's a power move. [29:09] >> You got to respect that for sure. [29:11] >> See, that's a [clears throat] post [29:11] singularity job. [29:13] >> Yeah, for sure. Come in, whisper into [29:15] the person's ear. It's good. [29:17] >> Yeah. [29:17] >> Um some of the most fascinating [29:19] questions ever asked in Tokyo. Mas [29:21] starts the meeting with you have 10 [29:22] minutes. We flew like 20 hours. [29:24] >> See, that's just a great way to get to [29:25] the [29:29] I think that's just a great way to get [29:31] really get to the meat right quickly. [29:33] Yeah. [29:34] >> Thank you for tuning in. Thank you for [29:35] watching TVP and today. We'll see you [29:37] tomorrow at 11:00 a.m. sharp. Give us [29:39] five stars on Apple Podcast and Spotify. [29:41] Sign up for our newsletter at tvp.com. [29:42] Goodbye.