Thoma Bravo Surrenders Medallia to Lenders After $5.1B Wipeout
TBPN
Soaring interest rates and severe sales underperformance forced Thoma Bravo to cede control of Medallia to private credit lenders after erasing $5.1 billion in equity.
## Background
This episode was published by the channel TBPN on YouTube on April 23, 2026. According to the page description, Diet TBPN delivers the best highlights of the daily TBPN live tech talk show hosted by John Coogan and Jordi Hays. The show covers major daily tech, market, and cultural news, drawing a wide audience in Silicon Valley.
## TL;DR
The hosts break down major corporate and technology developments, starting with Thoma Bravo's $5.1 billion equity wipeout in Medallia and foreign loan challenges under China's Belt and Road Initiative. They discuss OpenAI's release of GPT-5.5, a federal task force formed to address Chinese model distillation, and Meta's 10% workforce reduction. The episode concludes with a review of Kevin Kelly's tech predictions, earnings wins for Intel and Tesla, and Bob Iger's return to Thrive Capital.
## Atmosphere & style
The show carries a fast-paced, conversational, and energetic vibe driven by playful banter between the co-hosts. They balance sharp, insightful commentary on complex finance and macro topics with offbeat humor, referencing internet slang, creative show pitches, and personal celebrity encounters. The overall tone feels like a casual, plugged-in chat among Silicon Valley insiders who take tech seriously without taking themselves too seriously.
## Key takeaways
- [00:03](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=3s) Thoma Bravo is handing over Medallia to creditors, suffering a $5.1 billion equity wipeout due to execution issues and $3 billion in debt.
- [00:03](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=3s) Blackstone noted Medallia's downturn was driven by severe sales underperformance rather than direct AI displacement.
- [00:03](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=3s) Zero-interest-rate policies (ZIRP) inflated SaaS valuations, leaving private equity-backed software companies exposed as floating-rate debt ballooned.
- [06:00](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=360s) Analyzing Chinese diplomacy solely through tech competition misses broader geopolitical factors like supply chains, rare earths, and executive talent.
- [07:00](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=420s) OpenAI launched GPT-5.5, featuring enhanced multi-step agentic capabilities while maintaining GPT-5.4 token latency.
- [10:00](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=600s) The US government and frontier AI labs created a joint task force to combat 'Adversarial Distillation' by foreign entities.
- [10:00](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=600s) Foreign labs are using proxy networks and shell companies to exfiltrate proprietary capabilities from US frontier models.
- [13:00](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=780s) Meta announced an 8,000-employee workforce reduction as Big Tech continues corporate restructuring.
- [15:21](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=921s) Founders and tech media are exploring unconventional, entertainment-driven pitch formats to build visibility.
- [17:39](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1059s) Kevin Kelly's 2016 technological predictions have proven remarkably accurate in predicting modern tech trends like AI cognification and usage-based pricing.
- [20:10](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1210s) Bob Iger is returning to venture capital firm Thrive Capital following his tenure as Disney CEO.
- [20:10](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1210s) Intel beat Q1 earnings expectations and raised Q2 guidance behind strong data center performance.
- [20:10](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1210s) Tesla beat Q1 revenue and net income estimates while shifting toward a more realistic tone regarding project timelines.
The hosts kick off the show with bad news in private equity: Thoma Bravo is taking a massive write-down on customer/employee experience management software provider Medallia. After restructuring negotiations broke down, Thoma Bravo is reportedly preparing to hand over the business to its creditors. This represents a staggering $5.1 billion equity wipeout for the firm, which originally acquired Medallia for $6.4 billion back in July 2021. The enterprise software vendor has been struggling under $3 billion of debt held by major private credit lenders including Blackstone, KKR, Apollo Group, and Antares Capital. While initial market chatter pointed toward AI disruption as the culprit, Blackstone’s global head of private credit, Brad Marshall, noted on a recent call that Medallia’s struggles stem from "execution driven issues" rather than AI replacement, highlighting that sales reps were severely underperforming and delivering as little as 20% of their assigned quotas.
The hosts delve into Medallia’s history, noting its journey from a Sequoia Capital-backed startup—where Sequoia invested across multiple rounds, including $35 million in 2012, $50 million in 2014, and leading a $150 million round—to its 2019 NYSE IPO, before Thoma Bravo took it private near the peak of the low-interest-rate software boom in 2021. The discussion touches on how zero-interest-rate policies (ZIRP) inflated SaaS valuations and discounted cash flow models (DCFs), leaving private equity-backed companies vulnerable once floating-rate debt costs ballooned. With private debt lenders marking the company's debt down to 79 cents (FS KKR Capital Corp) and 74 cents (Apollo Debt Solutions) on the dollar, the equity has been completely erased. Despite Thoma Bravo installing a new management team in early 2025, the firm could not execute a turnaround, prompting JPMorgan's Jamie Dimon-style warnings about "cockroaches" lurking in private credit.
- **Medallia's Debt Load:** $3 billion owed across Blackstone, KKR, Apollo, and Antares Capital.
- **Valuation History:** Taken private in July 2021 for $6.4 billion (previously trading around a $5 billion market cap); equity value dropped by $5.1 billion.
- **Sales Quota Struggles:** Anonymous insider reports indicated sales reps were hitting as low as ~20% of quota due to intense competition from competitors like Qualtrics and newer AI startups (e.g., Aru).
---
[00:06](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=6s) Geopolitics, Belt and Road, and Multi-Faceted Chinese Diplomacy
Turning to global macro affairs, the hosts review a Wall Street Journal opinion piece covering Xi Jinping’s attempt to leverage the International Monetary Fund (IMF) to assist with distressed Chinese foreign loans. Commentator Noah Smith mocked the situation online, remarking on the perceived stumbles of the Belt and Road Initiative, prompting a lighthearted warning from the hosts about travelling to China after making such public jabs.
The hosts analyze how broader geopolitics intersects with the tech sector. Referring to observations from tech commentator Tyler Cowen, they emphasize that tech observers often fall into the trap of viewing Chinese diplomacy solely through the narrow lens of semiconductor chips, data centers, and AI competition. In reality, broader trade negotiations encompass a complex web of factors including Apple's manufacturing footprint, rare earth mineral exports, talent mobility (such as key executive moves to Meta), and historical geopolitical dynamics in Taiwan and the Middle East.
---
[00:07](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=7s) OpenAI Drops GPT-5.5 and the Rise of AI "Choppelgangers"
Breaking news interrupts the broadcast: OpenAI has officially introduced **GPT-5.5**, positioning it as a major step forward in AI-driven work and multi-step agentic capabilities. According to OpenAI’s announcement, GPT-5.5 is designed to execute complex multi-step goals, use tools, write and debug code, analyze data, operate software, and manage spreadsheets without sacrificing speed—matching the per-token latency of GPT-5.4. Demonstrations accompanying the release showed the model completing tasks like solving a Rubik's Cube and seamlessly switching between applications.
The discussion pivots humorously to Merriam-Webster adding new slang terms and "wambos" (word combinations) to its dictionary, such as "lorain" (combining "lore" and "explain") and "choppelganger"—defined as a less attractive or inferior version of a person or thing. The hosts joke that open-source models or heavily distilled versions derived from frontier AI models are the "choppelgangers" of official releases like GPT-5.5.
---
[00:10](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=10s) Government Task Force Formed Over "Adversarial Distillation" of US Frontier AI
The conversation turns serious as the hosts detail news from Michael Kratsios regarding a coordinated federal and industry response to model distillation. The US government is establishing a dedicated task force to combat what it deems "Adversarial Distillation" by foreign entities, primarily based in China. American frontier labs like OpenAI and Anthropic are reporting industrial-scale efforts where foreign labs utilize tens of thousands of proxy accounts, shell companies, and jailbreaking techniques to exfiltrate proprietary capabilities and fine-tuning data from models that cost hundreds of millions or billions of dollars to train.
The hosts highlight the positive "white-pill" aspect of American frontier AI labs collaborating to cross-reference traffic anomalies, spot shell entities, and map out coordinated distillation campaigns. Mentioning the recent release of Qwen 3.6—which users remarked performed closely to Opus 4.5 due to distillation—the hosts argue that protecting model weights, tightening Know-Your-Customer (KYC) API protocols, and coordinating security efforts are vital for maintaining the US's technological lead.
- **Threat Vectors:** Use of over tens of thousands of proxy accounts, multi-lab queries via shell organizations, and systematic jailbreaks to harvest model outputs.
- **Economic Impact:** Exploits the massive private R&D capital ($100M+ per training run) invested by US AI leaders.
---
[00:13](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=13s) Meta Layoffs, Corporate Restructuring, and Tech Media Concepts
Closing out the segment, the hosts cover breaking corporate restructuring news across big tech. Meta is initiating a 10% workforce reduction, impacting approximately 8,000 employees while simultaneously canceling 6,000 open roles. This mirrors similar belt-tightening across the industry, such as Microsoft offering early retirement incentives to roughly 7% of its workforce. The hosts debate whether these cuts are driven by AI efficiency gains, over-hiring, or strategic reallocations toward high-priority technical projects.
Finally, the hosts review creative pitch ideas from tech personality Christoph, who argued that tech media needs more entertaining, tech-positive content. Among the proposed show concepts were "Silent Library with Founders" (where entrepreneurs must endure ridiculous stunts without making a sound to win investment capital) and "Jackass with Founders" (involving physical stunts, such as riding high-speed robotic bulls or interacting with humanoid robotics).
<ElicitationsGroup message="Where would you like to go next with this transcript?">
<Elicitation label="Summarize Part 2 of this episode" query="Summarize Part 2 of this TBPN episode in the same detailed, timestamped walkthrough format."/>
<Elicitation label="Analyze the PE impact on SaaS valuations" query="Analyze how high interest rates and private debt restructuring are impacting SaaS valuations like Medallia's."/>
</ElicitationsGroup>
[15:21](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=921s) Unconventional Founder Pitches and Late-Night Flights
The hosts riff on wildly absurd pitch formats for tech founders, debating whether extreme antics like riding a mechanical bull or going ghost hunting could actually work as viable promotional content. The conversation pivots to Jesse Michels, who receives praise for ranking high on tech podcast charts while producing top-tier reporting on aliens, paranormal activity, and conspiracy theories. The hosts joke that founders could get all the PR they need just by making a guest appearance on one of Jesse’s videos. They brief briefly brainstorm pitches given while skydiving, though neither host has actually skydive themselves.
This leads into an unexpected airplane anecdote about being awake at 2:00 AM on a dark, overnight flight back from Europe. Out of the shadows walked Keanu Reeves—the actual actor—looking like John Wick coming down the aisle. The host ended up chatting with Keanu for about five minutes, describing him as a super nice guy before he returned to his seat. The banter wraps up with a quick nod to old party video channels like "I'm Shmacked" and "Supreme Milk," marking a nostalgic generational gap for the hosts.
- **Paranormal Content:** Jesse Michels’ show charts high in technology while covering UFOs, aliens, and conspiracy theories.
- **Pitch Formats:** Ideas include mechanical bull riding, ghost hunting, and skydiving pitches.
- **Celebrity Encounter:** A 2:00 AM flight conversation with Keanu Reeves.
---
[17:39](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1059s) Kevin Kelly’s 12 Inevitable Tech Predictions
The hosts review a post by Jason Schuman highlighting Kevin Kelly's 2016 book, *The Inevitable: Understanding the 12 Technological Forces That Will Shape Our Future*. Written around 2015 by the Wired co-founder, the book outlined 12 trends destined to shape the next 30 years. The hosts praise how remarkably accurate Kelly's forecasts turned out to be, noting how seamlessly his concepts match the modern tech landscape.
As they cycle through the trends, the hosts react to specific predictions:
* **Becoming:** The shift from fixed products to continuous upgrades and subscription models (which has now evolved into pure usage/consumption-based pricing).
* **Cognifying:** Making everything smarter using cheap, cloud-based AI ("Nailed it," the hosts remark).
* **Flowing & Screening:** Real-time streams and turning all surfaces into screens—joking that even your toaster has a display now.
* **Accessing & Remixing:** Moving from asset ownership to access, and unbundling/recombining primitive software components.
* **Tracking & Surveillance:** Total surveillance for consumer benefit sparks a brief debate. One host defends a "bull case" for encrypted, black-box tracking, arguing that if Netflix watches his viewing history privately to recommend the next perfect show, "I'm cool with that surveillance."
---
[20:10](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1210s) Bob Iger, Intel Earnings, and Tesla’s Reality Check
The show turns to rapid-fire tech and business news. They highlight reports of former Disney CEO Bob Iger returning to venture capital firm Thrive Capital, where he previously bought a stake and served as an LP, marking his next act alongside a broader cohort of veteran executives like Reed Hastings shifting roles.
Turning to public markets, Intel reported strong Q1 earnings after the bell, beating analyst expectations on both top and bottom lines and offering solid Q2 revenue guidance of $13.8 billion to $14.8 billion (versus the $13.0 billion Wall Street estimate). Driven by robust data center sales, Intel's performance drew a shoutout to X user "Bubble Boy," who posted two hours prior that he was doubling down with 25% of his portfolio in Intel calls.
Tesla also released its Q1 numbers, posting revenue of $22.4 billion (beating $21.4 billion estimated) and net income of $1.45 billion (beating $1.17 billion estimated). Despite the top and bottom line beats, *The Wall Street Journal* noted a more cautious, grounded tone from Elon Musk, who emphasized the need to "get realistic about some timelines" and anchor expectations back to business fundamentals.
- **Bob Iger:** Returning to Thrive Capital following his tenure at Disney.
- **Intel Q1:** Guidance set at $13.8B–$14.8B for Q2 (vs. $13.0B consensus); trades near a 100x P/E multiplier following data center growth.
- **Tesla Q1:** Reported $22.4B revenue and $1.45B net income, beating consensus across the board.
- **SpaceX Valuation:** Noted that at a $1.75 trillion valuation, SpaceX would surpass Tesla's current market cap of ~$1.1–$1.2 trillion.
The hosts wrap up the episode by asking listeners to leave a five-star rating on Apple Podcasts and Spotify, sign up for the newsletter at TBPN.com, and throw a "flashbang" as they sign off.
> [20:10](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1210s) get realistic about some timelines — Elon Musk
## Mentioned
Thoma Bravo, Medallia, OpenAI, ChatGPT 5.5, Bob Iger, John Coogan, Jordi Hays, The New York Times, Mark Zuckerberg, Sam Altman, Mark Cuban, Satya Nadella, Blackstone, KKR, Apollo Group, Antares Capital, Brad Marshall, Sequoia Capital, FS KKR Capital Corp, Apollo Debt Solutions, JPMorgan, Jamie Dimon, Qualtrics, Aru, Xi Jinping, International Monetary Fund, IMF, Noah Smith, Belt and Road Initiative, Tyler Cowen, Apple, Meta, Merriam-Webster, Michael Kratsios, Anthropic, Qwen 3.6, Opus 4.5, Microsoft, Christoph, Jesse Michels, Keanu Reeves, John Wick, I'm Shmacked, Supreme Milk, Jason Schuman, Kevin Kelly, The Inevitable: Understanding the 12 Technological Forces That Will Shape Our Future, Wired, Netflix, Disney, Thrive Capital, Reed Hastings, Intel, Bubble Boy, Tesla, The Wall Street Journal, Elon Musk, SpaceX
TBPN reports on mounting financial and strategic pressures across the tech sector, highlighted by Thoma Bravo's $5.1 billion equity wipeout in Medallia, federal actions against the adversarial distillation of frontier AI models, and widespread corporate restructuring.
Private Debt Distress and Medallia's Equity Wipeout (0:03)
Thoma Bravo is preparing to hand customer experience software provider Medallia over to its creditors after debt restructuring talks collapsed, marking a $5.1 billion equity wipeout. Thoma Bravo acquired the business for $6.4 billion in July 2021 near the height of zero-interest-rate software valuations. Medallia has been burdened by $3 billion in debt held by private credit lenders including Blackstone, KKR, Apollo Group, and Antares Capital. Secondary valuations reflect this strain, with FS KKR Capital Corp marking the debt to 79 cents and Apollo Debt Solutions marking it to 74 cents on the dollar.
While initial industry commentary questioned whether artificial intelligence disrupted Medallia, Blackstone global head of private credit Brad Marshall stated on an investor call that the distress stemmed from execution issues rather than AI replacement. Anonymous insider reports indicated sales representatives were achieving as little as 20% of their quotas amid stiff competition from Qualtrics and newer AI startups like Aru. The situation highlights vulnerabilities among private-equity-backed software companies facing ballooning floating-rate debt costs.
Frontier AI Releases and Countering Adversarial Distillation
OpenAI announced GPT-5.5, positioning the model as a major step forward for autonomous multi-step work, software control, data analysis, and coding, while matching the per-token latency of GPT-5.4. Alongside new product releases, the program highlighted growing national security and intellectual property concerns surrounding model theft. The U.S. government, advised by Michael Kratsios, is forming a dedicated task force to counter what it characterizes as adversarial distillation by foreign entities, primarily based in China.
American frontier developers including OpenAI and Anthropic reported industrial-scale campaigns where foreign groups deploy tens of thousands of proxy accounts, shell companies, and systematic jailbreaks to extract proprietary capabilities and fine-tuning data from models requiring over $100 million in training capital. Following the release of models like Qwen 3.6—which users observed performed close to Anthropic's Opus 4.5 due to distillation—U.S. labs are collaborating to analyze traffic anomalies and tighten API Know-Your-Customer protocols to protect domestic technological advantages.
Big Tech Headcount Reductions and Workforce Reallocations
Major technology companies continue to scale back headcount as part of ongoing corporate restructuring. Meta is cutting approximately 10% of its workforce, affecting roughly 8,000 employees, alongside the cancellation of 6,000 open job listings. This mirrors similar retrenchment across the industry, including Microsoft offering voluntary early retirement incentives to roughly 7% of its workforce.
The hosts debated whether these job cuts are the direct result of operational efficiency gains unlocked by internal AI tools, an ongoing correction of pandemic-era over-hiring, or strategic maneuvers to free up operating capital for high-priority technical infrastructure.
Executive Rotations and Divergent Big Tech Earnings (20:10)
In executive news, former Disney CEO Bob Iger is returning to venture capital firm Thrive Capital, where he previously held a stake and acted as a limited partner. Meanwhile, public market earnings yielded strong quarterly beats in semiconductor hardware and electric vehicles. Intel reported Q1 results that topped expectations on data center strength, issuing Q2 revenue guidance between $13.8 billion and $14.8 billion, ahead of Wall Street consensus of $13.0 billion.
Tesla also surpassed consensus estimates, posting Q1 revenue of $22.4 billion against an estimated $21.4 billion and net income of $1.45 billion against an expected $1.17 billion. Despite the financial beats, The Wall Street Journal noted that CEO Elon Musk struck a more reserved tone, urging a return to realistic timelines and core operational fundamentals.
While strong data center and hardware earnings demonstrate sustained demand for computing infrastructure, Medallia's restructuring and ongoing workforce contractions at Meta and Microsoft reveal that enterprise software and tech incumbents face heightened scrutiny over capital allocation, debt burdens, and execution.
## Background
本期视频由科技脱口秀频道 TBPN（主持人为 John Coogan 与 Jordi Hays）于 2026 年 4 月 23 日发布在 YouTube 平台。根据页面简介，这是精选自 TBPN 平日直播精华的 Diet TBPN 节目，聚焦硅谷科技界最新爆发的商业灾难、突发技术发布与行业前沿动态。
## TL;DR
节目首先聚焦顶级私募机构 Thoma Bravo 在 Medallia 投资上的溃败，因高杠杆债务与 AI 冲击导致重组破裂，其 51 亿美元股权彻底归零并移交债权人。随后，OpenAI 突发发布专为 Agent 和复杂自动化设计的 GPT-5.5，引发两位主持人对模型蒸馏攻击及跨实验室联防机制的热议。此外，节目还盘点了 Meta 与微软的新一轮裁员潮、鲍勃·艾格加盟 Thrive Capital 的人事动向，并结合凯文·凯利的技术预言，深入解析了英特尔与特斯拉最新财报释放出的产业信号。
## Atmosphere & style
整期节目节奏明快、插科打诨，洋溢着硅谷内行特有的辛辣幽默与自嘲氛围。两位主持人配合极具默契，不仅用滑稽音效和脱口秀般的犀利吐槽化解了私募暴雷与巨头裁员的沉重感，更在突发见证 GPT-5.5 官宣时爆发出极客专属的技术兴奋。从深夜航班偶遇基努·里维斯的奇趣闲聊，到对地缘债务与大厂财报的冷峻研判，整个对谈既有圈内沙龙的松弛感，又保持着极高密度的商业与技术信息量。
## Bulletin
- [00:03](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=3s) 私募巨头 Thoma Bravo 在 Medallia 的重组谈判破裂，准备向黑石、KKR 等债权人移交控制权，导致 2021 年私有化投入的 51 亿美元股权几乎彻底清零。
- [07:48](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=468s) OpenAI 突发发布 GPT-5.5，主打面向复杂 Agent 自动化工作流与端到端任务执行，并在性能大幅提升的同时保持了与 GPT-5.4 相当的单 Token 低延迟。
- [10:11](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=611s) 美国政府联手 Anthropic、OpenAI 等前沿实验室设立专项工作组，打击通过数万海外代理账号和越狱手段进行的工业级模型蒸馏窃密行为。
- [13:22](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=802s) 大厂组织架构持续瘦身，Meta 确认裁减 8,000 人并冻结 6,000 个在招岗位，微软则向 7% 员工提供提前退休方案以精简团队。
- [20:38](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1238s) 英特尔第一季度业绩与第二季度指引双双超出华尔街预期，受数据中心强劲需求提振其 P/E 估值接近 100 倍；特斯拉虽 Q1 超预期，但马斯克在电话会上对交付时间表展现出罕见的谨慎克制。
## Key takeaways
- [00:03](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=3s) Thoma Bravo 无法承受浮动利率飙升带来的 30 亿美元债务成本，Medallia 股权价值归零坐实了高杠杆 SaaS 收购案在加息周期的彻底崩塌。
- [03:50](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=230s) 传统体验管理软件面临一线销售崩溃与 AI 原生初创公司（如 Aaru）降维打击的双重困境，单纯依靠 PE 职业经理人难以抵御颠覆式创新。
- [06:15](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=375s) 科技地缘政治博弈不应局限于芯片和算力基础设施，中国一带一路不良海外债务的 IMF 处置方案、稀土管治与跨国供应链同样是深层筹码。
- [07:48](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=468s) OpenAI 正式发布 GPT-5.5，聚焦自主工具调用、多步验证与复杂现实工作流，首发即集成进 ChatGPT 与 Codex。
- [09:11](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=551s) 行业将过度蒸馏、阉割严重的开源翻版大模型戏称为“Choppelganger”，提醒开发者警惕廉价开源模型中隐藏的退化与缺陷。
- [10:11](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=611s) 前沿模型实验室开始共享异常流量情报，应对有组织的黑客利用大量代理空壳账户成体系抽取美方前沿 AI 能力的蒸馏攻击。
- [13:22](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=802s) 科技巨头在缩减数万常规编制与边缘业务的同时疯狂抢购算力芯片与核心 AI 人才，催生一边裁员一边押注的资本再分配周期。
- [14:07](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=847s) 科技圈公关内容亟需打破严肃沉闷，引入极端特技、恶搞竞技等娱乐化形式能为创业者带来更高的曝光收益。
- [17:39](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1059s) 重读凯文·凯利 2016 年著作《必然》，其关于从买断转向云服务（Becoming）以及 AI 智能赋能一切（Cognifying）的预测已被产业完全印证。
- [18:33](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1113s) 在端到端加密前提下，服务于用户自身的无感监控与深度偏好过滤能够大幅提升个性化体验，正在重塑数字消费内容生态。
- [20:09](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1209s) 迪士尼功勋 CEO 鲍勃·艾格加盟 Thrive Capital，标志着顶尖商业领袖退居二线后将风投作为职业下一阶段的新趋势。
- [20:38](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1238s) 英特尔受益于数据中心芯片反弹给出强劲指引推动股价拉升，而特斯拉马斯克则收紧预期表态，呼吁投资者回归基本面与现实交付节奏。
## Detailed walkthrough
### [00:03](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=3s) Thoma Bravo 将 Medallia 移交债权人：51 亿美元股权清零
节目录制现场一阵哄笑，主持人打趣着赶紧把悲报音效准备好，因为刚刚传来重磅坏消息：顶级私募股权机构 Thoma Bravo 在知名企业软件公司 Medallia 上遭遇惨烈减记。在长达数月的债务重组谈判最终破裂后，Thoma Bravo 据悉正准备将 Medallia 的控制权彻底移交给债权人。对于在 2021 年以 64 亿美元高位私有化该公司的 Thoma Bravo 而言，这意味着高达 51 亿美元的股权投资几乎彻底灰飞烟灭。
根据路透社发布的独家调查报道，Medallia 近几个月来彻底被高达 30 亿美元的沉重债务压垮，其主要债权方包括黑石集团（Blackstone）、KKR、阿波罗全球管理（Apollo Group）以及 Antares Capital。Medallia 的核心业务是向大企业提供用于收集、分析客户与员工反馈的体验管理软件，在行业内的主要竞争对手是 Qualtrics。然而随着生成式 AI 的爆发，红杉资本等知名 VC 支持的新一代 AI 原生初创公司正以更低的成本和更高的自动化程度蚕食该领域，导致市场对传统 SaaS 的估值逻辑产生了根本性动摇。
回顾 Medallia 的资本运作史，红杉资本曾是最大的赢家之一。红杉在 2012 年为 Medallia 领投了 3500 万美元，2014 年追加 5000 万美元，随后又主导了一轮 1.5 亿美元的巨额融资。Medallia 于 2019 年在纽交所敲钟上市，彼时估值在 50 亿美元上下浮动；到了 2021 年 7 月超低利率盛行的疫情狂热期，Thoma Bravo 以 64 亿美元将其溢价私有化。然而，当时基于 3% 极低利率制定的折现现金流（DCF）估值模型，在美联储激进加息后彻底失效。浮动利率导致债务利息支出成倍膨胀，而公司的现金流却未能跟上，最终将股权推向深渊。目前在二级与非公开市场中，FS KKR Capital 对其持有的债务估值仅为面值的 79 美分，Apollo Debt Solutions 更是将其标记为 74 美分。在偿还顺位上债权远先于股权，债权折价已然坐实了股权价值彻底归零的结局。
```mermaid
flowchart TD
accTitle: Medallia 资本结构崩溃与控制权转移流程
accDescr: 展示 Thoma Bravo 自 2021 年杠杆收购 Medallia 至浮动利率飙升、业绩崩塌、最终股权归零移交债权人的全过程。
subgraph Acquisition ["2021年 杠杆收购"]
TB["Thoma Bravo 出资<br/>51 亿美元股权"]
Debt["债权财团提供<br/>30 亿美元杠杆借款<br/>(Blackstone / KKR / Apollo)"]
Target["以 64 亿美元估值<br/>将 Medallia 私有化"]
TB --> Target
Debt --> Target
end
subgraph Crisis ["宏观转向与经营恶化"]
Macro["美联储加息<br/>浮动利率利息剧增"]
Execution["销售疲软与 AI 冲击<br/>仅达标 20% 配额"]
Target --> Macro
Target --> Execution
end
subgraph Liquidation ["重组破裂与控制权易主"]
DebtCut["债务大幅折价<br/>(标价 74~79 美分/美元)"]
EquityWipe["51 亿美元股权清零<br/>(Thoma Bravo 出局)"]
LenderTakeover["债权人全面接管控制权"]
Macro & Execution --> DebtCut
DebtCut --> EquityWipe
DebtCut --> LenderTakeover
end
```
- **收购与清零金额**：2021 年 Thoma Bravo 以 64 亿美元收购 Medallia，此次违约重组直接导致其 51 亿美元股权化为乌有。
- **债权人敞口**：30 亿美元债务由黑石、KKR、Apollo 和 Antares Capital 共同持有，目前债务在账面上仅值面值的 74% 至 79%。
- **红杉历史战绩**：红杉资本先后在 2012 年注资 3500 万美元、2014 年注资 5000 万美元并主导 1.5 亿美元融资，成功在公开上市与后续退潮前获利了结。
---
### [03:50](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=230s) 业绩不佳还是 AI 冲击？软件业寒冬与私有化困局
黑石集团全球私人信贷主管布拉德·马歇尔（Brad Marshall）此前在 2 月份的电话会议中曾对这笔坏账作出辩解，坚称 Medallia 陷入困境“与 AI 没有任何关系，而是源于我们认为的执行力问题”。然而现场主持人指出，来自一线的匿名销售人员爆料显示，公司的销售代表们在市场上被各路对手打得落花流水，多数人甚至只能完成 20% 左右的业绩配额。尽管 Thoma Bravo 在 2025 年初撤换了管理层并试图执行复兴计划，但高昂的杠杆利息与崩塌的现金流根本无法支撑重组周期的消耗。
这一暴雷事件在社交平台 X 上引发了投资者的普遍恐慌与嘲弄。有知名推主讽刺道：“是的，我非常确信这就是软件行业的底部了，未来绝不可能更糟了。” 主持人将 Medallia 比作摩根大通 CEO 杰米·戴蒙（Jamie Dimon）曾警告过的“蟑螂理论”——当你看到一只蟑螂（如早前爆雷的 First Brands 等破产事件）浮出水面时，暗处必然藏着成群的隐患。最讽刺的是，过去一年里身涉其中的各家机构都在巡回公关，坚称“一切安好”、“AI 会是 SaaS 的加速器”，但现实却无情地揭穿了这层虚饰。
深入探讨这一轮企业软件困境的根源，主持人们认为其致命弱点在于缺乏“创始人驱动”的狼性与敏锐度。在被私募股权控股后，职业经理人主导的公司往往只会固守传统的 CRUD 模式和存量客户，根本无法抵御新一代颠覆者的降维打击。以往做一个平庸的 Qualtrics 或 Medallia 替代品很难融资，但如今像 Aaru 这样利用 AI 模拟（Simulation）技术的颠覆型公司，以及新一批 Y Combinator 孵化的 AI 代理初创团队，正以前所未见的产品形态快速瓜分这块原本由老牌巨头垄断的蛋糕。
- **销售端执行崩溃**：内部一线销售人员普遍反映业绩严重受阻，配额达成率惨跌至 20% 上下。
- **私有化弊端凸显**：PE 控股模式下缺乏创始人领导力，叠加高额财务杠杆，使得传统软件面对外部敏捷竞争与 AI 转型时步履维艰。
- **新型颠覆者入局**：以 Aaru 为代表的 AI 模拟初创企业及 YC 孵化的新生代产品正在彻底改写客户体验与反馈分析行业的技术范式。
---
### [06:15](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=375s) 地缘政治与全球债务：一带一路救援与科技外交的更广博弈
话题从硅谷的企业债务转移到国际宏观债务的博弈。《华尔街日报》发表了一篇引人注目的观点文章，披露中国高层希望利用国际货币基金组织（IMF）来协助救助其在全球范围内面临违约风险的一带一路海外不良贷款。知名博主诺亚·史密斯（Noah Smith）在社交媒体上幸灾乐祸地嘲讽道：“笑死，一带一路彻底搞砸了，最高层真是无能。” 主持人笑着打趣调侃道：“诺亚，你近期可千万别去中国。”
针对这一国际金融风波，主持人援引泰勒·考恩（Tyler Cowen）的常年观点展开剖析：科技圈在谈论中美外交和大国博弈时，往往有极其狭隘的“芯片与 AI 隧道视野”，仿佛一切只关乎先进制程半导体和智算中心。但实际上，真正的全球外交是一场涵盖多维筹码的超级博弈，从一带一路的跨国金融杠杆、中东地缘利益，到稀土管制、苹果供应链动向，乃至海外科技顶尖人才流动。这笔巨额海外不良贷款的处置，极有可能成为未来中美及全球大国谈判中除芯片禁令与稀土博弈之外的又一张关键交换筹码。
- **债务风险向多边组织传导**：外媒报道称中国试图借助 IMF 机制化解分散在发展中国家的一带一路违约资产风险。
- **科技地缘视角的盲区**：硅谷往往过度聚焦于算力芯片与 AI 基础设施，忽视了稀土、国际债务重组、跨国制造业供应链与人才流动等交织在地缘政治深处的全盘筹码。
---
### [07:48](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=468s) OpenAI 重磅发布 GPT-5.5：面向 Agent 与复杂工作的智能跃进
演播室突然被突发的科技头条打断：OpenAI 毫无征兆地正式发布了 GPT-5.5！主持人当即现场宣读官方公告。OpenAI 将 GPT-5.5 定义为“面向实际工作并为 AI Agent 提供动力的新一代智能等级”，其核心定位是理解极其复杂的目标、自主调用多套工具、自我检查工作结果，并端到端地将繁杂任务执行到底。该模型代表了计算机工作自动化流程的全新飞跃，目前已在 ChatGPT 与 Codex 中全面上线。
官方随之释出了一段极具冲击力的解复原魔方演示视频。据技术说明，GPT-5.5 在代码编写与调试、在线全网调研、复杂数据深度分析、自动生成文档与专业电子表格、操作系统软件，以及在多款外部工具间无缝穿梭协作方面均展现了巨大飞跃。令人惊叹的是，在实现智商大幅跨越的同时，GPT-5.5 并没有牺牲推理速度，在实际服务部署中做到了与 GPT-5.4 相当的单 Token 延迟水平。主持人由衷感叹 OpenAI 团队这次确实是“火力全开、大显身手”（The team cooked），并向所有参与该项目的工程师表示祝贺。
节目最后，话题以轻松的语调跳向韦氏词典（Merriam-Webster）的新增词条。主持人提到现在的网络流行语更迭速度快到不可思议，韦氏词典的“俚语与趋势”编辑们显然在超负荷加班，不断收录各种全新的复合词（wambos / word combos），反映了当下文化与技术快速迭代带来的语言膨胀现象。
- **GPT-5.5 核心定位**：专为 Agent 设计，擅长复杂任务端到端交付、自主跨工具调用与自我修正纠错。
- **性能与速度表现**：在代码调试、表格制作与系统操作性能激增的前提下，其 Serving 时的单 Token 延迟依然维持在 GPT-5.4 的高速水平。
- **发布平台**：首发即同步上线 ChatGPT 与 Codex 平台供用户直接使用。
### [09:11](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=551s) GPT-5.5 的背景演进与“劣质克隆版”Choppelganger
节目开场，两位主持人围绕 GPT-5.5 的技术背景展开闲聊。如果给听众完整梳理 GPT 的前世今生，必须从早期的大模型 InstructGPT（Text-Davinci）、GPT-3.5，一路聊到 GPT-4.0 与 GPT-4.5，进而解析 GPT-5.5 究竟具备何种全新能力。不过，他们话锋一转，将视线聚焦到《韦氏词典》（Merriam-Webster）最近收录的一个新词——“Choppelganger”（由 chopped 与 doppelganger 组合而成，意指某人或某物较差、较丑陋的翻版）。他们调侃称，在 AI 领域，凡是有人利用大模型蒸馏技术把 GPT-5.5 压缩成一个缩水低配版本的，那个产物就是官方 GPT-5.5 的“Choppelganger”。
主持人幽默地告诫听众：除非你眼下实在走投无路、急需一个便宜甚至免费的开源模型来应急，否则千万不要碰这些翻版模型；在使用某些宣称性能接近前沿模型的开源模型时，务必保持“买者自负”（buyer beware）的心态，因为其底层很可能潜藏着大量未修复的缺陷与偏见。
- **Choppelganger 概念**：韦氏词典收录的新造词，被主持人用来指代对尖端闭源前沿模型进行模型蒸馏（Distillation）后得到的廉价缩水版模型。
- **开源模型防坑提示**：盲目依赖经过蒸馏提取的开源模型可能带来意料之外的模型退化与安全缺陷。
### [10:11](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=611s) 美国组建防蒸馏专项工作组与跨实验室联合防御
针对日益猖獗的模型蒸馏窃密行为，前白宫科技政策办公室官员迈克尔·克拉齐奥斯（Michael Kratsios）透露出最新动向：美国政府高度重视 Anthropic 与 OpenAI 等实验室关于“模型蒸馏攻击”的警告，并正式设立专项工作组，全力攻克如何在大规模层面上遏制对抗性蒸馏。主持人指出，随着前沿模型的训练成本飙升至数亿美元甚至数十亿美元级别，如果海外团队只需通过微小成本就能蒸馏提取权重、非法渗出数据并复制核心能力，这将造成无法估量的经济与战略损失。
令人欣慰的是，各大前沿 AI 实验室之间已经开始建立联防机制。单个实验室可能只会检测到零星的异常流量，误以为只是某个普通客户，但当多家机构协同对账时，才发现同一批由空壳公司操纵的数万个代理账户正分别向不同平台发起有针对性的探测提示词。如果将各家的碎片数据拼凑起来，攻击者便能在暗中完整绘制出整个前沿系统的能力图谱。节目中朗读了官方关于对抗性蒸馏的备忘录摘要：
> “美国在人工智能技术上的领先地位，源于数十年基础研究、大胆的创业冒险以及每年数千亿美元的私营部门投资……然而，美国掌握的信息表明，主要位于境外的实体正发起有组织的工业级行动，通过数万个代理账户规避检测并利用越狱手段，成体系地蒸馏抽取美国前沿 AI 模型的核心能力，借此剥削美国的专业知识与创新成果。”
```mermaid
flowchart TD
accTitle: 对抗性模型蒸馏攻击与实验室协同防御机制
accDescr: 展示海外代理账户分散提取不同前沿实验室数据，以及实验室跨机构联防和政府工作组的应对流程。
subgraph Attack["工业级对抗性蒸馏攻击"]
Proxy["数万个海外代理与空壳账户"]
Jailbreak["越狱探测与提示词窃密"]
Proxy --> Jailbreak
end
subgraph Defense["美方前沿实验室协同防御"]
LabA["OpenAI 接收 1/3 碎片请求"]
LabB["Anthropic 接收局部异常调用"]
LabC["其他前沿厂商数据拼图"]
DataShare["跨实验室威胁情报共享"]
Jailbreak --> LabA
Jailbreak --> LabB
Jailbreak --> LabC
LabA --> DataShare
LabB --> DataShare
LabC --> DataShare
end
subgraph Response["政府与行业应对机制"]
TaskForce["政府防蒸馏专项工作组"]
Security["强化 API 访问控制与严格 KYC"]
DataShare --> TaskForce
TaskForce --> Security
end
```
谈及前一天刚发布的通义千问 Qwen 3.6，外界惊叹其综合水平“几乎媲美 Claude Opus 4.5”，主持人相视一笑调侃道：“那是因为他们对顶级模型进行了高强度蒸馏。”目前，各大实验室正大幅收紧 API 权限管理、推行更加严格的“了解你的客户”（KYC）审核，以防范核心资产被系统性外流。
### [13:22](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=802s) 科技巨头裁员潮：Meta 裁员万人与微软精简组织
节目播报了科技巨头的最新裁员突发新闻。Meta 确认将裁减约 10% 的现有员工（约 8,000 人），同时冻结取消约 6,000 个在招岗位，总计削减体量高达 1.4 万人。与此同时，微软也传出为大约 7% 的员工提供提前退休激励方案，试图以此实现组织架构的“瘦身”与扁平化。
主持人分析认为，关于大厂裁员动机的争论仍将持续：一方面是过去几年的过度扩张导致机构臃肿；另一方面则是 AI 辅助编程与自动化带来的生产力跃升，使得企业能够在缩减人手的同时维持产能。更核心的因素在于资金再分配——大型科技公司需要将算力芯片与工程资源从边缘项目中腾挪出来，集中押注在以 AI 为代表的核心业务上，这也造就了“一边大规模优化减员、一边在核心领域狂热招人”的奇特现象。
- **Meta 人员调整**：直接裁撤 8,000 名现有员工，同步削减 6,000 个开放职位。
- **微软组织瘦身**：面向约 7% 的员工群体推行提前退休方案。
- **战略重心迁移**：资金与硬件算力正加速向关键 AI 基础设施倾斜。
### [14:07](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=847s) 科技圈专属整活真人秀脑洞与基努·里维斯深夜偶遇
在谈到科技行业公关时，主持人引用了风投人克里斯托夫（Kristoff）的观点——目前的科技圈媒体内容普遍严肃乏味，亟需更多积极正面且兼具娱乐性的创新传播形式。两人顺着这份整活清单逐一展开推演：
1. **创始人版《沉默图书馆》（Silent Library with Founders）**：模仿经典整蛊游戏秀，参赛的创业者必须在图书馆保持绝对安静，一人承受各种荒诞恶搞或心理惩罚，全员忍住不笑的人才能拿下投资意向书。
2. **创始人版《蠢蛋搞怪秀》（Jackass with Founders）**：让创始人们挑战极限特技，比如骑乘机械公牛或人形机器人对抗。主持人笑着吐槽这风险太大：“一个创始人为了推销产品上去摔断胳膊，接下来几个月都写不了代码，那可就得不偿失了。”
3. **边抓鬼边专访热门创始人（Ghost Hunting）**：主持人提到了著名博主杰西·迈克尔斯（Jesse Michels），称赞他将硬核风险投资背景与地外文明、超自然现象调研融为一体，内容长期在科技播客榜名列前茅。如果哪个创业者能客串他的探险视频，PR 效果必定拉满。
4. **高空跳伞路演（Pitching while Skydiving）**：创始人一边从万米高空自由落体一边向投资人讲解商业计划书。
聊到高空飞行，乔迪（Jordi）顺势分享了一段真实的私藏奇遇：某次他搭乘从欧洲飞回美国的红眼航班，时差导致他在凌晨两点依然毫无睡意；机舱昏暗的过道深处，一个修长的身影缓缓走来，竟然是“约翰·威克”（John Wick）的扮演者基努·里维斯（Keanu Reeves）本人。两人在机舱走道里小声攀谈了整整五分钟，基努为人极其谦逊随和，随后才返座休息。
此外，他们还提及了拍摄旧金山科技圈狂欢派对的整蛊短片（致敬当年风靡全美的大学派对实录“I'm Shmacked”），展现了不同圈层亚文化相互碰撞的奇妙趣味。
### [17:39](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1059s) 凯文·凯利神准预言：重读 2016 年《必然》中的技术大势
节目的最后部分将目光投向了《连线》（Wired）杂志创刊主编凯文·凯利（Kevin Kelly）。他在 2016 年出版的经典著作《必然》（*The Inevitable: Understanding the 12 Technological Forces That Will Shape Our Future*）中，预言了塑造未来 30 年的 12 大数字化技术趋势。回顾过去十年的发展轨迹，不得不惊叹于凯文·凯利早在 2015 年动笔时的超前眼光。
书中前两项核心趋势已在当下的科技产业中得到完美印证：
- **形成中（Becoming）**：商业模式彻底告别传统的固定买断制软件，全面转向“永远在更新”的在线服务与订阅模式；甚至连单纯的软件坐席（Seat-based）模式都在逐步瓦解，全面被按用量计费（Consumption-based）的基础设施模型所取代。
- **知化（Cognifying）**：通过云端获取廉价且强大的 AI 能力，让世界上的一切日常事物与工业软件都变得更加智能。两位主持人感叹，这一条在 ChatGPT 席卷全球的今天可谓是“完全命中”（Nailed it），完美预示了云端大模型 API 赋能千行百业的产业格局。
### [18:33](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1113s) 凯文·凯利的技术预言与“良性全面监控”之辩
两位主持人回顾了《连线》（Wired）杂志创始主编凯文·凯利（Kevin Kelly）过去的诸多前瞻性技术预言。凯利一生都深扎在科技前沿，他所描绘的未来图景如今大量成为现实：比如无休止的实时信息流主宰一切，以及“所有表面都将变成屏幕”——如今连家里的吐司烤面包机都能播放视频。此外，人类社会正在经历从“拥有资产”到“随时随地获取服务”的深刻转变。不过在凯利构想的大规模协作方面，主持人笑称如果按十进制的“共享指数”来打分，人类目前大概只走到了“2分”的阶段。
关于深度个性化与内容重组，凯利关于“过滤与极致个性化以提前预测欲望”的论断正在上演，而将现有产品拆解为最原始要素再进行无限排列组合的“混剪/重组”（Remixing），也已经在各种数字内容中彻底爆发，主持人甚至一本正经地拿“椰子模拟器”（Coconut Simulator）作为典型例证。在谈及“为了公民与消费者的利益而实行全方位追踪与全面监控”时，节目展开了一场幽默的争论。当被质疑“难道全面监控还能是好事”时，约翰·库根（John Coogan）给出了自己的“监控看多论”（Surveillance Bull Case）：如果存在一个端到端加密的黑盒子记录下他的所有 Netflix 观看习惯，没有任何 Netflix 员工能窥探隐私，但系统却能凭借这些数据精准推荐他真正喜爱的下一部大作，那这种“服务于约翰·库根本人”的监控就是完全可以接受的。至于凯利提出的“构建连接全人类与机器的全球行星矩阵”，主持人则认为大家还在等待其真正成真。知名投资人杰森·舒曼（Jason Shuman）在社交媒体上也惊叹于这些多年前预言的极高准确度。
- **屏幕无处不在**：从手机、手表扩展到家用电器，所有物理表面屏幕化。
- **所有权转向使用权**：即时访问与订阅制彻底颠覆传统的资产私有观念。
- **重组与解构**：产品要素被拆解至最原子形态并重新杂交组合。
- **提问的价值**：提出一个好问题远比获得现成答案更加珍贵。
---
### [20:09](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1209s) 鲍勃·艾格加盟 Thrive Capital 与巨头掌门人的“下一幕”
话题随后转向《华尔街日报》披露的重磅人事变动。乔迪·海斯（Jordi Hays）原本下意识猜测迪士尼传奇 CEO 鲍勃·艾格（Bob Iger）又要“重返迪士尼”，但约翰公布了出人意料的答案：艾格实际上是重返知名风险投资机构 Thrive Capital。事实上，艾格此前不仅长期担任 Thrive 的有限合伙人（LP），甚至还出资购买过 Thrive 的部分股权。
主持人感叹，对于功成名就的商业领袖而言，进军顶级风投是一段极佳的“职业生涯下一幕”。如今科技与传媒界正涌现出一批退居二线但并未退休的“元老校友群”——比如从 Netflix 卸任 CEO 的里德·哈斯廷斯（Reed Hastings）等人也都在寻找全新的发力点。外界非常期待看到这些昔日的行业巨擘如何利用顶级风投平台继续操盘前沿科技投资。
---
### [20:38](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1238s) 英特尔与特斯拉财报：数据中心强劲反弹与马斯克的保守转向
节目的最后部分聚焦于最新披露的重磅科技股财报。英特尔（Intel）在周四盘后公布了第一季度业绩，营收与利润双双超出华尔街分析师预期，并且由于数据中心芯片业务的强劲需求，英特尔给出了超预期的第二季度业绩指引：
```mermaid
flowchart TD
accTitle: Intel and Tesla Q1 Performance Overview
accDescr: Visual breakdown of Intel's strong guidance and Tesla's revenue beat paired with cautious forward outlook.
subgraph Intel["英特尔 (Intel) Q1 表现与指引"]
I1["Q1 营收与净利双双超预期"] --> I2["数据中心芯片销售势头强劲"]
I2 --> I3["Q2 营收指引: 138亿-148亿美元<br/>(华尔街预期 130亿美元)"]
I3 --> I4["市盈率 (P/E) 达到约 100 倍"]
end
subgraph Tesla["特斯拉 (Tesla) Q1 财务与预期调整"]
T1["Q1 营收: 224 亿美元<br/>(预期 214 亿美元)"] --> T3["业绩全面超预期"]
T2["Q1 净利润: 14.5 亿美元<br/>(预期 11.7 亿美元)"] --> T3
T3 --> T4["马斯克表态转趋谨慎<br/>呼吁对交付时间表保持现实"]
end
```
英特尔预计第二季度营收将在 138 亿至 148 亿美元之间，大幅高出华尔街此前预期的 130 亿美元，推动其市盈率（P/E）攀升至接近 100 倍的高位。社交媒体 X 上的知名散户投资者“Bubble Boy”甚至在财报公布前两小时发文透露，自己将 25% 的投资组合重仓押注了英特尔的看涨期权（Calls），在财报大涨后获得了主持人的连连祝贺。
与此同时，特斯拉（Tesla）也发布了第一季度财报，当季斩获 224 亿美元营收（超出市场预期的 214 亿美元），净利润录得 14.5 亿美元（超出预期的 11.7 亿美元）。然而，《华尔街日报》注意到埃隆·马斯克（Elon Musk）在财报电话会上的基调发生了显著变化，表现得极为克制与谨慎。马斯克坦言“我们需要对某些时间表保持现实态度”，不再像以往那样极力调动市场情绪，而是试图引导投资者重新聚焦公司的基本面。主持人指出，当前资本市场正在见证一个奇特的时间线：如果 SpaceX 估值如预期达到 1.75 万亿美元，它的体量将超越市值在 1.1 万亿至 1.2 万亿美元徘徊的特斯拉。
在节目尾声，两位主持人呼吁听众在 Apple Podcasts 和 Spotify 上打出五星好评，并前往官网 tbpn.com 订阅独家时事通讯。随着约翰喊出招牌口号“扔出闪光弹吧，乔迪”，两位主持人向观众道谢告别，结束了本期节目。
## Notable quotes
> [03:50](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=230s) 与 AI 没有任何关系，而是源于我们认为的执行力问题 — Brad Marshall
> [03:50](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=230s) 是的，我非常确信这就是软件行业的底部了，未来绝不可能更糟了。
> [06:15](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=375s) 笑死，一带一路彻底搞砸了，最高层真是无能。 — Noah Smith
> [06:15](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=375s) 诺亚，你近期可千万别去中国。 — 主持人
> [07:48](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=468s) 面向实际工作并为 AI Agent 提供动力的新一代智能等级 — OpenAI
> [10:11](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=611s) 美国在人工智能技术上的领先地位，源于数十年基础研究、大胆的创业冒险以及每年数千亿美元的私营部门投资……然而，美国掌握的信息表明，主要位于境外的实体正发起有组织的工业级行动，通过数万个代理账户规避检测并利用越狱手段，成体系地蒸馏抽取美国前沿 AI 模型的核心能力，借此剥削美国的专业知识与创新成果。 — 官方备忘录
> [10:11](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=611s) 那是因为他们对顶级模型进行了高强度蒸馏。 — 主持人
> [14:07](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=847s) 一个创始人为了推销产品上去摔断胳膊，接下来几个月都写不了代码，那可就得不偿失了。 — 主持人
> [20:38](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1238s) 我们需要对某些时间表保持现实态度 — Elon Musk
> [20:38](https://www.youtube.com/watch?v=7Xm-nhT_Y9k&t=1238s) 扔出闪光弹吧，乔迪 — John Coogan
## Mentioned
Thoma Bravo, Medallia, Qualtrics, Blackstone, KKR, Apollo Global Management, Antares Capital, Sequoia Capital, Brad Marshall, Jamie Dimon, Aaru, Y Combinator, Noah Smith, Tyler Cowen, International Monetary Fund (IMF), OpenAI, GPT-5.5, GPT-5.4, ChatGPT, Codex, Merriam-Webster, Michael Kratsios, Anthropic, Claude Opus 4.5, Qwen 3.6, Meta, Microsoft, Kristoff, Jesse Michels, Keanu Reeves, Kevin Kelly, 《必然》 (The Inevitable), Jason Shuman, Bob Iger, Thrive Capital, Reed Hastings, Disney, Netflix, Intel, Tesla, SpaceX, Elon Musk, John Coogan, Jordi Hays
## Video description
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.
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=7Xm-nhT_Y9k)
Source quotes
[3:50] Brad Marshall: 与 AI 没有任何关系，而是源于我们认为的执行力问题
[3:50] 是的，我非常确信这就是软件行业的底部了，未来绝不可能更糟了。
[6:15] Noah Smith: 笑死，一带一路彻底搞砸了，最高层真是无能。
[6:15] 主持人: 诺亚，你近期可千万别去中国。
[7:48] OpenAI: 面向实际工作并为 AI Agent 提供动力的新一代智能等级
[10:11] 官方备忘录: 美国在人工智能技术上的领先地位，源于数十年基础研究、大胆的创业冒险以及每年数千亿美元的私营部门投资……然而，美国掌握的信息表明，主要位于境外的实体正发起有组织的工业级行动，通过数万个代理账户规避检测并利用越狱手段，成体系地蒸馏抽取美国前沿 AI 模型的核心能力，借此剥削美国的专业知识与创新成果。
[10:11] 主持人: 那是因为他们对顶级模型进行了高强度蒸馏。
[14:07] 主持人: 一个创始人为了推销产品上去摔断胳膊，接下来几个月都写不了代码，那可就得不偿失了。
[20:38] Elon Musk: 我们需要对某些时间表保持现实态度
[20:38] John Coogan: 扔出闪光弹吧，乔迪
Source quotes
[3:23] Brad Marshall: execution driven issues
[10:00] Jamie Dimon: cockroaches
[10:00] Adversarial Distillation
[20:10] Elon Musk: get realistic about some timelines
Original transcript
[0:03] [laughter]
[0:04] >> We're going to get that sound ready
[0:06] because we got some bad news. Thoma
[0:09] Bravo is taking a massive write-down on
[0:12] a software company called Medallia.
[0:14] Thoma Bravo is reportedly handing over
[0:16] the software company Medallia to
[0:18] creditors after restructuring
[0:19] negotiations failed to materialize. This
[0:21] is a $5.1 billion equity wipeout for the
[0:25] firm who bought the business for $6.4
[0:27] billion in 2021. There's some background
[0:29] in Reuters that I think I should read
[0:31] through and then you can sort of give me
[0:32] your analysis and your and what you're
[0:34] seeing on the timeline around take. So,
[0:36] from Reuters, exclusive, Thoma Bravo
[0:38] nears agreement to turn software firm
[0:40] Medallia over to creditors. That does
[0:42] not sound good. Private equity firm
[0:43] Thoma Bravo is nearing an agreement to
[0:45] hand over software firm Medallia to
[0:47] lenders wrapping up months of
[0:48] restructuring negotiations. The move
[0:50] will wipe out $5.1 billion in equity.
[0:52] Medallia has struggled in recent months
[0:54] under the weight of $3 billion of debt
[0:57] which it owes to Blackstone, KKR, Apollo
[1:00] Group, and Antares Capital. Thoma Bravo,
[1:02] Blackstone, and KKR declined to comment.
[1:04] Following Medallia didn't immediately
[1:05] return requests for comments to Reuters.
[1:07] So, like other software companies,
[1:09] Medallia's valuation has been hit in
[1:11] recent months over concerns that its
[1:14] services will eventually be supplanted
[1:16] by artificial intelligence. What does
[1:18] Medallia do? Medallia provides software
[1:20] that collects and analyzes customer and
[1:23] employee feedback for companies. We've
[1:25] had some startups, some of them backed
[1:27] by Sequoia Capital, that are using AI to
[1:29] do this exact thing. That's a potential
[1:31] disruption. And then there's also
[1:33] >> Well, their their their main rivals
[1:35] Qualtrics.
[1:35] >> Yep.
[1:36] >> But yeah, and and and maybe you said
[1:37] this already, but you know that Sequoia
[1:39] was one of the big backers of Medallia
[1:41] when they were private company.
[1:42] >> interesting. So, they're
[1:43] >> I mean, they're still private, but
[1:45] >> There There's a long history of that
[1:46] with the firm that was before Zenefits
[1:48] was called like SuccessFactors. Lars
[1:51] Dalgaard did that deal.
[1:53] >> Sequoia's a real winner here. They They
[1:55] did 35 million in 2012. They did another
[1:58] 50 million in 2014 and then they later
[2:01] led a 150 million dollar round.
[2:03] >> Wow. And then eventually do they take it
[2:04] public or do they sell it for 6 billion
[2:07] to Toma Bravo directly from the private
[2:09] Was Medallia ever a public company? That
[2:11] would be interesting.
[2:12] >> they were.
[2:12] >> You look that up and I will keep reading
[2:15] from this to give some more back story.
[2:17] So private equity firms invested heavily
[2:18] in the software sector when interest
[2:20] rates were low following the peak of the
[2:22] COVID-19 pandemic. We all remember 3%
[2:24] interest rates. It was the golden era of
[2:26] startups and growth. All of the DCFs
[2:29] were massive and then of course the
[2:32] valuations came down once the interest
[2:34] rates went up. So investors have become
[2:36] increasingly nervous about the
[2:37] sustainability of high valuations
[2:39] assigned to some of those assets and the
[2:41] debt raised to buy them if you're on a
[2:42] floating rate and not you know these
[2:45] firms are not necessarily backed by a
[2:46] 30-year fixed mortgage like your house.
[2:48] Your interest rate went up and the debt
[2:51] payments ballooned and if the cash flow
[2:53] from the business has not also ballooned
[2:55] you could be in trouble like this
[2:56] company potentially is. Blackstone, KKR
[2:58] and Antares hold some of the debt in
[3:00] traded and non-traded funds. FS KKR
[3:03] Capital Corp marked the debt at 79 cents
[3:05] on the dollar in its last quarterly
[3:07] report and of course debt is senior to
[3:09] equity. So what does that mean for the
[3:11] equity? Not good and that's why this
[3:13] deal is happening. Apollo Debt Solutions
[3:15] marked it 74 cents on the dollar. So a
[3:18] question about can they even recover
[3:20] you know the full value of that debt?
[3:22] The equity is obviously in deep deep
[3:24] trouble. So Blackstone's global head of
[3:26] private credit Brad Marshall said on a
[3:27] conference call in February that
[3:29] Medallia had been quote underperforming
[3:32] not because of anything related to AI
[3:34] but due to what we believe to be
[3:36] execution driven issues. So there is a
[3:38] question of well if they restructure and
[3:41] this company gets in the hands of
[3:43] creditors maybe they can roll out AI
[3:45] features and become an AI winner and
[3:47] accelerate the top line and restructure
[3:49] the debt.
[3:50] >> I saw someone was sharing that
[3:52] uh, there were there were some sales
[3:53] people doing kind of some anonymous
[3:55] reporting and they were saying a lot of
[3:57] the reps were struggling to hit quotas,
[3:59] you know, like basically delivering like
[4:01] 20% of quota, something like that. So,
[4:03] just basically getting out-competed in
[4:05] the market.
[4:05] >> Sure. Yeah.
[4:06] >> And, uh, just go back to the earlier
[4:08] point, yes, Medallia went public in 2019
[4:10] >> 2019.
[4:11] >> on the NYSE, uh, was taken private July
[4:14] 26, 2021.
[4:15] >> Okay.
[4:16] >> Uh, it had been it had been, uh, trading
[4:18] around the $5 million mark prior to the
[4:21] take private at 6.4.
[4:22] >> So, Thoma Bravo, after taking it
[4:24] private, installed a new leadership team
[4:26] in early of 2025. Marshall, uh, the head
[4:29] of
[4:30] Blackstone's global head of private
[4:31] credit, said that they were working on a
[4:33] turnaround plan and we expect there to
[4:35] be discussions around the capital
[4:36] structure. And so, uh, people are
[4:39] blackballing on the timeline. Brandon
[4:40] says, "Yes, I am sure this is the bottom
[4:42] in software and we won't get worse from
[4:45] here." Uh, not good signs. This is, uh,
[4:48] you know, potentially one of those
[4:49] cockroaches that Jamie Dimon was
[4:51] worrying about. There's another post
[4:52] that seems like it was deleted, but
[4:55] >> Yeah, I mean, the so, the
[4:56] >> some of the screenshot here.
[4:57] >> The tough thing is that everyone
[4:59] involved here has been on a kind of a
[5:01] press tour saying like everything's
[5:03] fine.
[5:03] >> Yes.
[5:04] >> Like, we're all good.
[5:05] >> Yes.
[5:05] >> AI is going to be an accelerant, but,
[5:08] uh, even ignoring the AI question, a lot
[5:11] of these businesses no longer have, you
[5:14] know, our our founder-led. They're
[5:16] competing against other companies that
[5:17] that are founder-led. And yeah, it's
[5:20] it's just, uh, kind of a credit to Jamie
[5:23] Dimon for his comments saying like, you
[5:26] know, I I don't think this is over.
[5:28] >> Yeah.
[5:28] >> When you had First Brands and and some
[5:31] other, um, shutdowns.
[5:33] >> Yeah, in the previous startup era, there
[5:34] were I mean, I feel like it would have
[5:36] been hard to raise for just like a
[5:38] vanilla Qualtrics or Medallia competitor
[5:41] and just saying like, pre-AI, like,
[5:43] we're just going to build SaaS and we're
[5:45] just going to build a crud app and write
[5:48] software
[5:49] >> A new player is a company Aru.
[5:51] >> Yeah, that's a very different approach.
[5:54] Simulation. Yeah, that's a very
[5:56] different approach.
[5:56] >> And then we had another one. I'm
[5:57] blanking on the name. It was a YC
[5:59] founder. He
[6:00] He had gone through YC a few years ago,
[6:03] but his company is taking off doing
[6:05] doing something something similar. So,
[6:07] people have been well aware of the
[6:10] opportunity that Medallia and Qualtrics
[6:13] have owned.
[6:15] >> But, uh in other debt-related news, uh
[6:18] Xi Jinping wants to use the
[6:22] International Monetary Fund to rescue
[6:24] the array of distressed Chinese loans
[6:26] around the world. This is an interesting
[6:29] piece from the Wall Street Journal
[6:30] opinion. Noah Smith is laughing. He
[6:32] says, "Lol. Belt and Road failed so
[6:34] hard. Xi Jinping is incompetent."
[6:38] Uh that is This is a very
[6:41] >> Do not go to China, Noah. Do not go to
[6:44] China.
[6:44] >> Yeah, this is uh this is this is an
[6:46] interesting thing, but maybe a potential
[6:48] bargaining chip in all of the other
[6:49] discussions, you know, uh in in tech, uh
[6:53] Tyler always points out that uh tech is
[6:55] like too focused on chips when it comes
[6:57] to Chinese diplomacy, that there are
[7:00] many many other questions around trade
[7:03] and what Apple's doing and rare earths
[7:05] and uh there are so many and just the
[7:07] broad history of the Chinese empire that
[7:09] play into
[7:11] what their Taiwan policy is, how they
[7:13] will interact in the Middle East, and uh
[7:15] we tend to focus it all on AI. It's all
[7:17] about the data centers and the chips,
[7:19] and perhaps there are more things, and
[7:20] this is one example of something else
[7:22] that is a key factor in a new broader
[7:24] negotiation that includes chip exports
[7:27] and also rare earths and also talent and
[7:30] movements and whether or not Manes will
[7:33] be able to move over to Meta and a
[7:36] million other things. And so, that is
[7:37] the job of these world leaders is to
[7:39] swirl around all the different
[7:40] trade-offs and get to hopefully a uh
[7:44] a good deal for both sides.
[7:45] >> Some breaking news.
[7:48] >> Merriam-Webster
[7:48] >> GPT-5.5
[7:50] >> Oh.
[7:50] >> is out.
[7:51] >> is?
[7:51] >> Uh
[7:52] is out. Introducing GPT-5.5, a new class
[7:54] of intelligence for real work and
[7:56] powering agents built to understand
[7:57] complex goals, use tools, check its
[7:59] work, and carry more tasks through to
[8:02] completion. It marks a new way of
[8:03] getting computer work done.
[8:05] >> Mhm.
[8:06] >> Now available in ChatGPT and Codex.
[8:08] >> this demo of the Rubik's Cube.
[8:10] >> GPT-5.5 excels at writing and debugging
[8:12] code, researching online, analyzing
[8:13] data, creating documents and
[8:14] spreadsheets, operating software, and
[8:17] moving across tools until a task is
[8:19] finished. GPT-5.5 delivers this step up
[8:21] in intelligence without compromising on
[8:22] speed, matches
[8:25] uh GPT-5.4 per token latency in
[8:27] real-world serving. There's the model
[8:30] card there.
[8:31] >> And uh there are some other posts.
[8:33] >> We will wait for the reactions to come
[8:34] in. We will we'll gather them. We'll
[8:36] summarize them. But the team cooked.
[8:38] >> Cooked.
[8:39] >> Yeah, absolutely.
[8:40] >> Congratulations to everyone that worked
[8:41] on this.
[8:41] >> Um let's head over to Merriam-Webster,
[8:43] John. What do you got?
[8:44] >> Merriam-Webster Merriam-Webster added a
[8:46] new word to the dictionary, I believe.
[8:49] >> And I feel like the pace of
[8:52] >> added this?
[8:52] >> I don't know. I I mean, it's in
[8:55] Merriam-Webster.
[8:55] >> a slang section.
[8:56] >> Oh, okay. Slang and trending. They must
[8:58] be working overtime over there because
[9:01] the number of wambos, if you're not
[9:02] familiar with wambos, these are word
[9:05] combos,
[9:06] uh things like
[9:07] >> lorain
[9:08] >> lorain
[9:10] >> or explain
[9:11] >> or quee
[9:12] >> I could tell John, "Can you lorain
[9:15] GPT-5.5?"
[9:16] >> Yes, give you the lore
[9:18] >> and explain
[9:18] >> tell you about what happened with 4.5,
[9:20] 4.0, 3.5, 3 DaVinci, give you the whole
[9:24] lore, but then also explain what this
[9:26] model is capable of. That's the full
[9:28] lorain of GPT-5.5. Uh but we're not here
[9:32] to talk about lorain, we're here to talk
[9:34] about choppelganger, which is a new word
[9:36] in Merriam-Webster the dictionary. It's
[9:40] basically the main dictionary as far as
[9:41] I'm concerned. Choppelganger is a term
[9:44] for a less attractive version of someone
[9:47] or something. So, whoever's going out
[9:50] there and distilling 5.5 into sort of a
[9:53] chopped version of it, that will be the
[9:55] choppelganger of the official GPT 5.5.
[9:59] So, stay away from the choppelgangers
[10:01] unless you're unless you're really down
[10:02] your luck and you just said you need and
[10:04] you need a Chinese open source model to
[10:06] do something for you. Then, you know,
[10:07] buyer beware because it might it might
[10:09] have some flaws. Separately, there is
[10:11] there is new news from Michael Kratsios
[10:12] that the government seems to be taking
[10:15] Anthropic and Open AI's messaging around
[10:17] distillation very seriously and is
[10:20] setting up a task force. We can read a
[10:21] little bit more about this later, but
[10:23] setting up a task force to actually
[10:25] figure out how to prevent distillation
[10:27] at scale. As the models get bigger, it's
[10:29] more and more of an economic impact.
[10:31] When you're talking about, you know, a
[10:34] $100 million training run and a Chinese
[10:37] lab can distill it and sneak the the
[10:39] weights out and sneak the
[10:41] you know, exfiltrate the data, that's a
[10:42] lot different of an economic impact from
[10:44] stealing something that I don't know
[10:46] cost billions to to train or billions to
[10:48] to put together. Good news there.
[10:50] Hopefully they're successful. It was
[10:51] also sort of a white pill to see a lot
[10:53] of the labs working together to
[10:54] understand, hey, we're seeing this weird
[10:57] amount of data go to this particular
[10:59] company. Are you seeing the same thing?
[11:00] Or it's a shell company and we have five
[11:02] shell companies in these areas that are
[11:05] asking for these queries and and then
[11:07] the other lab says, oh, yeah, we're
[11:08] actually getting something similar. And
[11:10] if you puzzle piece those together, all
[11:12] of a sudden you're getting a really
[11:13] solid map of what a frontier system is
[11:15] capable of. And so, it feels like we,
[11:18] the industry, are the victim of a
[11:20] distillation attack even if it might not
[11:23] register among a single lab because that
[11:25] single lab is only getting hit with like
[11:27] 1/3 or 1/5 of the total uh
[11:31] >> Subject: Adversarial Distillation of
[11:33] American AI models. The United States
[11:34] leads the world in AI technologies. That
[11:37] lead uh that lead reflects decades of
[11:39] foundational research, bold
[11:40] entrepreneurial risk-taking,
[11:41] [clears throat] and hundreds of billions
[11:42] of dollars in annual private investment.
[11:44] American AI leadership drives economic
[11:46] growth, strength- strengthens national
[11:48] security, and advances the frontiers of
[11:49] science, medicine, and human knowledge.
[11:51] The breakthroughs emerging from American
[11:52] industry raise living standards, expand
[11:55] opportunity, and improve lives around
[11:56] the world. However, the United States
[11:58] has uh information indicating that
[12:00] foreign entities, principally based in
[12:01] China, are engaged in deliberate
[12:03] industrial-scale campaigns to distill US
[12:06] frontier AI systems, leveraging tens of
[12:08] thousands of proxy accounts to evade
[12:10] detection, and using jailbreaking
[12:12] techniques to expose proprietary
[12:14] information. These coordinated campaigns
[12:16] systematically extract capabilities from
[12:19] American AI models, exploiting
[12:22] >> I don't like that. That sounds bad.
[12:24] >> all.
[12:24] >> [laughter]
[12:24] >> Exploiting American expertise and
[12:27] innovation.
[12:28] >> Every AI researcher needs this button on
[12:30] their desk. If they detect this
[12:33] distillation attack, fire it off and the
[12:35] security team will come and make sure
[12:38] you're locked down.
[12:40] Uh and it does seem like overall uh all
[12:43] the labs are being more careful about uh
[12:45] how the APIs roll out, where how they're
[12:48] doing KYC, how they're doing uh
[12:50] different uh you know, partnerships and
[12:53] stuff. And I think it's all a back and
[12:55] forth and a dynamic. But uh it's fun
[12:57] that this new model is available and uh
[12:59] people will play with it and I'm sure
[13:00] we'll see some cool stuff. Some
[13:01] >> Yeah, and Qwen 3.6 rolled out yesterday.
[13:05] Uh people are like, "Wow, it's it's
[13:07] almost as good as Opus 4.5."
[13:10] >> [laughter]
[13:10] >> Like, "Yeah, because they train they
[13:13] they distill their models."
[13:15] Anyways, I'm glad that Kratos is on it
[13:18] and I'm glad that the labs can uh
[13:20] coordinate and work together on this
[13:21] one.
[13:21] >> Yes.
[13:22] >> More breaking news.
[13:23] >> Yes.
[13:23] >> And uh unfortunate, but we had talked
[13:26] about this pr- uh previously. Meta is
[13:27] cutting 10% of their workforce, which is
[13:30] 8,000 employees.
[13:31] >> But this has been talked about for a
[13:33] while, yeah.
[13:33] >> 6,000 open roles.
[13:36] >> I saw Microsoft was doing something
[13:37] similar. They're they're offering early
[13:39] retirement to something like 7% of the
[13:41] workforce, trying to to lean out a
[13:43] little bit. Uh I'm sure the debate will
[13:45] continue to rage over the the underlying
[13:47] motivations. These companies can be big
[13:49] and sometimes maybe they're too big,
[13:51] maybe there's AI gains, maybe there's,
[13:54] you know, different investment
[13:55] strategies. Uh we will have to see where
[13:58] they are moving chips around, what
[14:00] projects they are actually cutting, what
[14:02] projects they are doubling down on,
[14:03] because these companies are also hiring
[14:05] at all times effectively. Uh well, uh
[14:07] Kristoff had some actual media ideas for
[14:10] the tech industry. He says there's not
[14:13] enough media in tech.
[14:15] >> [laughter]
[14:15] >> We need more We need more We need more
[14:17] tech-positive
[14:18] >> media, for sure. So, he says, "Silent
[14:21] library with founders. Winner gets
[14:24] investment."
[14:25] >> What is silent library?
[14:26] >> Silent library? So, you're I Is this
[14:28] like a like a webcam that observes
[14:31] people
[14:31] >> is a television show. It had four
[14:34] seasons. It's a game show.
[14:36] >> a game show. Game shows are fun. I could
[14:38] see a game show being being fun, at
[14:40] least different.
[14:41] >> Six friends vying for a cash prize, if
[14:43] only they can remain silent as one of
[14:45] them is forced [laughter] to endure a
[14:47] bizarre stunt while seated in a library.
[14:50] >> Okay. Okay.
[14:51] >> Uh what else? Jackass with founders.
[14:54] >> What does that mean? How How do you make
[14:56] that tech-related? Maybe like humanoid
[14:58] robots? Like I I When I think of Johnny
[15:00] Knoxville, I just think of riding bull
[15:02] riding effectively. And so, I imagine
[15:05] like riding some sort of, you know,
[15:06] robotic bull would be it. But, you know,
[15:09] it's very dangerous. Like the the the
[15:11] the team that worked with like Bam
[15:13] Margera and and Johnny Knoxville, uh
[15:16] they had some serious injuries from time
[15:18] to time. And and I I don't know if it's
[15:21] in your best interest to be a founder
[15:22] and be like, "Yeah, I need to I need to
[15:24] launch my startup, so I got to go in the
[15:25] ring with a mechanical bull and
[15:28] potentially break my arm and not be able
[15:30] to code." Seems rough, but would be
[15:32] entertaining and with the right twist, I
[15:34] believe it could work. Interview hot
[15:36] founders while going ghost hunting. I
[15:38] feel like paranormal tech content Jesse
[15:41] Michaels doesn't get enough credit here.
[15:43] You know, he's been
[15:43] >> technology podcast.
[15:44] >> He's He's a
[15:45] It's in It's in the technology charts
[15:47] and it charts high. Like it's up there
[15:50] and he has an incredible pedigree, great
[15:52] investor, and also,
[15:55] you know, has some of the best reporting
[15:57] on aliens and paranormal activity and
[16:00] conspiracy theories. And I guess he just
[16:02] hasn't cracked the way to bring a
[16:04] founder along. But if you're if you're
[16:06] friends with Jesse Michaels or in touch,
[16:08] maybe you just tag along for a little
[16:10] cameo on one of his videos and that's
[16:13] all the PR you need.
[16:14] >> Founders give their pitches while
[16:16] skydiving. Can you
[16:18] >> I like skydiving.
[16:19] >> Can you talk while skydiving?
[16:21] >> Uh with the right headset.
[16:22] >> gone skydiving?
[16:23] >> No, but I like the idea and I like the
[16:25] videos.
[16:25] >> Have I told you the story where I was I
[16:28] was in on an overnight flight back from
[16:30] Europe?
[16:30] >> No.
[16:31] >> I'm sitting on a on a plane and I'm
[16:34] awake, I can't sleep.
[16:35] >> Okay.
[16:36] >> And out of the shadows, John Wick
[16:40] >> Oh.
[16:41] >> up the aisle.
[16:42] >> Keanu Reeves. Yeah, the real actor.
[16:45] >> It's effectively like 2:00 a.m.
[16:46] >> Yeah.
[16:47] >> And Keanu Reeves is like walking at me
[16:50] on this plane in the middle of the
[16:51] night.
[16:51] >> Yeah.
[16:52] >> Um we ended up talking for like 5
[16:53] minutes. It was cool. He went and then
[16:55] he went back to his seat.
[16:56] >> Really?
[16:57] And yeah, well, he stayed on the plane.
[16:59] He did not
[17:00] >> Super super nice guy.
[17:02] >> What else do they have here? SF party,
[17:04] I'm smacked videos. I don't know I'm
[17:06] smacked.
[17:07] >> Wow. Unk. You're unk. You got You guys
[17:09] know.
[17:10] >> I kind of remember, but uh
[17:11] >> You remember?
[17:12] >> Basically, this was his whole like like
[17:15] you could call it a media company, but
[17:16] it was really like a guy that would make
[17:18] party videos.
[17:20] >> What what era are we talking of?
[17:22] >> Okay.
[17:24] >> Supreme milk.
[17:26] >> Okay. Okay.
[17:35] >> Not speaking of milk.
[17:38] >> Yeah. Yeah. Yeah.
[17:39] >> Kevin Kelly, the founder of Wired, had
[17:42] some incredible predictions in 2016.
[17:44] Let's read through them. Summary of The
[17:46] Inevitable Understanding the 12
[17:48] Technological Forces That Will Shape Our
[17:51] Future. This was Kevin Kelly's book from
[17:53] 2016. According to Kelly, much of what
[17:56] will happen in the next 30 years is
[17:58] inevitable. The future will bring with
[17:59] it even more screens, tracking, and lack
[18:02] of privacy. In the book, he outlines 12
[18:04] trends that will forever change the ways
[18:06] we work, learn, and communicate.
[18:08] Becoming, moving from fixed products to
[18:10] always upgrading services and
[18:12] subscriptions. That has definitely
[18:14] happened. Uh we're moving away from even
[18:16] seats. Everything's consumption-based
[18:17] now. Cognifying, making everything much
[18:20] smarter using cheap, powerful AI that we
[18:22] get from the cloud. Nailed it.
[18:24] >> [laughter]
[18:24] >> That That is
[18:25] >> Blowing.
[18:25] >> fantastic prediction for 2020. Assuming
[18:28] he wrote it He actually wrote it
[18:29] probably in 2015, published in 2016.
[18:33] >> But he's the creator of Wired. He's been
[18:35] tapped into tech his entire career.
[18:37] Depending on Unstoppable Streams of
[18:39] Real-Time for Everything.
[18:41] For sure. Uh
[18:43] turning all surfaces into screens. That
[18:45] definitely happened. Your toaster can
[18:46] watch
[18:47] >> Screening.
[18:47] >> TV now.
[18:48] >> Uh
[18:49] >> Shifting society from one where we own
[18:51] assets to one where instead we have
[18:52] access to all services at all times.
[18:54] Collaboration at mass scale. On my
[18:56] imaginary sharing meter index, we are
[18:58] still two out of 10. Filtering,
[19:00] harnessing intense personalization in
[19:02] order anticipate our desires. Remixing,
[19:04] unbundling existing products into their
[19:06] most primitive parts and then
[19:07] recombining in all possible ways. That's
[19:09] definitely happening. Coconut Simulator,
[19:11] unironically, an example of that.
[19:13] Interacting immersing ourselves inside
[19:15] of computers to maximize their
[19:17] engagement. Tracking and playing total
[19:19] surveillance for the benefit of citizens
[19:20] and consumers. Sounds scary, potentially
[19:23] a good outcome if things are done
[19:24] properly.
[19:25] Promoting good questions is far more
[19:27] valuable than answers. What?
[19:29] >> You think total surveillance is
[19:31] potentially good?
[19:31] >> says total surveillance for the benefit
[19:34] of citizens and consumers. So, if there
[19:36] is a black box where my Netflix
[19:38] recommend where my Netflix activity
[19:40] exists, where no Netflix employee can
[19:42] see it because it's encrypted, but it
[19:44] can make great recommendations and
[19:46] recommend me the next great show that I
[19:47] will actually enjoy, I'm cool with that
[19:49] surveillance.
[19:50] >> For the benefit of John Coogan.
[19:51] >> Yeah, yeah. There there is there is a
[19:53] surveillance bull case. Constructing a
[19:55] planetary system connecting all humans
[19:57] and machines into a global matrix. Okay,
[19:59] that one we're still waiting on. But, a
[20:01] lot of good interesting predictions.
[20:03] Jason Schuman says wild how accurate
[20:06] these predictions were and they were in
[20:08] fact.
[20:09] >> Well, more predictions he knows in the
[20:10] journal. Bob Iger is returning to where?
[20:14] >> Disney?
[20:15] >> Thrive.
[20:15] >> Thrive? No way. That's amazing.
[20:18] >> Back to Thrive.
[20:19] >> Love it.
[20:20] >> Uh yeah, I think he he he's been an LP
[20:22] in Thrive. He also bought a piece of
[20:25] Thrive. Remember?
[20:26] >> Okay, well, that'll be a good next act
[20:28] for him. I'm very interested to see
[20:30] where he goes. There's a whole alumni
[20:32] class coming together. Reed Hastings is
[20:33] out and more on to the next thing. We'll
[20:36] see where they go. Hopefully
[20:38] >> Uh back to Intel. Intel announced its
[20:40] first quarter earnings after the bell on
[20:41] Thursday beating it analyst expectations
[20:44] on the top and bottom line and providing
[20:45] better than anticipated Q2 guidance on
[20:47] strong data center sales. Intel said it
[20:50] expects revenue of 13.8 to 14.8 billion
[20:53] for the second quarter. Wall Street was
[20:55] anticipating 13 billion and
[20:59] as of this morning they were they they
[21:01] were at around a 100 at 100 um
[21:04] times PE and so only up for Intel.
[21:09] >> Climbing for the ranks, climbing the
[21:10] ranks. Tesla also released Q1 earnings.
[21:14] Revenue of 2.22.4
[21:16] billion versus 21.4 billion estimated.
[21:20] So they beat on top line. They also beat
[21:22] on net income. 1.45 billion versus 1.17
[21:27] billion estimate. The interesting
[21:29] article in the journal was that
[21:31] Elon was being more cautious about Tesla
[21:35] talking saying I think we need to get
[21:37] realistic about some timelines. So he is
[21:41] certainly not, you know, pumping
[21:43] everyone up and and he's trying to sort
[21:45] of reset around
[21:47] the fundamentals and so
[21:49] we will see where that goes. It is a
[21:51] wild timeline that SpaceX if it goes out
[21:53] of 1.75 trillion will be bigger than
[21:56] Tesla which is sitting around 1.1, 1.2
[21:58] trillion these days. So
[22:00] still both huge companies.
[22:02] >> Credit to Bubble Boy over on X 2 hours
[22:04] ago he says everyone asked me about how
[22:07] I'm playing earnings. He says doubling
[22:09] down 25% of my portfolio is in Intel
[22:12] calls.
[22:13] >> Wow, there we go Bubble Boy. Congrats.
[22:16] >> Well done.
[22:16] >> Good.
[22:16] >> Well played.
[22:17] >> Stuff. Well
[22:18] >> Well played.
[22:18] >> Leave us five stars on Apple podcast and
[22:21] Spotify. Sign up for our newsletter at
[22:22] tbpn.com. Throw that flashbang Jordie
[22:24] cuz we are out of here.
[22:25] >> an honor.
[22:26] >> It's been an honor and a pleasure
[22:27] hanging out with us.
[22:28] >> We love you.
[22:29] >> Tomorrow.
[22:30] >> We'll see you soon.
[22:31] >> Goodbye.
