A Critical Development for Tesla & SpaceXAI ⚡️

以下是Electrified播客节目的内容摘要，包括所有提及的新闻： **1. 自动驾驶出租车服务尚未准备好进行规模化：** * 主持人迪伦·卢米斯（Dylan Loomis）强调，自动驾驶出租车服务未能规模化，症结在于服务本身尚未准备就绪，而非监管问题，也非埃隆·马斯克为促成合并而进行的操纵。 * **奥斯汀趣闻/轶事：** * **Abhimanyu：** 报告称等待时间为15-25分钟，但车辆实际抵达仍需55分钟。他注意到车队规模有所增加（约有90辆车已贴牌），但质疑实际有多少车辆在运行。预订后也无法编辑上下车地点。 * **Texas Tesla：** 分享了一段自动驾驶出租车绕圈六次、导致行程增加20分钟的视频，原因是车辆无法靠边停车。 * **Ethan McConaugh：** （特斯拉支持者，前实习生）经历了一次无人监督的乘车，车辆陷入无限循环接客状态长达20分钟。他称之为“新的最差接客体验”，但也承认自动驾驶汽车的“底线已经提高”。他还观察到，在过去三个月中，上下车性能“显著退步”，车辆会开过指定地点、反复进行前进/倒车操作，并在街区内绕圈。 * **Spencer & David：** 有类似的绕圈经历。 * **Arion：** 分享了一段自动驾驶出租车陷入停车场循环、需要协助的视频，以及另一辆车进行了40点掉头（最终解决但“不尽理想”）。 * 卢米斯总结称，该服务尚未准备好进行规模化，并且，截至最近，“情况正在朝着负面方向发展”。他预测，新城市（可能包括圣安东尼奥）的服务推出时，只会部署“少数车辆”。 **2. 通过Wi-Fi网络进行网络安全威胁：** * 微软报告称，一个与俄罗斯有关联的组织通过入侵酒店、会议中心和旅行网络中的强制门户（captive portal）Wi-Fi，重定向用户并窃取登录信息/Cookie。 * 主持人强调，网络本身现在已成为一种攻击媒介。 **3. FSD（全自动驾驶）客户体验：** * **负面片段：** AI Driver分享了一段他母亲的特斯拉在停车场直接倒车撞上购物车停放架的视频，强调了保持警惕的必要性。 * **正面亮点：** 一辆特斯拉成功向左避开了一辆白色宝马（该宝马弹离一辆卡车），展现出令人印象深刻的反应性能。 **4. 潜在的SpaceX-特斯拉合并讨论：** * 卢米斯表示自己对“合并疲劳”感到厌倦，但他评论道，埃隆的行动和言论似乎正在“为SpaceX收购特斯拉铺路”。 * 他提到最近SpaceX公开举行的全体员工大会是一个不寻常的举动，可能与此有关。 * **对埃隆薪酬的影响：** * 如果发生“控制权变更”的合并/收购，埃隆的运营里程碑（FSD、自动驾驶出租车、机器人、车辆交付量、调整后的EBITDA）将作废。 * 薪酬批次将根据12个市值阶梯进行衡量，且*仅*基于特斯拉在交易中的隐含估值（而非SpaceX的价值）。 * 董事会可能会重新谈判这些绩效奖励。 * 卢米斯驳斥了关于埃隆通过策划过高的SpaceX出价来“玩弄系统”或“盘剥特斯拉投资者”的说法；他澄清说，运营里程碑的作废并不意味着市值批次会自动获得。 **5. GrokBot与数字擎天柱（Digital Optimus）解释：** * **澄清：** 埃隆·马斯克表示，“数字擎天柱与GrokBot不同”；它是“实时视频人类模拟”。 * **主持人的GrokBot体验 (Cursor Ultra)：** * 每个机器人都有一个专用的、持久的虚拟机，允许任务离线继续运行。 * 机器人可以在没有API/MCP连接器的情况下登录网站，模拟人类交互。 * 机器人可以相互协作。 * 易于设置（10分钟即可完成Gmail、Google日历集成）。 * “教学例程”功能允许在虚拟机上录制操作（例如，登录梦幻足球游戏、查看分数）以实现自动化。 * GrokBot应用程序允许从Mac Studio桌面检索/压缩文件到手机。 * 用户界面友好，非技术人员也能轻松使用。 * **数字擎天柱的差异：** * 将看到计算机屏幕的实时视频（过去5秒），并像人类一样进行交互。 * 在本地AI推理硬件上运行，而非基于云的虚拟机。 * Grok将充当高级协调器，但大部分工作将在本地完成。 * 针对成本、延迟和规模进行了优化。 * 卢米斯认为，数字擎天柱的测试版可能会在今年晚些时候发布，他称之为“最被低估的产品”，在物理擎天柱（Physical Optimus）问世之前，它有潜力带来显著的收入/估值变化。 **6. Grok AI模型性能及未来：** * **加文·贝克（Gavin Baker）的赞扬：** 风险投资家加文·贝克赞扬了SpaceX AI和Cursor，称企业认为Grok是一个“很棒的模型”，并指出Anthropic的不可靠性。他称，即使在降价之后，Grok的表现也对竞争对手（Anthropic、OpenAI）构成了“绝对的帕累托优势”。 * **Grok 4.6：** 性能与Fable 5 Max大致相同，但价格便宜85%（输入成本便宜80%，输出Token便宜88%）。 * **数据呈现的批评：** 卢米斯批评SpaceX AI对Grok 4.6的图表呈现方式，指出即使Grok并非在所有类别中都表现最佳（例如，61.3%对比Fable 5的约65%），所有数据点仍被标记为领先。 * **性能飞跃：** 注意到Grok 4.5到4.6在各项基准测试中性能“令人难以置信的飞跃”。 * **Grok 4.7发布预告：** 埃隆·马斯克宣布Grok 4.7“显著优于4.6”，并应在3-4周内准备就绪。初步训练已完成，目前正在补充“大量的SpaceX公司数据”。埃隆预测，由于其独特的训练语料库，4.7“将超越所有现有模型”，并对实际工程应用而言“非同寻常”。 **7. 其他特斯拉和SpaceX新闻：** * **罗恩·德桑蒂斯（Ron DeSantis）与SunTracks：** 佛罗里达州州长在一个新闻发布会上强调了Waymo和特斯拉的CyberCab，该发布会重点关注今年晚些时候在SunTracks设施测试电动垂直起降（EVTOL）飞行器。 * **飞行汽车：** 埃隆·马斯克确认：“你们将会拥有飞行汽车。” * **Roadster 2.0：** 弗朗兹·冯·霍尔茨豪森（Franz von Holzhausen）最近表示Roadster“即将问世”。卢米斯对此表示怀疑，但建议其在潜在的合并事件中首次亮相，以达到最大的轰动效应。 * **TerraFactory工作岗位：** 埃隆·马斯克澄清，3,000个“高质量工作岗位”*仅*是第一阶段（共10个阶段）的，而第一阶段的成本接近170亿美元。卢米斯预测该项目资本支出将超过1000亿美元。 * **特斯拉App自动驾驶出租车队列：** 特斯拉App更新显示，自动驾驶出租车可能会推出“虚拟预订队列”（带有“需求旺盛”提示），但卢米斯认为这只是短期解决方案，长期来看并不可行。 * **Waymo Ojai车队：** 一张图片显示，亚利桑那州Waymo车场中有近1000辆Ojai车辆，可能占其已部署车队的25%。目前尚不清楚它们将取代捷豹I-Pace还是作为补充。 * **Starlink客户：** SpaceX目前拥有超过1300万Starlink宽带客户。 * **SpaceX计算能力：** SpaceX重申其目标是在明年年底前达到10吉瓦的额定计算能力，按照埃隆此前提及的费率（每瓦30-50美元），这可能产生3000亿至5000亿美元的收入。 * **今日股价表现：** 特斯拉收盘下跌1.59%，报327.51美元；SpaceX（隐含估值）收盘上涨9.65%，报146.15美元。纳斯达克100指数（NDX）上涨0.74%。

Here's a summary of the Electrified podcast episode, including every news item mentioned: **1. Robotaxi Service Is Not Ready to Scale:** * Host Dylan Loomis emphasizes that Robotaxi isn't scaling due to service readiness, not regulations or Elon Musk's manipulation for a merger. * **Austin Anecdotes:** * **Abhimanyu:** Reports 15-25 minute wait times, but rides still take 55 minutes to arrive. Notes an increased fleet size (around 90 vehicles badged) but questions how many are operational. Also couldn't edit pickup/drop-off after booking. * **Texas Tesla:** Shared a video of a Robotaxi looping six times, adding 20 minutes to a drive because it couldn't pull over. * **Ethan McConaugh:** (Pro-Tesla, ex-intern) experienced an unsupervised ride stuck in an infinite pickup loop for 20 minutes. He calls it a "new worst pickup," but acknowledges the "floor has risen" for AVs. He also observed "significant regression" in pickup/drop-off performance over the past three months, with cars passing pins, cycling drive/reverse, and looping blocks. * **Spencer & David:** Had similar looping experiences. * **Arion:** Shared a video of a Robotaxi stuck in a parking loop, requiring support, and another performing a 40-point U-turn (eventually resolved but "not ideal"). * Loomis concludes the service is not ready for scale and, as of recently, things are "moving in a negative direction." He predicts new city launches (possibly San Antonio) will only see a "handful of vehicles." **2. Cybersecurity Threat via Wi-Fi Networks:** * Microsoft reported a Russian-linked organization compromising captive portal Wi-Fi at hotels, conference centers, and travel networks to redirect users and steal logins/cookies. * The host highlights that the network itself is now an attack vector. **3. FSD Customer Experiences:** * **Negative Clip:** AI Driver shared his mother's Tesla backing directly into a shopping cart stall in a parking lot, underscoring the need for vigilance. * **Positive Highlight:** A Tesla vehicle successfully moved to the left to avoid a collision with a white BMW that bounced off a truck, demonstrating impressive reactive performance. **4. Potential SpaceX-Tesla Merger Discussion:** * Loomis expresses "merger fatigue" but comments that Elon's actions and commentary seem to be "paving the way for a SpaceX acquisition of Tesla." * He mentions the recent public SpaceX all-hands meeting as an unusual move, potentially related to this. * **Impact on Elon's Compensation:** * If a "change in control" merger/acquisition occurs, Elon's operational milestones (FSD, robotaxis, bots, vehicle deliveries, adjusted EBITDA) would be discarded. * Compensation tranches would then be measured against 12 market cap ladders, based *only* on Tesla's implied valuation in the transaction (not SpaceX's value). * Boards would likely renegotiate these performance awards. * Loomis dismisses claims that Elon is "gaming the system" or "fleecing Tesla investors" by engineering an overly high SpaceX bid; he clarifies that operational milestones falling off doesn't mean market cap tranches are automatically earned. **5. GrokBot & Digital Optimus Explained:** * **Clarification:** Elon Musk stated that "Digital Optimus is different" from GrokBot; it's "real time video human emulation." * **Host's GrokBot Experience (Cursor Ultra):** * Each bot has a dedicated, persistent virtual machine, allowing tasks to continue offline. * Bots can sign into websites without APIs/MCP connectors, mimicking human interaction. * Bots can collaborate with each other. * Easy setup (10 minutes for Gmail, Google Calendar integration). * "Teach routine" feature allows recording actions on a virtual machine (e.g., logging into fantasy football, checking scores) for automation. * GrokBot app allows retrieving/compressing files from a Mac Studio desktop to a phone. * User-friendly interface, accessible to non-tech-savvy individuals. * **Digital Optimus Differences:** * Will see a real-time video (past 5 seconds) of a computer screen and interact like a human. * Runs locally on AI inference hardware, not cloud-based virtual machines. * Grok will act as the high-level orchestrator, but most work will be local. * Optimized for cost, latency, and scale. * Loomis believes a Digital Optimus beta release might come later this year, calling it the "most slept-on product" with potential for significant revenue/valuation changes before Physical Optimus. **6. Grok AI Model Performance & Future:** * **Gavin Baker's Praise:** VC Gavin Baker praised SpaceX AI and Cursor, stating enterprises find Grok a "great model" and noting Anthropic's unreliability. He called Grok's performance "absolute Pareto dominance" over competitors (Anthropic, OpenAI) even after price cuts. * **Grok 4.6:** Roughly same performance as Fable 5 Max but at an 85% discount (80% cheaper input, 88% cheaper output tokens). * **Data Presentation Critique:** Loomis criticizes SpaceX AI's chart presentation for Grok 4.6, highlighting all data points as leading even when Grok wasn't the top performer in every category (e.g., 61.3% vs. Fable 5's ~65%). * **Performance Jump:** Notes an "incredible jump" in performance from Grok 4.5 to 4.6 across benchmarks. * **Grok 4.7 Announcement:** Elon Musk announced Grok 4.7 is "significantly better than 4.6" and should be ready in 3-4 weeks. Initial training is complete, and it's being supplemented with "massive amount of SpaceX company data." Elon predicts 4.7 "will exceed all current models" and be "something special" for real-world engineering due to its unique training corpus. **7. Other Tesla & SpaceX News:** * **Ron DeSantis & SunTracks:** Florida Governor highlighted Waymo and Tesla's CyberCab at a press conference focused on testing electric vertical takeoff and landing (EVTOL) vehicles at the SunTracks facility, starting later this year. * **Flying Cars:** Elon Musk confirmed, "you will get flying cars." * **Roadster 2.0:** Franz von Holzhausen recently said the Roadster is "coming soon." Loomis expresses skepticism but suggests a debut at a potential merger event for maximum excitement. * **TerraFactory Jobs:** Elon Musk clarified 3,000 "high-quality jobs" are for Phase 1 *alone* (out of 10 phases), with Phase 1 costing nearly $17 billion. Loomis predicts the project will exceed $100 billion in CapEx. * **Tesla App Robotaxi Queue:** Tesla App Updates suggests a "virtual booking queue" for Robotaxi ("high demand" messaging) might be coming, but Loomis views it as a short-term solution, not viable long-term. * **Waymo Ojai Fleet:** An image shows nearly 1,000 Ojai vehicles in a Waymo lot in Arizona, representing potentially 25% of their deployed fleet. It's unclear if they will replace Jaguar I-Paces or be an addition. * **Starlink Customers:** SpaceX now has over 13 million Starlink broadband customers. * **SpaceX Compute Capacity:** SpaceX reiterated its goal of 10 gigawatts of nameplate compute capacity by the end of next year, potentially generating $300-500 billion in revenue at Elon's previously mentioned rates ($30-50 per watt). * **Today's Stock Performance:** Tesla closed down 1.59% at $327.51, while SpaceX (implied) closed up 9.65% at $146.15. The NDX was up 0.74%.