Harry Stebbings
Why the AI infrastructure market is just like the energy market “There are very interesting parallels between AI infrastructure and the energy market. There are a lot of different ways to make money across the value chain. What Crusoe is focused on doing is building an AI super major, vertically integrated across upstream, midstream and downstream. Where the margin accrues is going to move around. We are seeing a similar phenomenon play out in AI, where our margins are going to move around across electrical data centers, chips and services.” @ChaseLochmiller Love to hear your thoughts @GavinSBaker @branninmcbee @stephenbalaban @Electron_Cowboy
中文: 为什么人工智能基础设施市场与能源市场一样 人工智能基础设施与能源市场之间存在非常有趣的相似之处。在整个价值链中，赚钱的方式有很多。 Crusoe 的重点是打造一个垂直集成于上游、中游和下游的人工智能超级巨头。 保证金累积的地方将会有其在移动。我们在人工智能领域也出现了类似的现象，我们的利润率将在电气数据中心、芯片和服务领域各地移动。@ChaseLochmiller 喜欢听你的想法 @GavinSbaker @branninmcbee @stephenbalaban @Electron_Cowboy
🎬
视频
#173 🔗 原推
2026-10-05 01:11:00
Harry Stebbings
How Crusoe builds a portfolio of different financially returning chip products “We have a range of different deals that we are doing with the compute we are buying. We are renting capacity on a long-term basis, call it five-year contracts, to credit-quality customers. There are shorter-term contracts that we will do at higher margins, but they are riskier. Then there are other services that we provide, like Crusoe Managed Inference and our serverless fine-tuning product. Those contracts are typically much shorter-term in nature, and they end up being higher margin to Crusoe.” @ChaseLochmiller Love to hear your thoughts @branninmcbee @bernhardsson @vipulved @tuhinone
🎬
视频
#172 🔗 原推
2026-10-04 22:50:00
Harry Stebbings
AI is making, not replacing, jobs. “Data centers are leading to this massive economic boom for the entire blue-collar labor economy in the United States. We are hiring people in factories and in the field, tons of skilled trade workers working with their hands to bring the infrastructure of intelligence to life. People are concerned that AI is going to take jobs, but yet it has so far only created tremendous amounts of jobs and economic development.” @ChaseLochmiller Love to hear your thoughts @Electron_Cowboy @mikeroweworks @erikbryn
中文: 人工智能正在创造而不是取代工作。 数据中心正推动美国整个蓝领劳工经济实现大规模经济繁荣。 我们正在工厂和现场招聘人员，这些技术娴熟的行业工人正在用手工作，以将情报基础设施建设实现现实。 人们担心人工智能将会占据就业机会，但到目前为止，它只创造了大量就业机会和经济发展。@ChaseLochmiller 喜欢听你的想法 @Electron_Cowboy @mikeroweworks @erikbryn
🎬
视频
#171 🔗 原推
2026-10-04 20:46:00
Harry Stebbings
Is it BS that data center buildout causes energy prices to go up for local communities? “When you look at markets where data centers have made investments and built big data centers, typically energy prices for communities have come down. It catalyzes more investment in energy generation technology and capacity, and you end up with more megawatts being amortized over the same transmission and distribution infrastructure. We are very supportive of the data center industry helping to bring online new power production in order to support those energy requirements.” @ChaseLochmiller Love to hear your thoughts @tylerhnorris @NatBullard @AriPeskoe @shaylekann
中文: 数据中心建设是否会导致当地社区的能源价格上涨? 当你观察数据中心投资并建设大数据中心的市场时，通常社区的能源价格已经下降。 它推动了对能源生产技术和容量的更多投资，最终通过相同的输电和配电基础设施实现了更多的兆瓦被摊销。 我们非常支持数据中心行业，帮助实现在线新的电力生产，以满足这些能源需求。@ChaseLochmiller 很喜欢听你的想法 @tylerhnorris @NatBullard @AriPeskoe @shaylekann
🎬
视频
#170 🔗 原推
2026-10-04 18:08:00
Harry Stebbings
One of the biggest commonalities of the most impressive founders I meet is that their parents are university lecturers/professors. I have no idea why, but it's a clear pattern.
中文: 我遇到的最令人印象深刻的创始人最大的共同点之一就是，他们的父母是大学讲师/教授。 我不知道为什么，但这是一个明确的模式。
#169 🔗 原推
2026-10-04 18:04:18
Harry Stebbings
Why do you have to be vertically integrated to be a data center provider today? “The supply chain to support large-scale AI data centers is sort of like a game of Whac-A-Mole. At different points, different bottlenecks come into play. One of the key bottlenecks was a power distribution center. The lead time for this one component was 100 weeks, and I said, ‘Well, I do not have 100 weeks.’ We had consciously vertically integrated electrical manufacturing. We were able to do it in 28 weeks, and having the resources internally gives you a lot more flexibility to unblock these key bottlenecks.” @ChaseLochmiller Love to hear your thoughts @stephenbalaban @MattLoszak @BrianGitt
🎬
视频
#168 🔗 原推
2026-10-04 15:41:00
Harry Stebbings
I think we often hear a lot of bullshit about venture partnerships. I will tell you about mine with @PaulBonnet and @Kieranleehill. They make me better. They make me think smarter, and I know that whatever situation we are in, we will get through it together, stronger. That, to me, is a venture partnership.
中文: 我觉得我们经常听到很多关于风险投资合作的胡说八道。 我会和@PaulBonnet和@Kieranleehill一起告诉你我的情况。 他们让我变得更好。他们让我更聪明地思考，我知道无论我们处于何种情况，都会共同度过，更加坚强。 对我来说，那是一种创业合作。
media 0 共 1 项
#167 🔗 原推
2026-10-02 22:23:17
Harry Stebbings
It is BS to think that open source runs away with it by being cheaper. “This whole cost dynamic does not mean that open source just runs away with it. The labs are obviously also going to offer much cheaper versions of their own models. Their access to compute is structurally very strong, and they have a lot of different ways to make money. It does not mean that open source will dominate, but it does mean that open source is going to be a big part of the market.” @jaltma Love to hear your thoughts @natolambert @bindureddy @alexatallah @Teknium
中文: 认为开源通过更便宜来摆脱它是一种理想。 这种整体成本动态并不意味着开源就会消失。 实验室显然也将提供更便宜的自有型号版本。他们对计算的获取在结构上非常强大，而且他们有很多不同的赚钱方式。 这并不意味着开源会占据主导地位，但意味着开源将成为市场的重要组成部分。 喜欢听你的想法 @natolambert @bindureddy @alexatallah @Teknium
🎬
视频
#166 🔗 原推
2026-10-02 18:26:00
Harry Stebbings
Should we invest in Jev at $10 billion? “Of course we should do it. After the last 20VC, I have decided to recommend up to 30% of the fund. Jev is already 17% of the traffic on OpenRouter and 20% of the traffic through Vercel’s router. Jev is a 70th of the price and 100 times faster. This is exactly the kind of bet we have to do. I do believe there is a 50X upside to $500BN.” @jasonlk Love to hear your thoughts @EGafni @hardimanjames @alexatallah @cramforce
中文: 我们应该以100亿美元投资捷英河吗? 当然我们应该这么做。在经历了最后20VC之后，我决定推荐最多30%的基金。 Jev 已经占 OpenRouter 流量的 17%，通过 Vercel 路由器的流量占其流量的 20%。杰夫的价格是它的70分之一，速度是现在的100倍。 这正是我们必须做的赌注。我相信有50倍的上涨空间，达到5000亿美元。”@jasonlk 很喜欢听你的想法 @EGafni @hardimanjames @alexatallah @cramforce
🎬
视频
#165 🔗 原推
2026-10-02 15:52:00
Harry Stebbings
Seed round should still be $2M to $3M. “You can do as much for $2M to $3M as you could 10 years ago. If you do not have folks dying to give you capital outside of demo day, that may still be the natural atomic amount of capital. A seed round should still be $2M to $3M.” @jasonlk Love to hear your thoughts @chudson @dunkhippo33 @TurnerNovak @semil
中文: 种子轮融资金额仍为200万至300万美元。 你可以用200万到300万美元来做到10年。 如果你没有人在演示日之外为你提供资本，那可能仍然是资本的自然原子量。 种子轮仍应是200万至300万美元。@jasonlk 很喜欢听你的想法 @chudson @dunkhippo33 @TurnerNovak @semil
🎬
视频
#164 🔗 原推
2026-10-02 01:03:00
Harry Stebbings
Why will we see many more Neo Lab acquisitions? “All the big foundation model companies are probably in the market to acquire some kind of robotic foundation model story. There are still 100 Neo labs, so you have got to be in the 10 that win. I think there will be a bunch of these big acquisitions over the next six to 12 months if the market continues to hold.” @rodriscoll Love to hear your thoughts @peteflorence @pathak2206 @lachygroom @BerntBornich
中文: 我们为何会看到更多新实验室的收购? 所有大型基础模型公司都可能在市场上获取某种机器人基础模型的故事。 仍有100个Neo实验室，因此你必须进入那场胜利的10强。 我认为，如果市场继续持有这种规模，未来六到十二个月内将会有大量此类收购。”@rodriscoll 很喜欢听你的想法 @peteFlorence @pathak2206 @lachygroom @BerntBornich
🎬
视频
#163 🔗 原推
2026-10-01 22:55:00
Harry Stebbings
Why the personal assistant space is very reminiscent of the AI coding space “It kind of reminds me of what happened with coding and with Cursor and Cognition in the face of the labs. These agents can interact with the entire third-party internet. It is not just talking to it anymore. This is something that primarily does things for you. When something is that important, a lot of things can win. There can be an amazing independent player like Instinct, and the labs will have offerings around this.” @jaltma Love to hear your thoughts @alexgraveley @pirroh @ScottWu46 @mattshumer_
中文: 为什么个人助理空间让人联想到人工智能的编码空间 这让我想起了编程以及Cursor和Cognition在实验室面前所发生的事情。 这些代理可以与整个第三方互联网进行交互。现在不再只是和它说话了。这主要是为你做事情的。 当某件事如此重要时，很多事情都可能获胜。可以有像《本能》这样出色的独立玩家，实验室也会围绕这一点提供产品。”@jaltma 喜欢听你的想法，@alexgraveley @pirroh @ScottWu46 @mattshumer_
🎬
视频
#162 🔗 原推
2026-10-01 20:23:00
Harry Stebbings
So @harmonic_ai just released their Hot 25 Report on the most in-demand early-stage companies. The top 3: 🥇 @resolveai AGAIN for the 3rd time. 🥈 @TheLatentCo making it's debut as the highest ranking newcomer on the list. 🥉@Starcloud_ returning after last being on the list back in Q3 2025 Newcomers to watch: @PrimeIntellect at #4, Strala AI #5 & @trajectorylabs at #7 Find the full report here: https://harmonic.ai/hot-25-startups/q4-2026
中文: 因此，@harmonic_ai 刚刚发布了关于最抢手的早期公司的热门25号报告。 前三: @resolveai 再次进入第3次。 🥈 @TheLatentCo 首次成为榜单上排名最高的新人。 🥉@Starcloud_ 在上一次进入榜单后回归，时间是在2025年第三季度 新来者观看:@PrimeIntellect 关注 #4，Strala AI #5 和 @trajectorylabs 发布 #7 请在此处查看完整报告:
🎬
视频
#161 🔗 原推
2026-10-01 17:30:00
Harry Stebbings
Spotify 👉 https://open.spotify.com/episode/0QYKPkttL5apwW0zE311R9 Youtube 👉 https://youtu.be/UNH5YtuK5UE Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-instinct-raises-%241b-at-%2410b-valuation-amd-buys/id958230465?i=1000792526682
#160 🔗 原推
2026-10-01 14:19:51
Harry Stebbings
The best products are created by people solving a problem for themselves. I wanted a podcast that discussed the biggest news in tech, every week, with amazing analysis and left the politics and ego aside. This is the result and the only show you have to listen to every week. - Instinct Raises $1B at $10B Valuation - AMD Buys Fei-Fei Li's World Labs for $8.2B - Meta Poaches MongoDB's CEO - Oura Pulls IPO - Nubank Eyes $8–12B Monzo Takeover My notes below with @jasonlk, @rodriscoll, and @jaltma 1. It Is BS to Think That Open Source Runs Away With It by Being Cheaper Open-source models will not win on price alone because frontier labs hold structural advantages in compute scale, distribution, and revenue. Closed providers can subsidize lower-tier models aggressively enough to match open-source pricing. Open source will still capture a meaningful share of developer workloads, but frontier labs will compete hard on price to defend their position. 2. Why the Personal Assistant Space Is Very Reminiscent of the AI Coding Space Consumer AI assistants like Instinct represent an "aggregator of aggregators" shift similar to Cursor's impact on coding. Instead of acting as passive chat boxes, these autonomous agents can execute complex, multi-step actions across the live internet. By fundamentally reshaping how consumers interact with software and services, this new paradigm creates room for both startups and incumbents to build enormous value. 3. Tyler Cowen's Prediction for the Future of Venture AI is driving greater variance across venture capital, concentrating returns around an even smaller number of breakout winners. As Tyler Cowen put it, "Variance is gonna go up with AI, and many of you will fail." Traditional compounding strategies are becoming less reliable as capital and value creation cluster around outliers. 4. Why Will We See Many More Neo Lab Acquisitions? Foundation model labs and semiconductor giants like AMD are actively acquiring Neo Labs to expand technical capabilities and defend strategic positions. Few of the roughly 100 existing research labs are likely to survive as independent businesses. Those with differentiated technical assets and specialized domain models could become highly valuable acquisition targets. 5. Seed Round Should Still Be $2M to $3M Despite the hype around $50 million seed rounds, $2 million to $3 million remains the atomic unit of early-stage investing. AI development tools now allow lean founding teams to achieve as much operational progress on $2 million to $3 million as startups required far more capital to achieve a decade ago. Outside capital-intensive research labs, disciplined initial rounds help limit dilution and preserve healthy fund mechanics. 6. Should We Invest in Jev at $10 Billion? Running top-tier frontier models for 10 to 12 hours a day is economically unsustainable for many enterprise workflows. Lower-cost alternatives like Jev, operating at a fraction of the price and up to 100x the speed, are already capturing significant token volume. As compute budgets become more constrained, specialized efficiency models could represent a major venture opportunity. (links in comments)
#159 🔗 原推
2026-10-01 14:19:41
Harry Stebbings
Agents require completely different search inputs, outputs, and latencies “The problem’s inputs are different, outputs are different, and constraints are different. Imagine someone running an agent built with a Luna model and someone running an agent built with a Fable model. They are very different models. How you want to optimize signal-to-noise and tokens for each of them is so different in terms of what you do in the web search stack.” @paraga Love to hear your thoughts @sarahmsachs @simonlast @RLanceMartin @charlespacker
中文: 代理需要完全不同的搜索输入、输出和延迟 问题的输入不同，输出不同，约束也不同。 想象一下，有人在运行一个使用Luna模型构建的代理，以及运行一个使用Fable模型构建的代理。它们是截然不同的模型。 在网页搜索栈中，你希望如何优化每个信号与噪音和令牌的方式，差异很大。 喜欢听你的想法 @sarahmsachs @simonlast @RlanceMartin @charlespacker
🎬
视频
#158 🔗 原推
2026-09-30 01:03:00
Harry Stebbings
Search compute is an economic trade-off against model compute “You are essentially allocating compute to web search in order to save compute on the model. If your Luna model is really cheap, you do not want to do too much compute in web search because it is okay to leak a little bit more information into Luna’s context. Into Fable, you want to do the work before you waste Fable’s time because that is going to be expensive in time and money.” @paraga Love to hear your thoughts @rweiss57 @jerryjliu0 @douwekiela @jeffreyhuber
🎬
视频
#157 🔗 原推
2026-09-29 22:49:00
Harry Stebbings
Why team sizes won't be impacted as much as people think “Our legal team is over 10 people, our customer success team is over 40 people. All of them use AI heavily. I definitely can say that I do not see any elimination. Over 60% of customer support requests can be handled with AI, but when it especially comes to B2B, AI just does not work.” @alexmashrabov Love to hear your thoughts @marty_kausas @jasonlk @searchbrat @maryshenocarro1
🎬
视频
#156 🔗 原推
2026-09-29 20:12:00
Harry Stebbings
Is the man high? You can’t my friend cos energy prices are 4x higher here than elsewhere. You can’t because you need 5 years and more regulation than ever to open a datacentre. You have to create the environment and ecosystem for AI to thrive. And we haven’t.
中文: 男人高吗? 我朋友不能，因为这里的能源价格比其他地方高出四倍。 你不能因为开设数据中心需要五年时间和更多的监管。 你必须创造环境和生态系统，才能让人工智能蓬勃发展。而我们还没有。
#155 🔗 原推
2026-09-29 17:50:42
Harry Stebbings
The 150-person content team powering Higgsfield's billion in ARR “We have an in-house team of over 150 creative professionals. It is almost half of the whole workforce. For 90 minutes of TV-quality content, it was over 100 hours of AI-generated content. Creative decision-making, picking the right piece, is still very important. That is what is driving most of the revenue.” @alexmashrabov Love to hear your thoughts @Diesol @bilawalsidhu @PJaccetturo @c_valenzuelab
🎬
视频
#154 🔗 原推
2026-09-29 17:40:00
Harry Stebbings
“Ads do not work with agents in their current form. Agents show up, no one sees ads, and you make no money. We are effectively building an AdSense for agents showing up to read your content. We like to pay content owners a variable amount of money every time an agent derives benefit from reading their information.” @paraga
中文: 广告目前不与代理人员合作。代理人出现，没有人看到广告，而你赚不到钱。 我们正在为出现阅读您内容的代理人员构建一个AdSense。 我们喜欢每次代理人通过阅读内容获得收益时，向内容所有者支付可变金额的费用。@paraga
🎬
视频
#153 🔗 原推
2026-09-29 15:14:00
Harry Stebbings
“Agents will use the web 1,000x more than humans. Hence, new tech is needed and new business models are needed. No tech built for a certain scale survives three orders of magnitude. When you need new business models alongside new technology, a problem becomes really interesting.” @paraga
中文: 特工们将比人类多使用1000倍的网络。因此，需要新技术，需要新的商业模式。 任何为特定规模而构建的技术能否存活三个数量级。 当你需要将新商业模式与新技术一起使用时，问题就会变得非常有趣。@paraga
🎬
视频
#152 🔗 原推
2026-09-29 00:51:00
Harry Stebbings
I think I must be the most unprofitable user for Instinct ever. Easily burning $30K plus for them! 🤣
中文: 我认为我一定是《本能》有史以来最无利可图的用户。 轻松烧掉3万美元以上!🤣
#151 🔗 原推
2026-09-28 23:23:14
Harry Stebbings
The controversial question: How do you really calculate revenue? “We look at revenue over the last 28 days and multiply it by 13. What is very important is that we take revenue, not sales. If that is an annual subscription or annual enterprise contract, we prorate this across 12 months. It is only live revenue. We are not taking three-year enterprise deals and baking them into a $1BN figure.” @alexmashrabov Love to hear your thoughts @BR_Murray @cjgustafson222 @Kellblog @poyark
中文: 有争议的问题:你如何真正计算收入? 我们审视过去28天的收入，并将其乘以13。 非常重要的是，我们需要的是收入，而不是销售。如果这是一份年度订阅或年度企业合同，我们将在12个月内进行正文。 这只是活费收入。我们不会接受三年的企业合作协议，而是将其收入转化为10亿美元。@alexmashrabov 很喜欢听你的想法 @BR_Murray @cjgustafson222 @Kellblog @poyark
🎬
视频
#150 🔗 原推
2026-09-28 22:38:00
Harry Stebbings
When I was a kid, I used to love MTV Cribs where you get to see inside the houses of mega-wealthy rappers. This is that for venture nerds and cap tables!!! 😂😂
中文: 小时候，我常常喜欢MTV小说，在那里你可以看到超级富豪说唱歌手的家。 这是针对风向型和帽式表格的!!!😂😂
#149 🔗 原推
2026-09-28 21:27:54
Harry Stebbings
My biggest lessons in the journey to finding product-market fit “We spent more than a year in search of a product that could work. We burned more than $10M out of $16M raised in seed fundraising. I feel I am responsible because I was focusing on the wrong things. I think I just lost touch with reality back then. I was optimizing for what is hype today, what is the right narrative, how we can hijack the attention, everything instead of building a good product.” @alexmashrabov Love to hear your thoughts @mwseibel @ElenaVerna @bbalfour @hnshah
🎬
视频
#148 🔗 原推
2026-09-28 20:08:00
Harry Stebbings
Every single day in this role I think to @arampell: The battle between a startup and an incumbent is a race to see whether the startup achieves scale and distribution before the established company replicates or acquires its innovation. The race is on.
中文: 在这个角色中，我每天都会向@arampell 想: 初创企业与现有企业之间的竞争，是争分量，以确定这家初创公司在成熟公司复制或获得创新之前是否实现了规模和分销。 比赛开始了。
#147 🔗 原推
2026-09-28 19:46:51
Harry Stebbings
I have said this before. This is 2021 on steroids. Yes, amazing companies are being built, faster than ever. But 99% are not and venture has never been less disciplined. This will not end well for most.
中文: 我之前说过这个。 这是2021年关于类固醇的问题。 是的，正在建设着令人惊叹的公司，速度比以往任何时候都快。 但99%的人没有，而且冒险的纪律性从未如此严格。 这对大多数人来说不会是好结局。
#146 🔗 原推
2026-09-28 19:13:38
Harry Stebbings
The power of the immigrant founder “My parents are from Uzbekistan. If a family of five people makes $1,000 a month, it is considered to be wealthy. Since I was eight, my parents told me that I must get to the United States because this is the place where technology matters. My mother had to work three jobs because my education was to compete in programming competitions and go to educational camps where I could learn from the best.” @alexmashrabov Love to hear your thoughts @amasad @deedydas @immad @aarthir
中文: 移民创始人的力量 我的父母来自乌兹别克斯坦。如果一个五口之家每月赚1000美元，那就被认为是富有的。 从我八岁起，父母就告诉我，我必须去美国，因为那里是科技重要的地方。 我母亲不得不做三份工作，因为我的教育是参加编程竞赛，去参加教育夏令营，在那里我能从最优秀的人那里学习。@alexmashrabov 喜欢听你的想法 @amasad @deedydas @immad @aarthir
🎬
视频
#145 🔗 原推
2026-09-28 17:15:00
Harry Stebbings
Spotify 👉 https://open.spotify.com/episode/6bsD2IoeAZlruFexYbX2xg Youtube 👉 https://youtu.be/jszn8rFtxm4 Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-%241bn-arr-in-18-months-the-untold-story/id958230465?i=1000791979678
中文: Spotify 👉 YouTube 👉 苹果播客 👉
#144 🔗 原推
2026-09-28 14:15:58
Harry Stebbings
Higgsfield is the most untold story in tech. $1BN in ARR in 18 months. Faster than everyone other than OpenAI and Anthropic. They spend $4M a month on models. They expect this to be $100K per person per month. They have 150 people working in a content machine. They will breed more millionaires than any other company in Kazakh history. For the first time, @alexmashrabov on the journey to $1BN in ARR. (below) 1. The Power of the Immigrant Founder Coming from Uzbekistan, Alex was pushed into competitive programming at age eight as his single path to reach the United States. For international founders, placing top in global competitions serves as the ultimate social elevator, instilling the relentless work ethic required to build breakout companies. 2. My Biggest Lessons in the Journey to Finding Product-Market Fit @higgsfield burned over $10 million of its $16 million seed round chasing hype and narrative rather than product quality. With under $5 million left, the team pivoted to product-led growth, solving camera control for creative directors, which immediately triggered organic hypergrowth without paid ads. 3. The 150-Person Content Team Powering Higgsfield's Billion in ARR Nearly half of Higgsfield's workforce consists of 150 in-house creative professionals producing tutorials, ads, and cinematic projects. Generating 90 minutes of TV-quality AI video requires 100 hours of raw output, proving human taste and curation remain the primary drivers of distribution. 4. We Spend $4 Million per Month on Models Higgsfield spends $4 million monthly on internal model usage, averaging $10,000 per employee so teams can freely vibe code and test workflows. Uncapped inference compute acts as a force multiplier, allowing top talent to discover breakthroughs at maximum velocity. 5. Why Chasing Benchmarks Is Bullshit and the Corporate Misalignment Occurring Public benchmarks have devolved into corporate psyops where lab researchers overfit test data to secure bonuses before job-hopping. Text-to-video benchmarks ignore real production workflows requiring 3,000-word prompts, proving direct customer iteration beats artificial leaderboards. 6. Why Team Sizes Won't Be Impacted as Much as People Think While AI handles over 60% of basic support requests, complex B2B environments cannot eliminate human teams. High product velocity constantly shifts rules and context, requiring smart, coordinated operators across legal and customer success. 7. Americans Are Way More Promiscuous When It Comes to Leaving Companies Silicon Valley workers routinely jump jobs every two years, prioritizing short-term trends over deep commitment. This transactional market gives international hubs an advantage, where cultural loyalty and team stability build compounding technical moats. (links in comments)
#143 🔗 原推
2026-09-28 14:15:30
Harry Stebbings
Some fricking wild stats from our show coming tomorrow with @alexmashrabov: - $1BN ARR in 18 months, faster than Cursor and Cognition. - Higgsfield spend $4M per month on different models - That is over $10K per head. - Alex expects to spend $50-$100K per month on tokens for his 10x people. - One engineer spent $30K in a single week on Astra vibe coding. WTF!! - They have a team of 150 people in content alone. The story that has not been told, coming tomorrow...
media 0 共 1 项
#142 🔗 原推
2026-09-27 21:50:48
Harry Stebbings
I just met a fan who said they love 20VC because it's like, "If Dwarkesh, Invests like the best and TBPN had a baby, it would be 20VC." I quite like that summary.
中文: 我刚刚遇到一位粉丝，他们说他们很喜欢20VC，因为“如果是Dwarkesh，投资者就是最好的，TBPN生了宝宝，那就是20VC。” 我非常喜欢那个总结。
#141 🔗 原推
2026-09-27 12:07:44
Harry Stebbings
Higgsfield have scaled to $1BN in ARR in 18 months; that is faster than Cursor. But did you know they have over 150 people in content production alone? For every product release, funding announcement, you name it, they have built a content machine that is unmatched. The next generation of companies will have content baked into their DNA.
中文: 希格斯菲尔德在18个月内以ARR的价格达到10亿美元;这比Cursor更快。 但你知道他们仅在内容制作方面就有超过150人吗? 每发布一次产品，发布资金，你的名字就是你的名字，他们打造了一台无与伦比的内容机器。 下一代企业将把内容融入他们的基因。
#140 🔗 原推
2026-09-26 15:13:58
Harry Stebbings
These posts are the single best posts on X when Paul does them. Passion returning $2BN on $18M is simply wild.
中文: 这些帖子是保罗做这些文章时在X上最好的一篇。 以1800万美元回报20亿美元的热情简直是疯狂。
#139 🔗 原推
2026-09-26 12:42:10
Harry Stebbings
Amazon is right to block Muse. “One, you do not get any revenue from your ad business, and Amazon’s ad business is now larger than their e-commerce profits. The second thing is the basket size gets reduced. If I do this, I just order the thing. If I block them, they will probably come to me anyway, because I am Amazon. So I have leverage.” @rodriscoll Love to hear your thoughts @juokaz @FredaDuan @scotwingo @harleyf
中文: 亚马逊阻止缪斯是正确的。 首先，你的广告业务收入不大，而亚马逊的广告业务现在已超过其电子商务利润。 第二点是篮子尺寸会减小。如果我这样做，我就直接点东西。 如果我屏蔽它们，它们很可能还是会来找我，因为我是亚马逊。所以我有筹码。” @rodriscol 很喜欢听你的想法 @juokaz @FredaDuan @scotwingo @harleyf
🎬
视频
#138 🔗 原推
2026-09-24 23:12:00
Harry Stebbings
Today Higgsfield hit $1BN in ARR. Monday next week our 20VC with @alexmashrabov is released. The story, as has never been told before 👇 https://twitter.com/HarryStebbings/status/2103248038227480838/photo/1
中文: 今天，希格斯菲尔德的ARR达到了10亿美元。 下周周一，我们与@alexmashrabov合作的20VC将发布。 这个故事，正如此前从未被讲述过的👇
media 0 共 1 项
#137 🔗 原推
2026-09-24 22:19:43
Harry Stebbings
Why Muse and Instinct are the biggest credible threats to ChatGPT “Muse is a Trojan horse to fight ChatGPT because the LLM is pretty good. It is a darn good, normal consumer-grade LLM. If it is free, it has agents that are truly autonomous, which ChatGPT does not. It does everything ChatGPT can do, and it has autonomous agents. It does not have to just be agents. It is doing all of it.” @jasonlk Love to hear your thoughts @srcasm @wailord @jgreze @mignano
中文: 为何Muse和Instinct是ChatGPT面临的最大可信威胁 缪斯是一匹与ChatGPT对抗的木马，因为LLM相当不错。 它是一种不错的普通消费级LLM。如果它是免费的，它的代理是真正自主的，而ChatGPT则没有。 它能完成ChatGPT的所有用途，并且具有自主代理。不必只是代理人。它正在做这一切。”@jasonlk 很喜欢听你的想法 @srcasm @wailord @jgreze @mignano
🎬
视频
#136 🔗 原推
2026-09-24 21:08:00
Harry Stebbings
Spotify 👉 https://open.spotify.com/episode/47N1rZwXfPZH7JSb6qhK0v Youtube 👉 https://youtu.be/5FnXlCQxV5o Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-metas-muse-hits-no-1-chatgpt-finally-has-a/id958230465?i=1000791420110
中文: Spotify 👉 YouTube 👉 苹果播客 👉
#135 🔗 原推
2026-09-24 18:58:38
Harry Stebbings
90% of podcasts you listen to will actually not make you smarter. Today’s show is hotter than Meta stock… - Meta's Muse Hits No. 1. - ChatGPT Finally Has a Rival - Menlo Sounds the AI Bubble Alarm - Factory Triples Its Valuation to $5 Billion - Keith Rabois vs Airwallex: Who is Right? - Crusoe's $3.9 Billion Round. Is the Data Centre Trade Overheating? Imagine Gavin Baker, Squawkbox and Jerry Springer had a baby… that’s this episode today 🤣 1. Why Anthropic pushed its $2 trillion IPO from October to November Anthropic delaying its IPO from October to November is a tactical move to present a clean Q3 prospectus. Rather than listing while Q3 metrics are unaudited, waiting lets bankers include fully audited, record-setting October numbers so the financial results speak for themselves. 2. Why Muse and Instinct are the biggest credible threats to ChatGPT Autonomous agent interfaces like Muse act as a Trojan horse against ChatGPT by bundling solid consumer LLMs with free autonomous agent execution. Offering daily conversational AI and action-taking agents for free with higher token limits makes paying $20 monthly for basic chat hard to justify. 3. Amazon is right to block Muse Amazon blocking Meta's agentic commerce, Muse, defends its core profit model. Autonomous agents bypass sponsored ad listings, Amazon's main e-commerce profit driver, while shrinking total basket sizes. Amazon's massive consumer leverage forces Meta to negotiate a value exchange rather than freely harvesting its store infrastructure. 4. What should the API policy be for every big company? Aggregated consumer demand dictates tech adoption far more than theoretical payment protocols. When platforms like Meta aggregate massive user volume against backend systems, incumbents like Amazon, OpenTable, and Resy must immediately formulate agent API policies. Demand aggregation remains the ultimate driver of enterprise urgency. 5. Why I would make the investment into Factory at 5 billion Investing in Factory at $5 billion is a high-conviction bet on data sovereignty and model choice. C-level executives deeply distrust frontier labs with confidential codebases, fearing competitive data leaks through shared LLMs. As coding inference surges, air-gapped, model-agnostic enterprise platforms will capture massive market share. 6. Why would Rory invest in Factory at $5 billion? Software coding is the mother lode of AI value creation. While corporate boards demand rapid AI adoption, enterprises prefer buying end-to-end dev tools from independent players like Factory rather than frontier model labs whose data retention policies pose existential IP risks. 7. Why Jack at Airwallex needs an army of people fighting for him Running a decacorn requires a dedicated bench of public advocates to disarm narrative attacks. Founders shouldn't waste executive bandwidth arguing on social media against public attacks. Mobilizing an army of vocal supporters lets CEOs focus strictly on execution while third parties manage reputation. (links in comments)
中文: 你收听的播客中，有90%实际上不会让你变得更聪明。 今天的节目比Meta的股价更热...... - Meta的缪斯乐队排名不中。1. - ChatGPT 终于有了竞争对手 - 门洛发出人工智能气泡警报 - 工厂估值增长三倍，达到50亿美元 - 基思·拉博伊斯 对阵 艾尔瓦尔克斯:谁是对的? - 克鲁索的39亿美元轮融资。数据中心贸易是否过热? 想象一下，加文·贝克、斯阔博克斯和杰瑞·斯普林格生了个宝宝......今天就是这一集🤣 1。为什么Anthropic将其2万亿美元的IPO从10月推迟至11月 Anthropic将首次公开募股推迟至10月，此举是为呈现一份清白的第三季度招股说明书而采取的战术举措。等待银行在未经审计的情况下，不会在第三季度公布时列出全面审计、创纪录的10月份数据，以便财务结果能够自行衡量。 2。为何Muse和Instinct是ChatGPT面临的最大可信威胁 像Muse这样的自主代理接口通过将稳健的消费类LLM与免费的自主代理执行相结合，作为对抗ChatGPT的特洛伊木马。免费提供每日对话式人工智能和行动型代理，并增加代币使用限额，使得每月支付20美元用于基本聊天，难以证明其合理性。 3。亚马逊阻止缪斯是正确的 亚马逊阻挠Meta的代理业务Muse，为其核心盈利模式辩护。自主代理绕过赞助广告列表，亚马逊的主要电商盈利驱动力，同时缩小了总篮子规模。亚马逊庞大的消费杠杆迫使Meta需要协商建立价值交换，而不是自由地获取其门店基础设施。 4.每家大公司都应该制定哪些API政策? 消费者需求的聚合决定技术采用的远不止于理论上的支付协议。当像Meta这样的平台在后端系统上聚合大量用户量时，像Amazon、OpenTable和Resy这样的现有公司必须立即制定代理API策略。需求聚合仍然是企业紧迫性的最终驱动力。 5。为什么我会投资50亿的工厂 以50亿美元投资工厂是对数据主权和模型选择的高说服力投资。C级高管对拥有机密代码库的前沿实验室深感不信任，担心通过共享的LLM泄露具有竞争性的数据。随着编程推理的激增，与模型无关的企业平台将占据巨大的市场份额。 6.为什么罗里会投资50亿美元的工厂? 软件编码是人工智能价值创造的主人。尽管企业董事会要求快速采用人工智能，但企业更倾向于从Factory等独立企业购买端到端开发工具，而不是采用数据留存政策带来存在风险的前沿模型实验室。 7.为何杰克在艾尔沃雷克斯需要一支为他而战的队伍 经营一个十项式的事件需要专门的公众倡导者来解除叙事攻击。创始人不应浪费高管在社交媒体上针对公众攻击的争论。动员大批发声的支持者，让首席执行官能够严格专注于执行，而第三方则负责管理声誉。 (评论中的链接)
#134 🔗 原推
2026-09-24 18:58:31
Harry Stebbings
In the last 48 hours I have done shows with: - @ChaseLochmiller @CrusoeAI - @paraga @p0 - @alexmashrabov @higgsfield_ai 11 years in, this remains the best job in the world. 👇 https://twitter.com/HarryStebbings/status/2103165459507900551/photo/1
中文: 在过去48小时内，我用以下内容做了节目: - @ChaseLochmiller @CrusoeAI - @paraga @p0 - @alexmashrabov @higgsfield_ai 11年过去了，这仍然是世界上最好的工作。👇
media 0 共 4 项 media 1 media 2 media 3
#133 🔗 原推
2026-09-24 16:51:35
Harry Stebbings
@jasonlk @rodriscoll Spotify 👉 https://open.spotify.com/episode/47N1rZwXfPZH7JSb6qhK0v Youtube 👉 https://youtu.be/5FnXlCQxV5o Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-metas-muse-hits-no-1-chatgpt-finally-has-a/id958230465?i=1000791420110
中文: @jasonlk @rodriscoll Spotify 👉 YouTube 👉 苹果播客 👉
#132 🔗 原推
2026-09-24 16:28:40
Harry Stebbings
Here are the biggest news I’ve discussed with @jasonlk and @rodriscoll this week: - Meta's Muse Hits No. 1. ChatGPT Finally Has a Rival - Menlo Sounds the AI Bubble Alarm - Factory Triples Its Valuation to $5 Billion - Keith Rabois vs Airwallex: Who is Right? - Crusoe's $3.9 Billion Round. Is the Data Centre Trade Overheating? I’ve condensed my notes below: 1. Why Anthropic pushed its $2 trillion IPO from October to November Anthropic delaying its IPO from October to November is a tactical move to present a clean Q3 prospectus. Rather than listing while Q3 metrics are unaudited, waiting lets bankers include fully audited, record-setting October numbers so the financial results speak for themselves. 2. Why Muse and Instinct are the biggest credible threats to ChatGPT Autonomous agent interfaces like Muse act as a Trojan horse against ChatGPT by bundling solid consumer LLMs with free autonomous agent execution. Offering daily conversational AI and action-taking agents for free with higher token limits makes paying $20 monthly for basic chat hard to justify. 3. Amazon is right to block Muse Amazon blocking Meta's agentic commerce, Muse, defends its core profit model. Autonomous agents bypass sponsored ad listings, Amazon's main e-commerce profit driver, while shrinking total basket sizes. Amazon's massive consumer leverage forces Meta to negotiate a value exchange rather than freely harvesting its store infrastructure. 4. What should the API policy be for every big company? Aggregated consumer demand dictates tech adoption far more than theoretical payment protocols. When platforms like Meta aggregate massive user volume against backend systems, incumbents like Amazon, OpenTable, and Resy must immediately formulate agent API policies. Demand aggregation remains the ultimate driver of enterprise urgency. 5. Why I would make the investment into Factory at 5 billion Investing in Factory at $5 billion is a high-conviction bet on data sovereignty and model choice. C-level executives deeply distrust frontier labs with confidential codebases, fearing competitive data leaks through shared LLMs. As coding inference surges, air-gapped, model-agnostic enterprise platforms will capture massive market share. 6. Why would Rory invest in Factory at $5 billion? Software coding is the mother lode of AI value creation. While corporate boards demand rapid AI adoption, enterprises prefer buying end-to-end dev tools from independent players like Factory rather than frontier model labs whose data retention policies pose existential IP risks. 7. Why Jack at Airwallex needs an army of people fighting for him Running a decacorn requires a dedicated bench of public advocates to disarm narrative attacks. Founders shouldn't waste executive bandwidth arguing on social media against public attacks. Mobilizing an army of vocal supporters lets CEOs focus strictly on execution while third parties manage reputation. (links in comments)
#131 🔗 原推
2026-09-24 16:28:30
Harry Stebbings
Top three rules to crushing pre-interview: 1. Do an astonishing amount of prep. We do 10+ ref calls per episode. 2. Never have a pre-interview calls. Total BS. 3. Send the schedule before. People can prepare thoughts. This isn’t a game of tricking people with hard questions.
中文: 前面试前最严厉的三条规则: 1。做大量准备。每集进行10次以上的回复。 2。从来没有面试前的电话。总计 BS。 3。先把时间安排寄出。人们可以准备想法。这并不是用棘手的问题欺骗别人的游戏。
#130 🔗 原推
2026-09-23 19:47:15
Harry Stebbings
The venture market right now is more frothy than it has ever been. Yes, outcomes are larger than ever. Yes, it has never been a more exciting time. But three rounds in three weeks is not cool, it’s unhealthy. $5BN for an idea, fine for a genius, not when there are 100. The Ponzi scheme is off the charts right now. And all investors know it behind the scenes.
中文: 目前的风险投资市场比以往任何时候都更为泡沫。 是的，结果比以往任何时候都要大。 是的，这从未是一个更令人兴奋的时刻。 但三周内三轮并不酷，而且不健康。 一个想法需要50亿美元，天才可以没问题，而不是在有100个想法时。 庞氏骗局目前已不在排行榜上。 所有投资者都在幕后了解这一点。
#129 🔗 原推
2026-09-23 15:09:58
Harry Stebbings
Muse and Instinct are the first credible threats to ChatGPT since launch. It is accessible. It is everywhere consumers already are. The model performance is great for consumer use cases. Bad for OpenAI.
中文: 自推出以来，Muses 和 Instinct 是 ChatGPT 面临的首个可信威胁。 可访问。消费者已经无处不在。 模型性能非常适合消费者使用案例。 对OpenAI不利。
#128 🔗 原推
2026-09-22 21:05:42
Harry Stebbings
"Moats" are the most BS thing in startups. No such thing exists when you start. Google, Apple, Meta can always "just build it". Defensibility is built in increments. Small, immensely focused teams outpace the biggest companies on earth. Investors, get over the moat obsession.
#127 🔗 原推
2026-09-22 12:59:08
Harry Stebbings
So jokes, sent an incredible entrepreneur to two VCs; one in London, one in SF. SF VC met 3 hours later and has partner meeting tomorrow. London VC hasn't responded to the intro yet. Speed is a feature not a bug.
中文: 笑话，把一位了不起的企业家送到了两家风险投资公司，一家在伦敦，一家在旧金山。 SF VC 在三小时后相识，明天将举行合作伙伴会议。 伦敦风险投资公司尚未对该介绍作出回应。 速度是一个功能，而不是错误。
#126 🔗 原推
2026-09-22 10:00:02
Harry Stebbings
I have never had an investment hit $500M in revenue in such a short time as fomo. One of the most under-discussed companies today.
中文: 在如此短的时间内，我的投资从未达到过5亿美元的收入。 当今讨论最少的公司之一。
#125 🔗 原推
2026-09-21 22:31:17
Harry Stebbings
@danieldines @UiPath Spotify 👉 https://open.spotify.com/episode/6RhLqVjA1htI3yl4Vpp4tK Youtube 👉 https://youtu.be/N9U-RoNXYZE Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-why-ai-cannot-replace-humans-in-enterprise-why/id958230465?i=1000790894201
#124 🔗 原推
2026-09-21 14:18:20
Harry Stebbings
“Chinese labs should be classified as the good guys. AI is not ready to replace humans in enterprise.” For all the hyperbole and BS we hear, very few speak their truth like @danieldines. He is a dear friend and incredible entrepreneur (@UiPath) speaking so openly today, the way only true friends can. I asked if he would be ok with me sharing my notes from the discussion. Attached below. 1. Why Chinese AI Labs Should Be Classified as the Good Guys Chinese AI labs should be viewed as part of the constructive global AI ecosystem rather than automatically villainized. Calls to “pace the frontier” under the banner of safety can also restrict open-source development, while genuinely bad actors are unlikely to comply with voluntary coalitions anyway. 2. Managing Enterprise Workforce Transformation Without Mass Panic Founders and executives should be honest with employees about AI-driven transformation without blindly cutting headcount. Task execution is only one part of a job. Rushing to replace people ignores the relationships, institutional trust, and customer connections that often hold an enterprise together. 3. Why Jensen Huang and Nvidia Depend on Open Source to Thrive Nvidia’s long-term success is closely tied to a thriving open-source model ecosystem. If a closed duopoly captures the market, those frontier labs will have every incentive to build proprietary silicon, potentially threatening Nvidia’s core hardware dominance. 4. Why the Map of Work Holds More Value Than Interchangeable Models AI models are increasingly becoming interchangeable infrastructure. The more defensible value lies in the “map of work”: the institutional knowledge, exact workflows, and edge-case exceptions that define how a specific enterprise actually operates. 5. The Critical Difference Between Model Memory and True Learning on the Job Loading context or notes into a model’s scratchpad is fundamentally different from human learning. Humans change through experience and adapt their internal judgment, while current models do not update their weights on the job, limiting their ability to develop true initiative. 6. What Enterprises Actually Fear When Working With Frontier Model Labs Large enterprises are less concerned that frontier labs like OpenAI will enter their specific verticals than they are about exposing sensitive source code and operational data. The bigger fear is that proprietary IP sent to external model providers could somehow leak to existing competitors. 7. Why Vibe Coding Breaks Down When Moving From Prototype to Production AI coding agents have made building early prototypes incredibly fast, but prototyping is not where the hardest engineering work begins. Production software requires maintaining connectors, enforcing security, managing permissions, and building rigorous test suites that vibe-coded applications often struggle to handle. (links in comments)
#123 🔗 原推
2026-09-21 14:18:07
Harry Stebbings
Said with pure desire to learn: How does a frontier model provider go public when they can’t answer the question of liability? Swarms of rogue agents commit terrible acts. Who is liable? Silence in the room. How can there be an IPO? Truly want to learn.
中文: 以纯粹的学习欲望说道: 当前沿模型提供商无法回答责任问题时，他们该如何公开? 一群流氓特工犯下了可怕的行为。谁应承担责任? 房间里一片寂静。 怎么可能进行IPO? 真心想学习。
#122 🔗 原推
2026-09-21 13:34:48
Harry Stebbings
Why I would not invest in the new Instinct round “This is the threat that every VC worried about, and we all got a hall pass since the start of AI because the LLMs did not build any apps. This is the one that they are building. Instinct is slow. That is a sign of compute costs. That is why they have to raise $1BN. I worry when the incumbent has infinite capabilities here and wants to build the app. That is why I would say no, but I might be wrong.” @jasonlk Love to hear your thoughts @FundamentEdge @wailord @pranavreddy @nabeelqu
中文: 为什么我不会投资新的本能回合 这是每位风险投资公司都担心的威胁，而且自从人工智能启动以来，我们都获得了一次通道通行证，因为LLMs没有开发任何应用程序。这是他们正在建设的。 本能是缓慢的。这是计算成本的一个标志。这就是为什么他们必须筹集10亿美元。 我担心现任者在这里拥有无限功能，并希望构建应用程序。这就是为什么我会说不，但我可能错了。@jasonlk 很喜欢听你的想法 @FundamentEdge @wailord @pranavreddy @nabeelqu
🎬
视频
#121 🔗 原推
2026-09-18 15:44:00
Harry Stebbings
Why Matteo might just win with Muse “As software, it is very, very good. It instantly works. This is the definition of great software. Meta already has the infrastructure, and this is running on its own LLM. From an infrastructure perspective, almost no one can compete. That is why it is fast. That is why it works well. You get all the VM, all the infrastructure, all the storage, and they have their own LLM.” @jasonlk Love to hear your thoughts @bigT_sheesh @wailord @hwchase17 @bernhardsson
🎬
视频
#120 🔗 原推
2026-09-17 23:55:00
Harry Stebbings
Why is it overblown and not nearly as dramatic as Anthropic makes it? “If the feds really thought there was someone in downtown San Francisco building a technology that had a 10% chance of blowing up the world, they would shut it down. Replace the word AI with, ‘We are building a nuclear reactor. It is totally safe right now, but there is a 10% chance it goes wrong in the next five years and blows up the world.’ Either the US government is asleep, which I doubt, or it is looking at this going, ‘This is a bunch of excited teenagers. We will step in later if it gets crazy.’” @rodriscoll Love to hear your thoughts @alexolegimas @deanwball @notRichardRen
中文: 为什么它被夸大了，而不像人类那样夸张? 如果联邦政府真的以为旧金山市中心有人在制造一种技术，而这项技术有10%的可能性会炸毁世界，那么他们就会将其关闭。 将“AI”一词替换为“我们正在建造一座核反应堆”。目前完全安全，但未来五年内出现问题并彻底炸毁世界的可能性为10%。 要么美国政府睡着了，我对此表示怀疑，要么就是看着这种情况发生，“这是一群兴奋的青少年。”如果这会让我们发疯，我们稍后再介入。” @rodriscol 喜欢听你的想法 @alexolegimas @deanwball @notRichardRen
🎬
视频
#119 🔗 原推
2026-09-17 21:25:00
Harry Stebbings
IS Dario simply being strategic ahead of an IPO with his pacing argument “My view was this was just a risk factor in an S-1 done live. Anthropic is going public, and he is just getting ahead of a risk factor so that when the $2TN IPO happens, it is a non-issue. We are going to debate it as a society, and so when we go on the roadshow to New York and everywhere else, no one cares.” @jasonlk Love to hear your thoughts @ByrneHobart @EricNewcomer @GavinSBaker @TheZvi
中文: 在首次公开募股前，他只是在策略上表现出色，且主张平价 我认为这只是完成S-1生活中的一个风险因素。 Anthropic 正在上市，他刚刚领先于一个风险因素，因此当 $2TN 首次公开募股发生时，它就不再是问题。 我们将作为一个社会来讨论这个问题，因此当我们前往纽约及其他任何地方进行路演时，没有人会在意。” 很喜欢听你的想法 @ByrneHobart @EricNewcomer @GavinSbaker @TheZvi
🎬
视频
#118 🔗 原推
2026-09-17 18:31:00
Harry Stebbings
@jasonlk @rodriscoll Spotify 👉 https://open.spotify.com/episode/2CYdHhU2iXhwkUm4Ws8OXt Youtube 👉 https://youtu.be/_tfUQrusRGM Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-why-pacing-the-frontier-is-bs-instinct-raising/id958230465?i=1000790315549
中文: @jasonlk @rodriscoll Spotify 👉 YouTube 👉 苹果播客 👉
#117 🔗 原推
2026-09-17 14:16:16
Harry Stebbings
The biggest news in tech this week: - Dario Calls to "Pace the Frontier" - Instinct Raising $1BN at $10BN & Meta Launches Muse - Miro Sells for $1.35BN After a $17.5BN Valuation - Mistral Raises €3BN My notes below with @jasonlk and @rodriscoll: 1. Summary of the Problems and the Solutions Cyber risk is real, mass economic unemployment feels overblown, and losing control of recursive agents is difficult to assess. Meanwhile, proposed solutions range from unlikely to practically impossible to implement. 2. Why Is It Overblown and Not Nearly as Dramatic as Anthropic Makes It? If the government truly believed a San Francisco lab had a 10% chance of destroying humanity, it would intervene immediately. The lack of intervention suggests authorities do not view extreme existential risk claims as an imminent threat. 3. Why Is It Bullshit, Dario, to Think We Will Have International Alignment on AI? Global coordination on frontier AI ignores geopolitical reality. Getting democratic allies to agree is difficult enough, while expecting enforceable alignment with China and Russia makes international pacing agreements even harder to imagine. 4. Is the Personal Assistant Market a Market Where You Can Build a Standalone Company? Standalone AI assistants face a fundamental question: can they build durable businesses while valuations race into the billions and platform incumbents pursue the same market? 5. Why Meta Might Just Win With Muse Meta has a structural advantage because it owns the infrastructure and underlying LLM. While competitors spend $3 to $4 per user on compute, Meta can deliver fast, capable agentic workflows without the same infrastructure burden. 6. Is Dario Simply Being Strategic Ahead of an IPO With His Pacing Argument? One interpretation is that Anthropic’s pacing argument is effectively an S-1 risk factor discussed in public. By addressing existential risk now, leadership can make it a debated, understood issue rather than a roadshow problem later. 7. Why I Would Not Invest in the New Instinct Round Standalone AI assistants face brutal economics when incumbents have their own models, infrastructure, and enormous resources. Subsidizing $5 to $10 per user can turn millions of users into a massive annual compute bill that venture capital must keep funding. (links in comments)
#116 🔗 原推
2026-09-17 14:15:58
Harry Stebbings
Why talking about returns without a timeline attached means nothing “There is a rule in our office that you are not allowed to talk about returns without also talking about time. It is very common on the private side to say you are up 2X, 3X, 5X, whatever. That tells you nothing. If you are up 5X over 30 years, that is horrible. If you are up 5X in five months, that is amazing.” @CIO_Baylor
中文: 为何在没有附加时间线的情况下谈论回报毫无意义 我们办公室有一条规定，即你不得在不谈论时间的情况下谈论回报。 在私人方面，说你的2倍、3倍、5倍的起重都很常见。这什么也不告诉你。 如果你在30年内增加5倍，那就太糟糕了。如果你在五个月内能增加5倍，那就太神奇了。@CIO_Baylor
🎬
视频
#115 🔗 原推
2026-09-16 21:01:00
Harry Stebbings
Why is velocity of capital more important than multiple of capital? “Historically they were 10, 12-year funds. Now they are 15, 18-year funds. You get your money back in 15 or 18 years, and let’s say you are up 15X. If you were in a growth equity fund that was up 3X in six years, and you did that three times, over 18 years you would be up 27X. What we are really after is the velocity of capital, not just returns on capital. I am trying to optimize for the biggest pile of money for our students.” @CIO_Baylor
中文: 为什么资本的速度比资本的多重更重要? 从历史上看，它们是10年、12年期的基金。现在它们是15年、18年期的基金。你在15或18年内就能拿回你的钱，假设你的奖金增加了15倍。 如果你在一家在六年内上涨3倍的成长型股票基金中，并且你这样做了三次，那么在18年内，你的股价将上涨27倍。 我们真正追求的是资本的速度，而不仅仅是资本回报。我正努力为学生优化最大的资金。@CIO_Baylor
🎬
视频
#114 🔗 原推
2026-09-16 18:03:00
Harry Stebbings
You have a blank canvas. What do you do next? “If you are starting with a blank sheet and you are going to do privates, you really need to nail down the private side first. You need to figure out what sort of liquidity environment you can live with on the private side and determine what that allocation is going to be. You need to box it and set it aside, because it is really, really hard to move a private book around.” @CIO_Baylor
🎬
视频
#113 🔗 原推
2026-09-16 15:47:00
Harry Stebbings
I do not talk about this often, but when I was eight years old, my grandparents lost everything. Bailiffs were at the door of the house, and given a bag where you have 20 mins to put all our belongings into it. If it does not fit in the bag, you do not get to take it. Within 20 mins, lives are turned upside down. Seeing this happen to your family changes your worldview. It gave me a fire, a hunger, honestly, a ferocity that I would never let my family suffer like this again. Yesterday I had such a proud moment of showing my now 86-year-old grandparents 20VC HQ, the studios, etc etc. An emotional moment showing them what we are building. Them in the studio 👇
中文: 我很少谈论这件事，但当我八岁的时候，我的祖父母失去了一切。 包在房子门口，并给了一个袋子，你有20分钟时间把我们所有的物品都放进屋里。 如果包里不合，就不要拿。 不到20分钟，生活就被颠倒了。 看到这种事情发生在家庭中，就改变了你的世界观。 这让我感到火，一种饥饿，说实话，那种我再也不会让家人遭受这样的痛苦。 昨天，我有了一个值得骄傲的时刻，向现年86岁的祖父母展示20VC总部、工作室等。 一个感人的时刻，向他们展示了我们正在构建的内容。工作室里的👇
media 0 共 1 项
#112 🔗 原推
2026-09-16 10:13:33
Harry Stebbings
What Seems Crazy Today but Will Be Commonplace in Five Years “Probably your bot buying a cybercab that starts making money for you. You’ll buy a cybercab that makes money for you, and your bot will run it. That would be pretty sick.” @m_franceschetti You think this is possible with that 5 year timeline? What does no one see that everyone should know about autonomous? @bradtem @ARK_Tasha @jdouma
🎬
视频
#111 🔗 原推
2026-09-15 15:39:00
Harry Stebbings
I am genuinely asking because I do not understand: If we "pace AI" and China does not. What happens then? Not a trick question, legitimately do not understand how this can end well.
中文: 我真诚地在问，因为我不明白: 如果我们“与人工智能步道”，而中国不会。 那会发生什么? 问题并非棘手，而是完全不明白这怎么能好而结束。
#110 🔗 原推
2026-09-14 22:22:10
Harry Stebbings
We Do $100 Million on Email Marketing With No Employees “Before, we had an email marketing team. Now we have zero. Everything is done through AI. We were forced into it. Our team was down to two people, and when the person leading it was leaving, my co-founder Alexandra jumped in to see how she could use AI to make it work. Within three days, she built multiple bots that now run all our email marketing. So now we have a team of zero, and email marketing makes close to $100 million.” @m_franceschetti What does no one see about the future of email marketing that everyone should see @denk_tweets @jon4growth @ecomchasedimond @MattPRD
🎬
视频
#109 🔗 原推
2026-09-14 18:34:00
Harry Stebbings
How to Do Paid Marketing Really, Really Well “Go one by one. Start with Meta. It’s the biggest, and it’s the one you can prove fastest. Based on the size of the company, spend a certain amount per week. Set the cap at the CAC you want to achieve, and only scale if you’re within that CAC. Otherwise, you start throwing money out the window. Your growth people will always want to spend more, and then you lose control. You need to be disciplined now if you want to be successful next year. Otherwise, you’ll have to pull back, spend less, and stop growing year over year.” @m_franceschetti Love to hear your thoughts on this @TaylorHoliday @andrewjfaris @DenneyDara
🎬
视频
#108 🔗 原推
2026-09-14 16:16:00
Harry Stebbings
Spotify 👉 https://open.spotify.com/episode/65jcPltqpyFMIyMU4ngqZO Youtube 👉 https://youtu.be/U_ManIWISzw Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-how-lps-allocate-to-venture-in-2026-what-they/id958230465?i=1000789506935
#107 🔗 原推
2026-09-14 14:07:58
Harry Stebbings
The single biggest mistake VCs make is they do not build relationships with LPs in between funds. You need to know how the best allocators to venture in the world think: - What they like? - What they do not like? - What would make them pull from a fund? - What would make them double down in a fund? - How a fund can build urgency to get them over the line? And so so much more. The challenge is not many CIOs speak openly about this. And an even bigger challenge is many are not independent thinkers and tend to follow what others around them do. David Morehead is one of the most independent-thinking CIOs in the endowment fund world. He has scaled Baylor to $2.6BN and one of the most respected institutions. I sat down with @CIO_Baylor to understand what every VC needs to hear from LPs but does not. I summarised my notes below: 1. Even the Best Venture Returns Don’t Matter to Some Funds For mega-endowments managing $40 billion to $60 billion, writing $20 million checks into elite venture funds may simply not move the needle. Even an extraordinary 50x return that generates $1 billion has limited impact on the overall portfolio, pushing mega-LPs toward much larger platform allocations. 2. How This Endowment Made Millions Betting That Vibe Coding Would Not Replace Core Software When software stocks fell 50% to 60% on fears that AI “vibe coding” would make SaaS obsolete, Baylor looked at how enterprises actually operate. Traditional businesses are unlikely to replace mission-critical systems requiring 100% precision with AI that is only 93% accurate, turning the sell-off into a major buying opportunity. 3. How We Think About Position Sizing A $400,000 distribution from a 7x exit means little to a multi-billion-dollar endowment. Rather than focusing solely on fund size, Baylor works backward from underlying company exposure, targeting roughly $3M per portfolio company so major wins can generate a meaningful dollar impact. 4. Why Is Velocity of Capital More Important Than Multiple of Capital? A 15-18-year venture fund generating 15x can produce less compounded capital than 3 consecutive 6-year growth funds returning 3x each, compounding to 27x. Endowments need capital returned and redeployed quickly enough to maximize long-term compounding. 5. Why Talking About Returns Without a Timeline Attached Means Nothing A 5x return means nothing without knowing the timeframe. A 5x over 30 years is terrible, while a 5x over five months is extraordinary. Endowments therefore evaluate returns alongside duration, because the velocity of capital can matter just as much as the headline multiple. 6. Why We Learn Way More From Our Public Managers Than Our Private Managers Private venture managers can get caught up in speculative technology narratives, like predicting self-driving cars on every road within three years. Public-market managers often provide more grounded perspectives because their frameworks must account for immediate regulatory hurdles, adoption friction, and continuous price discovery. 7. You Have a Blank Canvas. What Do You Do Next? When constructing a multi-asset portfolio from scratch, institutional allocators should establish their private-market allocation first. Because illiquid investments constrain liquidity and future reallocation, allocators need to define strict risk boundaries and box off the private portfolio before deploying capital across public markets. (links in comments)
#106 🔗 原推
2026-09-14 14:07:37
Harry Stebbings
Single biggest mistake people make when requesting an intro... "I want to meet someone at Revolut..." This tells me nothing. 1. State the specific person at the company you want to meet. Provide a Linkedin link. 2. State the objective and what you want. 3. State in the intro request that you have sent a separate email to this one that can be forwarded on with a note. Increase your chances of getting the intro. Do the work. Do not be vague. It will 10x chance of intro happening.
中文: 人们在申请介绍时犯的最大错误...... 我想在Revolut见一个人...... 这什么都没告诉我。 1。说明您要联系的公司的具体人员。提供链接链接。 2。明确目标和你想要什么。 3。在介绍请求中，您已向此发送了一封单独的邮件，可转发该邮件并发送一张便条。 增加你获得介绍的机会。 做这项工作。不要含糊其辞。 这将是介绍发生的10倍概率。
#105 🔗 原推
2026-09-14 12:03:19
Harry Stebbings
How Eight Sleep Partnered With Charles Leclerc “Charles already had our product for two years before we met, and we didn’t even know. That’s the best way to start a relationship. I’m an obsessed F1 fan. Two years before he reached F1, when he was racing in F2, I reached out to his team saying, ‘I’m a huge Charles fan. I think he’ll make it big. Can Eight Sleep sponsor him?’ But we were too tiny to put together a real deal. When I finally met him, I still had that email. I shared our medical vision. He really cares about that, and he’s incredibly into AI. By the end of dinner at the Miami GP, we said we’d work together. The deal was done in a matter of days.” @m_franceschetti
中文: 八人睡眠如何与查尔斯·勒克莱尔合作 查尔斯在见面前已经拥有我们的产品两年了，我们甚至都不知道。这是开始一段关系的最佳方式。 我是一个痴迷的F1粉丝。两年前，当他在F2赛马场上冲球时，他才进入F1，我联系了他的车队，说:“我是查尔斯的忠实粉丝。”我觉得他会把它做大。八个睡眠能赞助他吗?但我们太小了，无法达成真正的协议。 当我终于见到他时，我依然有那封邮件。我分享了我们的医学愿景。他非常关心这一点，而且他非常关注人工智能。 晚餐结束时，我们在迈阿密全科医生诊所吃完晚饭，说我们会一起努力。这笔交易在几天内就完成了。” @m_franceschetti
🎬
视频
#104 🔗 原推
2026-09-14 00:14:00
Harry Stebbings
Eleven years ago I was introduced to @a_zatarain about sponsoring a 20VC episode. She did not know but this payment was used to pay for some of my mother’s MS treatment. The partnership happened and a friendship blossomed between me, Alexandra and Matteo. I remember 10 years ago, everyone thought they were crazy. Italian CEO, not an engineer, doing hardware with a supply chain in China. Haha, good luck. How wrong they were. A multi-BN business today doing hundreds of millions in revenue. An insanely fun chat with my friend @m_franceschetti 👇 1. How to Do Paid Marketing Really, Really Well Disciplined paid marketing means scaling one channel at a time, starting with Meta, while maintaining a strict cap on customer acquisition cost (CAC). Overspending destroys unit economics and eventually forces pullbacks that hurt year-over-year growth. Holding firm on CAC today creates the foundation for predictable, compounding growth tomorrow. 2. The Story Behind How Eight Sleep Partnered With Charles Leclerc The best partnerships come from authentic product adoption and long-term conviction. Eight Sleep tried to sponsor Charles Leclerc years before he reached F1. When they eventually met, Leclerc was already an unprompted Eight Sleep user, allowing a shared vision around AI and performance to turn into a partnership within days. 3. We Do $100 Million in Email Marketing With No Employees Operational constraints can force breakthrough automation. When Eight Sleep’s email lead left, co-founder Alexandra built AI bots in three days to manage the entire workflow. Today, the channel operates with zero dedicated employees while generating nearly $100 million in revenue. 4. What Seems Crazy but Will Be Very Commonplace in Five Years’ Time? Autonomous assets managed by personal AI agents could create entirely new forms of passive income. Within five years, an individual’s AI agent could buy an autonomous cybercab, manage its daily operations, and generate revenue with minimal human involvement. 5. Why Angel Investing as a Founder Has Taught Me So Much About Company Building Angel investing can become a powerful market intelligence engine for active founders. Reading investor updates from non-competing startups provides real-time insight into valuation shifts, hiring profiles, and acquisition channels that are working, all of which can directly inform operating decisions. 6. With AI, Every Company Will Turn Into a Holding Company With Multiple Products AI-powered internal tooling dramatically lowers the cost and friction of launching new ventures, allowing single-product startups to evolve into conglomerates. Like Chinese giants such as Xiaomi, AI-native companies could use shared internal infrastructure to launch and manage as many as ten distinct business lines under a single holding company.
media 0 共 1 项
#103 🔗 原推
2026-09-13 14:10:00
Harry Stebbings
Spotify 👉https://open.spotify.com/episode/3S9UP1SfvTc3tujDM8nlxY?si=f191f9758655416d Youtube 👉https://youtu.be/KHwaNKfiASw Apple Podcasts 👉https://podcasts.apple.com/us/podcast/20vc-7-predictions-for-how-ai-changes-the-world/id958230465?i=1000789146856
#102 🔗 原推
2026-09-13 14:10:00
Harry Stebbings
There’s Going to Be No Financial Math You Can Use to Buy the Stock “It’s a portfolio and worldview bet. You have a company exploding in interest, with an early lead but not a ton of monetization. Your choices are to put money in at four or five billion, or to say, ‘It’s easy to clone, there are 10 more like it,’ and back one at 50 pre in the hope that Meta acquires it instead. Like Google early on, there’s going to be no financial math you can use to buy the stock. You’re just saying it’s a huge category, and it’s been proven that if you get enough traction, the monetization follows.” @rodriscoll When you are doing Instinct or Town, what do you tell yourself to quell the concern that Meta will absorb you @saranormous @chetanp @martinmignot
🎬
视频
#101 🔗 原推
2026-09-12 20:45:00
Harry Stebbings
We have no model companies. We have no data centres ready. We have energy 4x the price of anyone else. We let clowns rule our country. Time for a change Britain. Stop being so arrogant. Be a student. Learn from the US and China.
中文: 我们没有模范公司。 我们还没有数据中心准备就绪。 我们的能源价格是其他人的四倍。 我们让小丑统治我们的国家。 是时候改变英国了。 别再这么傲慢了。做个学生。向美国和中国学习。
#100 🔗 原推
2026-09-12 08:53:38
Harry Stebbings
The People Making Money Are Running Fastest and Evolving Quickest “My aha is that the prize goes to the companies that can evolve the quickest. In this market, the people making money are the ones running fastest and evolving quickest. That extra 10% of grind can have a massive payoff in a world where fortunes are being made in 12 to 24 months.” @rodriscoll What has been the single biggest needle mover in increasing shipping speed for you @karrisaarinen @adamguild @Romain_Lapeyre @jacsebl
中文: 赚钱的人跑得最快，发展得最快 我的愿望是，奖品会颁给那些能发展最快的公司。 在这个市场中，赚钱的人跑得最快，发展最快。 在一个财富在12到24个月内不断致富的世界里，额外增加10%的磨难可能会带来巨大回报。” 提高运速的单一最大转手是什么 @karrisaarinen @adamguild @Romain_Lapeyre @jacsebl
🎬
视频
#99 🔗 原推
2026-09-12 01:02:00
Harry Stebbings
Why the Oura IPO Will Be a Success “Oura is a somewhat understood consumer brand with 74% growth, which obviously probably can’t last forever. It’s the kind of thing people are going to want to buy. They understand it, and this isn’t 18% or 20% growth. There’s competition and downside, and maybe it’s Peloton 2.0, but for the moment, it’s pretty attractive. I think it’ll be a pretty successful IPO, which, at the margin, is good for everybody.” @jasonlk Love to hear your thoughts on this @mrsharma @msuster @AndrewDudum @euriekim
中文: 为何我们的IPO会取得成功 Oura 是一个增长率达到 74% 的消费者品牌，这显然无法永远持续下去。 这是人们想要购买的东西。他们理解这一点，而这并非18%或20%的增长。有竞争和不利因素，或许是Peloton 2.0，但目前它相当有吸引力。 我认为这将是一次相当成功的首次公开募股，在边际上对每个人都有好处。”@jasonlk 很喜欢听你对这个 @mrsharma @msuster 的看法 @AndrewDudum @euriekim
🎬
视频
#98 🔗 原推
2026-09-11 22:05:00
Harry Stebbings
Why the Majority of Neo Labs Will Not Be Good Investments “The two foundation models were Neo Labs themselves five years ago, and they’ve turned out to be the best venture bets of all time. Just because that’s true doesn’t mean the other 100 Neo Lab bets you can make today will also be amazing. Now you have other companies with the capital already and other companies with the distribution. The question is which of those bets will be orthogonal enough to the foundation model companies to be an interesting investment.” @rodriscoll Love to hear your thoughts on this @deedydas @EnoReyes @swyx @lilatretikov @PaulBonnet
中文: 为什么大多数Neono Labs投资不好 这两款基础模型五年前就是Noo Labs，它们被证明是自创最佳的投资模式。 事实并非如此，意味着你今天能进行的其他100次新实验室投注也将非常精彩。现在你已有其他公司与资金合作，其他公司也拥有该分销。 问题在于，哪些赌注对基础模型公司来说足以成为一项有趣的投资。 @rodriscol 很喜欢听听你对这个问题的看法:@deedydas @EnoReyes @swyx @lilatretikov @PaulBonnet
🎬
视频
#97 🔗 原推
2026-09-11 18:08:00
Harry Stebbings
When Someone Goes Risk-On, Everyone Goes Risk-On “Salesforce is the dominant SaaS company at $180 billion to $200 billion, and Service Cloud is maybe 25% of that, so a pure Service Cloud replacement is only worth 50. To justify $15 billion or $20 billion in market cap, Sierra can’t just build a slightly better next-generation Service Cloud. It has to become the entire customer ecosystem for your whole business, coming right at its former parent. And then everyone else has to follow. When someone goes risk-on, everyone goes risk-on.” @rodriscoll Love to hear your thoughts; does Bret’s clear ambition to eat Salesforce require all providers to move beyond customer service @thejessezhang @destraynor @jaminball @chetanp @stevehind
中文: 当有人冒险时，每个人都会冒险 Salesforce 是占主导地位的 SaaS 公司，市值在 1800 亿至 2000 亿美元之中，而 Service Cloud 可能占其25%，因此纯粹的 Service Cloud 替代服务仅值 50 项。 为了证明150亿美元或200亿美元的市值合理性，Sierra 无法仅仅打造一个略好于下一代服务云的服务。它必须成为整个企业的整个客户生态系统，直接成为其前任母公司。然后其他人都必须关注。 当有人冒险时，每个人都会冒险。 @rodriscol 喜欢听你的想法;布雷特明确希望吃Salesforce，是否要求所有服务机构都超越客户服务 @thejessezhang @destraynor @jaminball @chetanp @stevehind
🎬
视频
#96 🔗 原推
2026-09-11 15:06:00
Harry Stebbings
How Robinhood Could Disrupt the Whole IPO Market “There are a lot more IPOs we need to get done, and it would be neat if Robinhood flipped the script so you really could have a decent IPO driven primarily by retail. There are downsides. You hope institutional investors hold for two years, and more often than not they do, but that playbook sort of works. If you could do a $200 million to $400 million IPO led through Robinhood, mostly by retail investors, that would be disruptive for a subset of startups and great for the ecosystem. Nvidia can’t buy everything.” @jasonlk Love to hear your thoughts on this and how realistic you think this could be @bgurley @howardlindzon @infoarbitrage @vladtenev.
中文: 罗宾汉如何颠覆整个IPO市场 我们需要完成更多的IPO，如果罗宾汉改变剧本，你完全可以以零售为主，进行一场像样的IPO。 有缺点。你希望机构投资者能持有两年，而且往往不持有这种运作，但这种剧本却是一种剧本。 如果你能通过Robinhood(主要由散户投资者主导)进行2亿至4亿美元的IPO，这将对部分初创企业造成影响，并对生态系统造成巨大影响。英伟达买不到所有东西。@jasonlk 很喜欢听听你对此的看法，以及你认为这件事可能有多现实:@bgurley @howardlindzon @infoarbitrage @vladtenev。
🎬
视频
#95 🔗 原推
2026-09-11 01:08:00
Harry Stebbings
The #1 Priority for Mark Zuckerberg Right Now “I would imagine, as we speak, there are 20 engineers locked in a room somewhere in Palo Alto, literally with guards at the door, saying, ‘Nobody eats and nobody leaves until you ship an Instinct clone.’” @rodriscoll What despite all the resources, is Zuck unable to do, that you are able to do @noahrshinn @jgreze @therealnirs @Altimor @frydwia
中文: 目前马克·扎克伯格的首要任务 我想，在我们说话的时候，帕洛阿尔托某处有20名工程师被锁在一间房间里，门口有警卫，说:“没人吃饭，没人离开，直到你寄出一个本能的克隆。” @rodriscol 尽管资源齐心，扎克却无能为力，你仍可以去 @noahshinn @jgreze @therealnirs @Altimor @frydwia
🎬
视频
#94 🔗 原推
2026-09-10 22:02:00
Harry Stebbings
Spotify 👉 https://open.spotify.com/episode/4L7mwnCTIDOMgnW9ViHVAx?si=51e19c4fc7074b84 Youtube 👉 https://youtu.be/4ioSi39yjJQ Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-jensen-huang-declares-agi-has-arrived-gpt-astra/id958230465?i=1000788813207
中文: Spotify 👉 YouTube 👉 苹果播客 👉
#93 🔗 原推
2026-09-10 14:01:16
Harry Stebbings
I get that we’re all busy, but if you choose to not listen to this podcast, you will be materially less intelligent. On the agenda this week: - Jensen Huang Declares AGI Has Arrived - GPT Astra and Fable 5.1 Accelerate the Model Race - Tesla Launches Cybercabs - Index Pulls Out of Town & Anthropic Pulls From Descartes Acquisition My notes below with @jasonlk and @rodriscoll 1. How Robinhood Could Disrupt the Whole IPO Market Allowing retail investors through Robinhood to lead $200M to $400M tech IPOs could create a disruptive new path to liquidity for mid-stage startups. With traditional institutional listings constrained, retail-driven distribution could unlock critical exit opportunities across the venture ecosystem. 2. Why the Majority of Neo Labs Will Not Be Good Investments Early Neo Labs became multi-trillion-dollar leaders, but backing new entrants today carries enormous risk given the capital and distribution moats of incumbents. Unless a new research lab develops a truly orthogonal model architecture, it will struggle to compete with established frontier labs. 3. Why the Oura IPO Will Be a Success Strong consumer brand recognition, combined with 74% revenue growth and 85% subscriber retention, positions Oura for a successful public debut. Unlike typical consumer apps that suffer from severe churn, Oura’s sticky hardware-plus-subscription model provides the predictability public markets value. 4. The #1 Priority for Mark Zuckerberg Big Tech incumbents are moving at unprecedented speed to defend their distribution against consumer AI startups. Meta’s top operational priority should be putting engineers in a room to rapidly clone emerging agentic applications before new entrants can establish lasting consumer habits. 5. When Someone Goes Risk-On, Everyone Goes Risk-On Venture capital operates in momentum cycles where a single high-profile, risk-on deal can push the entire market to follow. In hypergrowth AI categories, investor behavior is often driven more by competitive pressure and deployment speed than by conservative financial modeling. 6. The People Making Money Are Running Fastest and Evolving Quickest In a market where software features can be cloned in weeks, long-term success belongs to founders who continuously expand and adapt their products. Winning startups survive not by defending legacy features, but by executing relentlessly and compounding capabilities faster than their competitors.
中文: 我得说我们都很忙，但如果你选择不听这个播客，你就会变得不那么聪明了。 本周议程: - 詹森黄宣布AGI已抵达 - GPT Astra 和 Fable 5.1 加速模型竞赛 - 特斯拉推出Cybercab - 指数从笛卡尔收购中撤出城镇和人类 我下面用 @jasonlk 和 @rodriscoll 的笔记 1。罗宾汉如何颠覆整个IPO市场 允许通过Robinhood的零售投资者牵头2亿至4亿美元的科技IPO，可能为中端初创企业创造一条颠覆性的流动性新路径。由于传统机构信息受限，以零售为导向的分销可能为整个创业生态系统带来关键的退出机遇。 2。为什么大多数Neono Labs投资不好 早期的Noo Labs成为数万亿美元的领导者，但鉴于现有企业的资本和分销护城河，如今支持新进入者将面临巨大风险。除非一个新的研究实验室开发出真正正交的模型架构，否则将难以与已建立的前沿实验室竞争。 3。为何我们的IPO会取得成功 强大的消费者品牌知名度，加上74%的收入增长和85%的用户留存率，使 Oura 成功实现了公众的亮相。与遭受严重流失的典型消费者应用程序不同，Usera 的粘性硬件加订阅模式提供了市场市场的可预见性价值。 4.马克·扎克伯格的首要优先事项 大型科技公司的现有企业正以前所未有的速度推进，以捍卫其对消费者人工智能初创企业的分销。Meta的首要任务应该是让工程师们迅速克隆新兴的代理应用，然后让新进入者建立持久的消费习惯。 5。当有人冒险时，每个人都会冒险 风险投资在动量周期中运作，一项备受关注的高风险交易可能推动整个市场跟进。在高增长型人工智能类别中，投资者行为往往更多地受竞争压力和部署速度的推动，而不是由保守的金融模型所驱动的。 6.赚钱的人跑得最快，发展得最快 在一个软件功能可在数周内克隆的市场中，长期成功取决于那些持续扩展和调整其产品的创始人。获胜的初创企业生存的并非通过捍卫传统特征，而是通过比竞争对手更快地执行持续且具有复合能力。
#92 🔗 原推
2026-09-10 14:01:16
Harry Stebbings
Next on the block is Snap. 2027 Prediction: Bending Spoons signs agreement to acquire Snap Inc. Watch this space.
中文: 下一块是 Snap。 2027年预测:Bending Spoons签署协议收购Snap公司 观看这个空间。
#91 🔗 原推
2026-09-10 13:36:47
Harry Stebbings
Why Memory and GPUs Have to Be Co-Located “If you want to do large-scale training like we do, you need the memory to be co-located with a large cluster of GPUs. I can’t just rent from Google, Microsoft, or even Base10 and run training at the scale I want because I need a gigantic memory card next to the GPUs, containing all my data and accessible by the entire cluster.” @cliffweitzman Love to hear your thoughts on this @ctnzr @tri_dao @matei_zaharia @NaveenGRao
中文: 为什么内存和GPU必须共同定位 如果你想像我们一样进行大规模训练，需要将内存与大量GPU集群共同定位。 我不能只从谷歌、微软甚至Base10租借，然后按我想要的规模运行训练，因为我需要一张巨大的存储卡，旁边是GPU，包含我所有的数据，并可被整个集群访问。”@cliffweitzman 很喜欢听听你对这个问题的看法:@ctnzr @tri_dao @matei_zaharia @NaveenGRAo
🎬
视频
#90 🔗 原推
2026-09-09 23:36:00
Harry Stebbings
Why Every Company Should Buy, Not Rent, GPUs “If I wanted to rent an H100 for a year at $3.50 to $5 an hour across cloud providers, multiplied by 24 hours a day and 365 days a year, I’d end up paying $35,000 to $50,000. But I could buy that same GPU for $30,000.” @cliffweitzman Love to hear your thoughts on this @GavinSBaker @stephenbalaban @dylan522p
🎬
视频
#89 🔗 原推
2026-09-09 21:32:00
Harry Stebbings
John Terminus for goodness sake. None of us care about a foldable screen. We just want you to buy Wispr Flow and Instinct and having a working Siri. Is this too much to ask… 🤣
中文: 约翰·特里尼乌斯，为天而有名。 我们谁都不在乎可折叠屏。 我们只希望您购买Wispr Flow和Instinct，并拥有一个可工作的Siri。 这太令人难以接受了吗......🤣
#88 🔗 原推
2026-09-09 21:31:12
Harry Stebbings
The best sign to look for when investing in a company: When the building manager of the building they are in hates them? They stay too late. They always need more office space. They have so many meetings etc etc. Great companies, piss of building managers. Always!
中文: 投资一家公司时最需要注意的标志: 当大楼的经理讨厌他们时? 他们待得太晚了。他们总是需要更多的办公空间。他们有那么多会议等。 优秀的公司，建筑经理的小便。永远!
#87 🔗 原推
2026-09-09 21:23:58
Harry Stebbings
The highest value per sentence person on X. 👇
#86 🔗 原推
2026-09-09 20:28:47
Harry Stebbings
What We Care About When Hiring, and Why It’s Easier for Seed-Stage Companies Than Ever “Even six months ago, we really cared whether you had read a ton of textbooks and whether your handcrafted code was amazing. What I care about most today is raw technical intelligence because I know we can teach you everything else. In six months, you could be a machine. So we hire math Olympiad competitors, Kaggle winners, and physicists who might not have even coded before. Almost anyone can become someone you hire now if they’re smart and hardworking because you can teach them incredibly fast.” @cliffweitzman What is your most unconventional tip to founders about hiring today @ceo_clickhouse @matansf @m_franceschetti @alanchanguk
🎬
视频
#85 🔗 原推
2026-09-09 18:45:00
Harry Stebbings
Why None of the AI Assistants Have Product-Market Fit Today “None of the products today actually have really deep product-market fit yet. GrokBot’s really cool, but it’s a power-user product, not a mainstream one. Town is great, but we have a lot of work to do to make it a true mainstream product. So even before I worry about defensibility, I’m still asking: What is the right product experience that’s going to resonate with the mainstream?” @jgreze Love to hear your thoughts on this. @andrewchen @benthompson @gregisenberg @blader
中文: 为什么目前没有一个人工智能助手具有产品市场的契合度 如今的产品实际上还没有真正具有深度的产品与市场契合度。 GrokBot 非常酷，但它是一款功耗为用户的产品，而不是主流产品。小镇很棒，但我们还有很多工作要做，才能使其成为真正的主流产品。 因此，即使在我担心可防御性之前，我还在问:什么才能与主流产品产生共鸣?@jgreze 很喜欢听听你对此的看法。@andrewchen @bentompson @gregisenberg @blader
🎬
视频
#84 🔗 原推
2026-09-09 15:34:00
Harry Stebbings
Well that is reassuring. I have never been one to want to pass posts through compliance first. But feel Dario might be calling Evan into his office for this one 🤣
#83 🔗 原推
2026-09-09 07:43:34
Harry Stebbings
Why Instinct Is Not a Competitor to Town “I don’t think Instinct (@noahrshinn) and Town are trying to do the same thing or monetize the same way. We generate revenue from companies using us for work, with network effects around multiple team members. Their product doesn’t do any of that. I see a strategy that’s more like customer acquisition with a free, fully subsidized product. The harnesses have similar capabilities, but the ICPs and where the marketing is going feel pretty different to me.” @jgreze How do you think about the future of agent to human interaction? How many agents will a single human have… @martinmignot @altcap @scottbelsky @joshm @omooretweets @ankrgyl
🎬
视频
#82 🔗 原推
2026-09-09 00:32:00
Harry Stebbings
@alexandr_wang When does it come out in the UK?
#81 🔗 原推
2026-09-08 22:19:25
Harry Stebbings
To me, Instinct is very similar to Lovable. Lovable started off as a thin wrapper on top of models. It was able to be copied fast. But it wins because of: 1. Speed of execution. 2. Brand. 3. Momentum/fundraising/hires. And over time, the product moves from being easily copied to being immensely feature-rich and deep. Jury is out whether Instinct can do the speed of execution but they have the brand and momentum.
中文: 对我来说，本能与洛夫莱姆非常相似。 可爱的最初是一款轻薄的包装纸，采用了型号。它能够被快速复制。 但它之所以能胜出，是因为: 1。执行速度。 2。品牌。 3。动量/筹款/招聘。 随着时间的推移，产品从易于复制，转向功能丰富且深度极致。 陪审团排除了Instinct能否完成执行速度，但他们拥有品牌和动力。
#80 🔗 原推
2026-09-08 22:14:08
Harry Stebbings
To me, Instinct is very similar to Lovable. Lovable started off as a thin wrapper on top of models. It was able to be copied fast. But it wins because of: 1. Speed of execution. 2. Brand. 3. Momentum/fundraising/hires. And over time you build the product moves from being easily copied to being immensely feature-rich and deep. Jury is out whether Instinct can do the speed of execution but they have the brand and momentum.
中文: 对我来说，本能与洛夫莱姆非常相似。 可爱的最初是一款轻薄的包装纸，采用了型号。它能够被快速复制。 但它之所以能胜出，是因为: 1。执行速度。 2。品牌。 3。动量/筹款/招聘。 随着时间的推移，你将产品从易于复制，发展到功能丰富且深度极致。 陪审团排除了Instinct能否完成执行速度，但他们拥有品牌和动力。
#79 🔗 原推
2026-09-08 22:13:16
Harry Stebbings
What Will Separate the Winners From the Losers in AI Agents “The product that wins this category will have a network effect at the agent level. One feature our power users love most is Agent-to-Agent. You ask your townie a question; it realizes a coworker’s townie has the answer, so it just goes and asks it. Once your whole team is on that, it’s really difficult to imagine moving to a different product. No one has figured out multi-user, multiplayer AI yet. I think that’s the moat.” @jgreze Love to hear your thoughts on this @lessin @hwchase17 @Altimor @medinism
中文: 人工智能代理中赢家与输家的分离 赢得此类别的产品将在代理级别产生网络效应。 我们的用户最喜爱的一个功能是代理代理。你向你的镇上人问了一个问题;它意识到同事的镇上人有答案，所以就去问一下。 一旦你的整个团队都加入进来，就很难想象会换个产品了。目前还没有人发现多人多用户AI。我觉得这就是护城河。@jgreze 喜欢听你对这个@lessin的看法 @hwchase17 @Altimor @medinism
🎬
视频
#78 🔗 原推
2026-09-08 21:18:00
Harry Stebbings
“The question no one knows is how much of the workload for any company stays close to the frontier, where it’s very expensive. For human-level tasks like scheduling, calendars, answering emails, and daily research on competitors, things are trending far from the frontier, and open-weight models already do them really well. Prices halve every nine to twelve months, so you can price your product today to generate 20 to 30% margins in 18 months. But whether you’re left with 10%, 20%, or 30% at the frontier will change the economics of these companies.” @jgreze 12 months time, what percent of enterprise workflows will run open vs closed @lqiao @chetanp @alexatallah @nikesharora @benioff
中文: 没有人知道的是，任何公司的工作量都在靠近前沿领域，而那里非常昂贵。 对于诸如日程安排、日历、回复邮件以及日常竞争对手研究等人类层面的任务，情况远非前沿，而开放量级模型已经很好地完成了这些任务。 价格每九到十二个月减半，因此您今天就可以为产品定价，在18个月内产生20%至30%的利润率。但无论你留在10%、20%还是30%的前沿，都会改变这些公司的经济经济状况。 12个月时，企业工作流程的开放运行率将达到与已关闭的 @lqiao @chetanp @lexatallah @nikesharora @benioff 的对比
🎬
视频
#77 🔗 原推
2026-09-08 18:03:00
Harry Stebbings
Why My R&D Is Just Spent Getting to Product Parity With the Giants “These products are expensive to build. A lot of my R&D is just keeping up with the Joneses. Your agent has to be at least as capable as everyone else’s, because if Codex can do something you can’t, and that thing matters to users, it’s over. It’s not like I have two engineers keeping up with Codex. Codex is 100 people making that thing better, and if we’re not as good, the user says, ‘Why would I pay $50 a month for Town when I can pay $24.99 for OpenAI?’” @jgreze Love to hear your thoughts on this @amasad @thisisgrantlee @nadonomy @jacsebl
中文: 为什么我的研发只是与巨人队在产品上取得了对比 这些产品的制造成本很高。我的很多研发都跟上了琼斯一家。 你的代理必须至少能像其他人一样，因为如果Codex能做一些你无法做到的事情，而这件事对用户来说很重要，那就结束了。 不像我有两个工程师在跟上Codex的要求。Codex 是 100 个人让这个东西变得更好，如果我们不这么好，用户会说:“我为什么每月能为 Town 支付 50 美元，才能支付 24.99 美元购买 OpenAI?”@jgreze 很喜欢听你对这个@amasad的看法@thisisgrantlee @nadonomy @jacsebl
🎬
视频
#76 🔗 原推
2026-09-08 15:05:00
Harry Stebbings
A few weeks ago, a company we invested in shut down. The founder sent a very detailed email to the investor base explaining the situation. I responded with three clear questions: 1. I am very sorry to hear this, are you ok? Losing a company is a very personal loss. I want to make sure that you yourself are okay. 2. Is there anything that I can do in the next steps with the investor base to help? 3. What are your thoughts on your next steps? Is there anything I can do to help facilitate those? This is not the time to discuss lessons or negotiate for cents. This is the moment of peak pain. Be gentle. The way you act in hard times is the way you will be treated in good times. People will always appreciate you for thinking long term and being there for them when it matters.
#75 🔗 原推
2026-09-08 12:09:59
Harry Stebbings
My Biggest Strategic Mistake in the History of Speechify “Passing on ElevenLabs was the biggest strategic mistake I made in the history of Speechify. It was 100% on me. I met Piotrek and Mati in 2022, and we were very impressed. But I thought a text-to-speech API would become commoditised over time, so I didn’t want to go into that business. What I didn’t understand was that the point of an AI lab is to continuously innovate. The first product you release is the wedge that gets people to later adopt your other technology.” @cliffweitzman How do you think about the decision about where AI labs should productise vs where they should not @AnjneyMidha @levie @JaroslavBeck @andi_blatt
🎬
视频
#74 🔗 原推
2026-09-08 00:03:00
Harry Stebbings
The hottest category in tech right now is AI assistants. The question is; who is going to win? Instinct raised at $2.5BN. Has Benchmark and Index behind them. Grok Bot is ripping with Elon and has the distribution machine of X. And then there is Town, one of the only ones that’s actually making real money from real businesses. Don’t write Zuck off. This will be his next play and integrated into every WhatsApp user’s product. I sat down with Town Founder, @jgreze to understand WTF is going on, who wins and who loses? Condensed my notes below! 1. We Have Passed the Point Where Humans Look at Lines of Code Software engineering has crossed a threshold where machines increasingly write and ship code into production under model-based guardrails. Outside critical security controls, humans will spend far less time reviewing raw code. The future belongs to systems that manage autonomous AI agents writing software, running tests, and validating their own work. 2. You Can Build at the Speed of Machines, but You Can Only Learn at the Speed of Humans AI development tools allow competitors to clone features in weeks, erasing traditional software head starts. But deep user feedback cannot be automated. While machines accelerate execution, true competitive advantage comes from maximizing human learning cycles and understanding customers faster than anyone else. 3. Why None of the AI Assistants Have True Product-Market Fit Today Despite immense market hype, no current AI assistant has achieved deep product-market fit with mainstream users. Most products still cater primarily to power users rather than everyday workers. Before worrying about moats, founders need to create frictionless experiences that resonate with the mass market. 4. Network Effects at the Agent Level Will Separate AI Winners Sustainable moats in AI assistants will come from multi-user network effects at the agent level, not single-player productivity. When autonomous assistants collaborate across teams to resolve queries and execute work, switching becomes increasingly difficult. Multiplayer workflows create organizational lock-in that personal assistants cannot replicate. 5. How Much of the Workload Stays at the Frontier Versus Open Weight? AI application margins depend heavily on how much work requires expensive frontier models versus cheaper open weights. Complex reasoning may still demand frontier intelligence, while routine tasks like scheduling and email tagging continue moving down the cost curve. Shifting 80% of workloads to open weights over time could create far more sustainable economics. 6. My R&D Is Just Spent Getting Product Parity With the Giants Competing with giants like OpenAI requires massive investment simply to maintain feature parity. Startups cannot rely on unique distribution if their underlying harness falls behind on core capabilities. AI assistants must invest heavily to match the execution speed and product depth of frontier teams. 7. Why Instinct Is Not a Competitor to Town Consumer assistants like Instinct focus on rapid acquisition through subsidized personal tools, while enterprise platforms build monetizable team workflows. Although their technical harnesses may overlap, their target ICPs and business models diverge sharply. Enterprise agents monetize by embedding collaboration directly into daily operations. (links below)
#73 🔗 原推
2026-09-07 15:49:15
Harry Stebbings
Spotify 👉 https://open.spotify.com/episode/3bF9C6632HFkbuhX6aTWG9?si=2b07c2778c784e52 Youtube 👉 https://youtu.be/9ISWVzQ85Po Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-the-%24100-billion-ai-assistant-race-town-vs/id958230465?i=1000788254110
中文: Spotify 👉 YouTube 👉 苹果播客 👉
#72 🔗 原推
2026-09-07 15:49:15
Harry Stebbings
Open feedback: @noahrshinn ability to do voice calls would be a game changer. Connect to ElevenLabs and have that functionality. Would pay $1,000 per month for this.
中文: 开放反馈: @noahrshin 能够进行语音通话将彻底改变游戏规则。 连接到 ElevenLabs 并具备此功能。 每月会为此支付1000美元。
#71 🔗 原推
2026-09-07 11:18:11
Harry Stebbings
24 months ago I stopped using Google in favour of ChatGPT. This weekend I stopped using ChatGPT in favour of Instinct. With every transition, my search volume has exploded and the tasks asked has expanded immensely. This is about to get wild.
中文: 24个月前，我停止使用谷歌来支持ChatGPT。 本周末，我不再使用ChatGPT来支持Instinct。 每一次过渡，我的搜索量都激增，而所要求的任务也大大扩展了。 这即将变得疯狂起来。
#70 🔗 原推
2026-09-06 23:30:17
Harry Stebbings
Why Human-Computer Interaction Is Shifting From Screens to Voice “Google and ChatGPT both won because of a very simple interface: a text box and a button. The even simpler version is just having a conversation. I say something, and I hear something in response. Soon, people are going to be talking to their computers, phones, and wearables constantly throughout the day while using screens a lot less.” @cliffweitzman What does no one know about our future interaction with devices that everyone should know @tankots @_allanguo @scottbelsky
中文: 人为的互动为何正从屏幕转向语音 谷歌和ChatGPT都因界面非常简单而获奖:一个文本框和一个按钮。 更简单的版本就是进行对话。我说了些什么，也听到了回应。 很快，人们就会整天不停地与电脑、手机和可穿戴设备聊天，同时使用屏幕却少了很多。”@cliffweitzman 关于我们未来与设备交互的哪些工作，每个人都应该知道 @_allanguo @scottbelsky
🎬
视频
#69 🔗 原推
2026-09-06 23:19:00
Harry Stebbings
Sometimes I do edit shows and think my word this guest is full of s*** 😂😂😂 Then we can the show. 34% of recorded shows do not get published. Quality of the product is everything.
中文: 有时候我会编辑节目，觉得这个嘉宾的字里满是***😂😂😂 然后我们才能演出。 34%的录制节目未能发布。 产品质量就是一切。
#68 🔗 原推
2026-09-04 22:03:15
Harry Stebbings
We did not end up doing more with less. We did more with more and that's why European startups fail. “It is not true that we are going to do more with less. That turned out to be the great fallacy of late 2025, early 2026. We are doing much more with more. That is the meta point. That is why your European startups, most of them, are going to fail, at least in the US. Their little point solutions are just going to disappear in six months.” @jasonlk Love to hear your thoughts @antonosika @christianreber @chrija @destraynor @Jameswise @taavet @surangac
中文: 我们最终没有用更少的多做。我们做得更多，这就是欧洲初创企业失败的原因。 我们不会用更少的人做更多的事情。事实证明，这是2025年末、2026年初的重大谬误。 我们正在做更多的事情。这就是元点。 这就是为什么你的欧洲初创企业(其中大多数)将会失败，至少在美国是这样。他们的小点解决方案将在六个月后消失。@jasonlk 很喜欢听你的想法 @antonosika @chrianreber @chrija @destraynor @Jameswise @taavet @surangac
🎬
视频
#67 🔗 原推
2026-09-04 16:41:00
Harry Stebbings
This might surprise you but the majority of European VC is quite hollow and parrot things they hear on good podcasts 😉 Paul is the exception. Insanely smart and this is a must read.
中文: 这可能会让你感到意外，但大多数欧洲风险投资公司都相当空洞，而且会在优秀的播客节目中听到这些内容😉 保罗是个例外。理智巧妙，这是必读之法。
#66 🔗 原推
2026-09-04 16:32:46
Harry Stebbings
Hugging Face acquisition explained “A man making $120BN a year selling compute decides to buy a company that helps make compute more cost-effective so he can sell more compute. If end users have $1TN to spend on tokens, Nvidia would prefer that money flow through open-source people at 30% gross margins, rather than 70% gross margins at OpenAI or Anthropic. Open source is good for compute salespeople. If you are selling GPUs, you want everyone else’s margin to be lower so yours can be higher.” @rodriscoll Love to hear your thoughts @vipulved @NaveenGRao @bindureddy @realGeorgeHotz
中文: 收购Hugging Face 一名年收入1200亿美元、销售电脑的人决定收购一家有助于使计算更具成本效益的公司，以便他能卖出更多的计算产品。 如果最终用户有1美元可以投入代币，英伟达更倾向于通过开源人员以30%的毛利率流动，而不是在OpenAI或Anthropic上投入70%的毛利率。 开源对计算销售人员有好处。如果你在销售GPU，你希望其他人的利润率更低，这样你的GPU就能更高。 喜欢听你的想法 @vipulved @NaveenGRAo @bindureddy @realGeorgeHotz
🎬
视频
#65 🔗 原推
2026-09-04 00:54:00
Harry Stebbings
How Instinct could follow the same path as Replit and Lovable “When these products came out, they were all built in a month. It was so easy to clone these products in the early days and do nothing. Now they are so complicated. Replit and Lovable of a year ago were not a moat. Today they have massive moats. If Instinct is going to do what we claim it does, in a year it has got to do 100 times more than it does today. All the use cases it has to accomplish become a moat.” @jasonlk Love to hear your thoughts @sarahtavel @nabeelqu @danshipper @gregisenberg
中文: 本能如何能像《Replit》和《Lovable》一样 这些产品推出时，都是一个月内建成的。早期克隆这些产品非常容易，而且无所事事。 现在它们太复杂了。一年前的“复制与可言不透”。如今他们有巨大的护城河。 如果本能会做我们声称会做的事，那么一年内它的完成次数就比今天多100倍。它必须实现的所有使用案例都变成了一条护城河。@jasonlk 喜欢听你的想法 @sarahtavel @nabeelqu @danshipper @gregisenberg
🎬
视频
#64 🔗 原推
2026-09-03 22:39:00
Harry Stebbings
The three ways the wheels come off the bus for Nvidia “There are only three things that can go wrong. Either their direct customers stop buying compute, the financing breaks, or end-user demand weakens. Really, the only thing that can go wrong at some point, and it is not today, is end-user demand. The whole thing works provided the end customers keep exploding, and right now they are.” @rodriscoll Love to hear your thoughts @GavinSBaker @BenBajarin @TheStalwart
中文: 英伟达的车轮从公交车上起飞的三种方式 只有三件事可能会出错。要么他们的直接客户停止购买计算，要么融资中断，要么最终用户需求减弱。 真的，唯一可能在某个时候出错的，而现在却不是，就是终端用户需求。 只要终端客户不断爆炸，整个事情就奏效了，而现在他们却在。@rodriscoll 很喜欢听你的想法 @GavinSbaker @BenBajarin @TheStalwart
🎬
视频
#63 🔗 原推
2026-09-03 19:53:00
Harry Stebbings
The bull case for Clay being a $100 billion company “The bull case is that agentic GTM has just started. We thought the TAM was the same as it was. It turns out when agents can run these GTM motions, they will consume 10 to 100 times more usage than humans ever could. They can run GTM around the clock. The usage is just going to explode, and Clay is a clear breakout winner. That is the bull case for Clay being a $100BN company.” @jasonlk Love to hear your thoughts. What is your bull case @NicolaeRusan @edsim @acharoo @joshk @andrew__reed @scottbelsky @lennysan @shishirmehrotra
中文: 克莱公司是一家价值1000亿美元的公司 事实就是，代理性GTM才刚刚开始。我们以为TAM和它是一样的。 事实证明，当特工能够运行这些GTM运动时，其使用量将比人类所能消耗的多出10到100倍。它们可以全天候运行GTM。 使用率将大打成一罄，而克莱显然是一个突破性的赢家。这就是克莱成为一家价值100亿美元的公司的大理由。”@jasonlk 喜欢听你的想法。你的靶心是什么 @NicolaeRusan @edsim @acharoo @joshk @andrew__reed @scottbelsky @lennysan @shishirmehrotra
🎬
视频
#62 🔗 原推
2026-09-03 17:05:00
Harry Stebbings
Spotify 👉 https://open.spotify.com/episode/7yClsnQWvXnK6YENx4fR6S Youtube 👉 https://youtu.be/8skbt979n2o?si=HAyz7NfaS7FEJsuU Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-nvidia-crushes-quarter-and-buys-hugging-face/id958230465?i=1000787591862
#61 🔗 原推
2026-09-03 14:08:14
Harry Stebbings
This podcast is the single most important podcast to know what is going on in tech every week. On the agenda this week: - NVIDIA Crushes Quarter and Buys Hugging Face - OpenAI Cuts Off Cursor - Instinct Hits $2.5BN Valuation and The Race for AI Assistants - Cognition Raises at $46BN, Linear $2.5BN and Clay $7BN My notes with @rodriscoll and @jasonlk below: 1. How Instinct Could Follow the Same Path as Replit and Lovable In the early days, cloning lightweight AI tools is trivial. Defensibility emerges by rapidly adding complex workflows like security automation and multi-agent orchestration. Products that start without a moat can build formidable ones over time by solving dozens of evolving customer requirements faster than anyone else. 2. Hugging Face Acquisition Explained As the maker of compute, NVIDIA benefits when AI token traffic flows through 30% gross-margin open-source models rather than 70% gross-margin closed models where platforms capture more of the economics. Driving down software margins allows a greater share of total ecosystem spend to flow directly into GPUs. 3. The Bull Case for Clay Being a $100 Billion Company Autonomous agents executing go-to-market strategies around the clock could consume 10x to 100x more tokens and software usage than human sales teams ever could. As a leader in agentic GTM, Clay is positioned to capture an enormous wave of automated outreach, campaign analysis, and global prospect engagement. 4. The Bull Case for Linear When software teams build 100x more features at 50x the speed using AI, legacy project management tools and manual Kanban boards begin to break down. Linear can become the agent-friendly system of record for coordinating, tracking, and managing thousands of issues generated simultaneously by human-agent development teams. 5. The Three Ways the Wheels Come Off the Bus for NVIDIA NVIDIA’s record-breaking momentum faces one fundamental existential threat: a sudden collapse in end-user demand for AI intelligence. Hyperscaler CapEx buildouts and complex vendor financing arrangements work only as long as customers continue aggressively buying frontier-model tokens throughout the supply chain. 6. We Are All Building Compound Startups Today AI development tools have accelerated code production dramatically, making narrow point solutions increasingly vulnerable. To survive rapid competitive convergence, software startups must embrace becoming compound companies that ship expansive, multi-module product suites covering the entire customer workflow. 7. We Did Not End Up Doing More With Less. We Did More With More, and That’s Why European Startups Fail The belief that AI would allow companies to shrink headcount and simply do more with less has not played out as expected. Winners are compounding capital and talent to do vastly more with more, putting underfunded point solutions, particularly across Europe, at risk of being overwhelmed by aggressively scaling U.S. competitors. (links in comments)
#60 🔗 原推
2026-09-03 14:08:03
Harry Stebbings
The Bull Case for Linear Being a $100BN Company: "Linear is the clear winner. They have built an agentic product first that allows us to build 100x more software, and that means 100x more features than ever before. Humans cannot keep up with it, and humans still have to work with agents. If every human on your team's gonna build 500 features and 1,000 issues and you have 10 people on your team, you need a process and a system of record for managing all these issues with your agents. We need a new system of record for it. The team at Linear has figured it out. We have seen an explosion, 50x more agent usage than 90 days ago. This will seem cheap when Replit's at $15BN, Lovable's at $100BN, because Linear will be the one powering them all". @jasonlk Love to hear your thoughts, what is your bull case for Linear @karrisaarinen @stephzhan @adambain @dickc @soleio @zoink @Mkclements
中文: 线性作为一家价值100亿美元公司的牛市案例: 线性是明显的赢家。 他们先构建了一个代理产品，使我们能够多构建100倍的软件，这意味着其功能比以往多100倍。 人类无法跟上它，人类仍然需要与特工合作。如果你团队中的每个人都要构建500个功能和1000个问题，而团队中有10个人，那么你需要一个流程和记录系统，才能与你的代理人员一起管理所有这些问题。 我们需要一个全新的记录系统。 线性团队已经弄清楚了。我们看到爆炸了，比90天前的代理使用量多了50倍。 当“以150亿美元”(售价为1500亿美元)的“Replit”(售价为100亿美元)时，这看起来似乎很便宜，因为Linear将是它们全部的动力之一。@jasonlk 喜欢听你的想法，你为线性@karrisaarinen @stephzhan @adambain @dickc @soleio @zoink @Mkclements 提供什么
🎬
视频
#59 🔗 原推
2026-09-02 22:39:43
Harry Stebbings
What is required to get the most out of models? “To take the most advantage out of models, you need to do something a little different from just routing. You need your agent or system to dynamically understand the task it is working on and understand how to allocate intelligence in a much more stateful way. You need to know what just happened and what is going to happen in the future. You have to be in there, in the task.” @EnoReyes How do you think about this @jerryjliu0 @tomas_hk @sarahwooders @AstasiaMyers
中文: 最需要什么才能充分利用模型? 要充分利用模型，你需要做一些与路由不同的事情。 你需要你的代理或系统动态地理解它正在处理的任务，并理解如何以更有状态的方式来分配情报。 你需要知道刚刚发生了什么以及未来会发生什么。你必须在任务中待在那里。@埃诺雷耶斯 你怎么看这个 @jerryjliu0 @tomas_hk @sarahwooders @AstasiaMyers
🎬
视频
#58 🔗 原推
2026-09-02 15:58:00
Harry Stebbings
Are large U.S. enterprises scared to work with frontier model providers? “People do not trust the statement ‘zero data retention.’ It is basically you saying, ‘Just trust me, I have got you.’ A lot of companies worry about sending their source code, for example, to a frontier lab. You worry about the output from that code generation. You worry about third-party indemnification if you were to consume code that is being derived from another repository.” @ceo_clickhouse Love to hear your thoughts @lorenc_dan @mitchellh @tqbf @NaveenGRao
中文: 美国大型企业是否害怕与前沿模式提供商合作? 人们不信任“数据保留”这一说法。基本上就是你说:“相信我，我已经拥有你了。” 许多公司担心将源代码发送到前沿实验室。 你担心那代代码的输出。如果你要使用从其他仓库衍生出的代码，就会担心第三方赔偿。@ceo_clickhouse 喜欢听你的想法 @lorenc_dan @mitchellh @tqbf @NaveenGRAo
🎬
视频
#57 🔗 原推
2026-09-02 00:04:00
Harry Stebbings
The two ways that large language models will win “If you are a model provider, you are basically looking to dominate the platform era and get really good at selling inference. Or you want to move up to become an application-layer company that has really good models. Anthropic seems to be following the application path, while OpenAI seems to be dipping its toes in both.” @EnoReyes How do you think about this @charlespacker @swyx @bennstancil @jaminball
中文: 大型语言模型将赢得的两种方式 如果你是一家模型提供商，你基本上希望主导平台时代，并真正擅长销售推理。 或者你想上台，成为一家拥有非常优秀模式的应用层公司。 Anthropic 似乎正沿着应用路径走下去，而 OpenAI 似乎正在两者兼而有色。@埃诺雷耶斯 你怎么看这个@charlespacker @swyx @bennstancil @jaminball
🎬
视频
#56 🔗 原推
2026-09-01 21:05:00
Harry Stebbings
The age of the opinion-less media brand is over. You need to stand for something. You need to have bold opinions. Just too crowded otherwise.
中文: 无意见媒体品牌的时代已经结束。 你需要为某事挺身而上。你需要有大胆的意见。 否则就太拥挤了。
#55 🔗 原推
2026-09-01 20:12:38
Harry Stebbings
Why is everyone underestimating the outcome sizes today? “People are looking at outcome sizes of AI and saying, ‘That is ludicrous. That is crazy.’ That is underestimating by an order of magnitude how massive a transformation this is going to be. The types of businesses that are going to become massive do not look like businesses 20, 30, 40 years ago. It is basically collections of people that understand what the future looks like a little bit more clear-eyed than other people.” @EnoReyes Do you think we are fundamentally massively underestimating outcome sizes of the next generation of companies or will it truly be for a very select small few that are outliers (Anth, OAI) @anjneymidha @mmurph @lessin @Curiousjorge65 @danhockenmaier @danshipper
中文: 为什么今天每个人都低估了结果的大小? 人们正在关注人工智能的结果尺寸，并说:“这太荒谬了。”这太疯狂了。这低估了这一转变将是多么巨大。 那些将会变得规模庞大的企业类型，看起来与20年、30年、40年前的企业并不像。 基本上，人们懂得未来看起来比其他人更清醒一些。@埃诺雷耶斯 你认为我们从根本上低估了下一代公司的成果规模，还是真正针对少数一些例外的公司(Anth，OAI)@anjneymidha @mmurph @lessin @Curiousjorge65 @danhockenmaier @danshipper
🎬
视频
#54 🔗 原推
2026-09-01 18:47:00
Harry Stebbings
What changes when you're building software for agents not for humans? “The experience is going to be defined by the slowest point in that chain. The number one requirement for agentic query patterns is low latency, because they are executing dozens of SQL queries simultaneously across all these different systems. The most important requirement is the unpredictability of those query patterns, the responsiveness, and the fact that they are much more exploratory than a traditional human report or query.” @ceo_clickhouse Love to hear your thoughts @glcst @kiwicopple @bernhardsson @ashashutosh
🎬
视频
#53 🔗 原推
2026-09-01 15:09:00
Harry Stebbings
Just to say for all those who critiqued Rishi Sunak, the man is still an active MP, working hard for his constituents... just saying.
中文: 就所有批评里希·苏纳克的人而言，这名男子仍然是一位活跃的议员，为选民努力工作......只是说。
#52 🔗 原推
2026-09-01 12:01:29
Harry Stebbings
What should investors be worried about today that they are not? “The single biggest risk would be durability of revenue, because the switching costs are very high for infrastructure software. The switching costs can be very low for agentic applications, and we are seeing that with these model providers leapfrogging one another every other week. I would call into question the durability of some of the revenue for some of these AI applications.” @ceo_clickhouse Love to hear your thoughts @mmurph @ttunguz @jaminball @Altimor
中文: 投资者今天应该担心什么? 最大的风险是收入的持久性，因为基础设施软件的更换成本非常高。 对于代理应用程序而言，更换成本可能非常低，而且我们看到这些型号供应商每隔一周就会相互跨越一次。 我质疑部分人工智能应用收入的持久性。@ceo_clickhouse 喜欢听你的想法 @mmurph @ttunguz @jaminball @Altimor
🎬
视频
#51 🔗 原推
2026-09-01 01:57:00
Harry Stebbings
How does this AI cycle compare to prior technology shifts and transitions? “Those cycles, in my experience, were much more gradual. This seems to be accelerating at an unprecedented pace in terms of how quickly these agentic experiences are maturing and how quickly these companies are growing. We have not seen revenue growth like this in our lifetime, and the demands on the systems of these agentic applications are unlike anything we have ever seen.” @ceo_clickhouse Love to hear your thoughts @rick @jasonlk @patio11 @infoarbitrage
中文: 与以往的技术变化和转型相比，这种人工智能循环是如何进行的? 根据我的经验，这些周期要循序渐进得多。 从这些代理体验的成熟速度以及这些公司发展的速度来看，这似乎正以前所未有的速度加速。 我们一生中从未见过这样的收入增长，这些代理应用程序系统的需求与我们以往见过的任何需求都大不相同。 喜欢听你的想法 @rick @jasonlk @patio11 @infoarbitrage
🎬
视频
#50 🔗 原推
2026-08-31 23:40:00
Harry Stebbings
@ceo_clickhouse Spotify 👉 https://open.spotify.com/episode/3HN3a51A8b4cbwJmgOBRA8 Youtube 👉 https://www.youtube.com/watch?v=Wna2Ds4dt50 Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-the-ai-bubble-is-wrong-ai-margins-need-to/id958230465?i=1000786926609
#49 🔗 原推
2026-08-31 16:37:36
Harry Stebbings
The story of ClickHouse is truly insane. Started as an open-source project; scaled into the fastest-growing database product ever. Year 1: $0 Year 2: $12M Year 3: $50M Year 4: $200M Year 5 (not complete): My bet is $450M. My notes from our discussion with @ceo_clickhouse below 👇 1. Are Large U.S. Enterprises Scared to Work With Frontier Model Providers? Large enterprises remain skeptical of “zero data retention” claims and wary of sending proprietary source code to frontier labs due to IP indemnification and data leakage concerns. Rather than exposing production code, companies may limit frontier model usage to less sensitive workflows like code review while turning to open-weight alternatives for critical data. 2. How Do You Assess Defensibility and Moat in Companies That Scale Faster Than Ever Before? When an application scales from zero to $100M in ARR in a single year, investors must rigorously question its underlying moat. Hypergrowth without high switching costs leaves companies vulnerable to rapid churn as customers move effortlessly to the next model or tool that leapfrogs the incumbent. 3. How Does This AI Cycle Compare to Prior Technology Shifts and Transitions? Unlike the gradual adoption curves of the internet and mobile eras, the current AI wave is accelerating at an unprecedented pace. Agentic experiences are maturing rapidly, driving explosive revenue growth and placing historically unique performance demands on underlying data infrastructure. 4. What Job Does Not Exist Today That Will Be Very Prevalent in Five Years? A critical new corporate role could be an AI finance function dedicated entirely to managing token consumption and resource allocation across the enterprise. But the role may ultimately be short-lived as autonomous AI agents increasingly manage their own infrastructure spend and budget execution. 5. What Should Investors Be Worried About Today That They Are Not? The biggest overlooked risk in AI today is revenue durability. While infrastructure software benefits from high switching costs, agentic applications can have exceptionally low barriers to switching, raising questions about long-term retention as models and products continually leapfrog one another. 6. Why Revenue Concentration Is a Real Concern Operators and investors should treat revenue concentration as a critical risk, with any single customer or vertical accounting for more than 10% of revenue representing significant exposure. Sustainable enterprise value requires a diversified customer base so losing one account never threatens the company’s overall growth trajectory. (links in comments)
media 0 共 1 项
#48 🔗 原推
2026-08-31 16:37:17
Harry Stebbings
When I was eight years old, I was a ball boy at Fulham Football Club. I was an avid fan, living down the road, and this was the pinnacle of all pinnacles. So to have the chance to walk out onto the pitch, to place the ball down before the first game of the season, against Chelsea, was truly a dream come true. @ceo_clickhouse I cannot thank you enough for this my friend. You do know @ClickHouseDB could IPO, return me a s*** ton of money and it still would not likely beat this feeling! Stay tuned for a banger of an episode coming with Aaron tomorrow.
media 0 共 1 项
#47 🔗 原推
2026-08-30 14:34:53
Harry Stebbings
Six months ago I asked a lawyer friend: How much do you use Legora today? They responded: maybe for 10% of tasks. It is helpful. I asked them again last week, the same question: They responded: I just check the output Legora does. If it were taken away, I would be SO SO upset. It has gone from doing 10% to doing 80%. Law will follow the same path coding has done.
中文: 六个月前，我问了一位律师朋友: 你今天使用Legora多少钱? 他们回应了:或许有10%的任务。很有帮助。 上周我再次问了他们，同一个问题: 他们回应道:我只需查看Legora的输出结果即可。如果被带走，我会非常生气。 从做10%变成了80%。 法律将遵循编程相同的路径。
#46 🔗 原推
2026-08-29 23:20:30
Harry Stebbings
Contrarian view: I think London property will absolutely rip over the next 24 months. Why? Huge amount of wealth created by IPOs and acquisitions. SpaceX, Anthropic, OpenAi, Cursor etc etc. Americans love two things. 1. Euro Summer. 2. London! 50% of NVIDIA are worth more than $25M. A load of Americans are about to buy a pied a terre in London. Money on!
中文: 反政府观点:我认为伦敦房产在未来24个月内绝对会被收购。 为什么? 通过首次公开募股和并购创造了巨额财富。SpaceX、人类、OpenAi、Cursor等 美国人喜欢两件事。 1。欧洲夏季 2。伦敦! 50%的英伟达价值超过2500万美元。 大量美国人即将在伦敦购买一个铁饼。 钱!
#45 🔗 原推
2026-08-29 18:46:35
Harry Stebbings
@matanSF @MattEvantic But you do look like Matt Damon in Goodwill Hunting so… 🤷‍♂️
中文: @matanSF @MattEvantic 但你在《善意狩猎》中看起来确实像马特·达蒙一样......🤷 ♂️
#44 🔗 原推
2026-08-29 16:04:15
Harry Stebbings
@matanSF @MattEvantic He was much more charming than you my friend 🤣
中文: @matanSF @MattEvantic 他比你的朋友 🤣 更迷人
#43 🔗 原推
2026-08-29 15:56:58
Harry Stebbings
Spotify 👉https://open.spotify.com/episode/1K7yijPmWta84oqajDwRAg?si=00e5af72b0c74b4e Youtube 👉https://youtu.be/h9VNB9TA2Hk
中文: Spotify 👉 YouTube 👉
#42 🔗 原推
2026-08-29 15:28:07
Harry Stebbings
I first met @matanSF following a kind intro from @MattEvantic.   We went for a walk in Hyde Park. I was in my trusty short shorts and it was kinda awkward as a walk.  It was awkward because after 5 mins, he was clearly a genius and for the next 55 mins, I just had to pretend like I had more questions to ask before saying with intense eagerness, “can I invest”.  Thank the lord he let me. And through that I got to spend time with his co-founder, @EnoReyes.  Eno is this insane combination of a truly brilliant technologist with an intense awareness of what it takes to build an insanely high-margin, efficient business in AI.  I sat down with Eno when he was in London recently (episode in comments) and have added my handwritten notes below.  Special thanks to @rabois @shaunmmaguire @byersblake @Sabina_Smith_ @laurenmhreeder for some amazing question suggestions. -- 1. The Frontier in AI Right Now The true frontier of AI is building verification systems where no benchmarks exist. Creating systematic frameworks for what “good” looks like allows AI to reliably execute complex, high-friction human tasks that were previously difficult to automate. Is this the true beauty of Instinct @noahrshinn @saranormous 2. The Two Ways That Large Language Models Will Win Frontier model providers face two paths: dominate infrastructure through high-volume inference or move up the stack into high-margin applications. But model-locked applications can conflict with what enterprises actually want: the best possible outcome across multiple models. Love to hear your thoughts on this specifically @AnjneyMidha @mmurph 3. What Is Required to Get the Most Out of Models? Simple gateway routing outside the execution layer delivers only basic cost savings. Maximizing agentic performance requires operating statefully inside the workflow itself, dynamically understanding context, execution history, and what needs to happen next. How do you think about this @alexatallah @shensi @ThibaultJaigu @rauchg @koblovinamerica 4. Why 80% of Neo Labs Will Die and What Separates the Winners From the Losers The winners will anchor themselves to durable enterprise workflows that do not disappear as underlying frontier models improve. Single biggest advice to VCs on investing in neolabs today @LiamFedus 5. Why the Harness Is So Valuable and Who Ultimately Is the Sovereign of Your Intelligence Continuous learning and workflow optimization happen at the harness layer, not inside closed model APIs. True enterprise sovereignty requires owning your harnesses and learning loops so critical intelligence remains proprietary rather than being ceded to third-party labs. 6. Why We Should Not Be Scared to Use Chinese Open-Source Models Labeling open-source weights as dangerous “Chinese models” can obscure the distinction between model provenance and actual security risk. Open models can reflect creator biases, but their risks should be evaluated technically rather than by origin alone. As open weights improve, they could power a growing share of standard enterprise workflows.
media 0 共 1 项
#41 🔗 原推
2026-08-29 15:27:10
Harry Stebbings
Dear @noahrshinn, I want to try Instinct. I have deep fomo. 🤣 Heard the most wonderful things. Please see below. 😢 https://twitter.com/HarryStebbings/status/2093713423808836067/photo/1
中文: 亲爱的@noahrshinn， 我想试试本能。 我有深厚的表情。🤣 听到最美妙的事情。 请看下方。😢
media 0 共 1 项
#40 🔗 原推
2026-08-29 14:52:34
Harry Stebbings
You need $500,000 a year as a salary just to live in San Francisco today! “Rents in the mediocre apartments in Dogpatch are $10,000 a month now. If it’s $10,000 a month to rent a one-bedroom at The Avalon, how much do you have to make to feel rich? A lot. You need $240K in California pre-tax just to pay the rent. You probably need $480K to feel good about yourself.” @jasonlk Love to hear your thoughts @kimmaicutler @garrytan @lessin @Noahpinion
🎬
视频
#39 🔗 原推
2026-08-27 23:15:00
Harry Stebbings
Why Hugging Face and TBPN have more in common than you think: “If someone does buy Hugging Face, the deal’s gotta be you don’t touch it. Because if you touch it, you break it. It’s a much bigger version of the TBPN challenge. If it becomes an OpenAI commercial, TBPN has no value. If you mess with this marketplace for 10,000 models, even if you put a little ad at the top, you destroy it. If anyone actually spends $3 billion, let alone $13 billion, they’ve got to leave it alone for 24 to 36 months.” @jasonlk Single biggest advice to HF on how to stay “neutral” in this new world of being acquired @paraschopra @dabit3 @bindureddy @mattshumer_
🎬
视频
#38 🔗 原推
2026-08-27 20:14:00
Harry Stebbings
“Everyone in IT has woken up and realized that these two frontier models could steal a lot of their TAM. Everyone is saying, ‘We better have a different story.’ The enterprises are saying it, Palantir is saying it. If you are an enabling technology for open-weight models, now is peak moment.” @rodriscoll Love to hear your thoughts @NaveenGRao @natolambert @simonw @AnjneyMidha @swyx
中文: IT中的每个人都意识到，这两个前沿模型可能会窃取他们的大量TAM。 人人都在说:“我们最好有不同的故事。”企业在说，帕兰提尔在说。 如果你是一款适用于敞口重量模型的赋能技术，那么现在已成为巅峰时刻。@rodriscol 很喜欢听你的想法 @NaveenGRao @natolambert @simonw @AnjneyMidha @swyx
🎬
视频
#37 🔗 原推
2026-08-27 17:36:00
Harry Stebbings
@rodriscoll @jasonlk Spotify 👉 https://open.spotify.com/episode/1AMW9bPLbJNtDXTBCQQrih Youtube 👉 https://youtu.be/j_JlQm8DB80 Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-nvidia-bonanza-buys-poolside-invests-in-mercor/id958230465?i=1000786216809
#36 🔗 原推
2026-08-27 14:05:49
Harry Stebbings
If you do not listen to this weekly show, you are deliberately choosing to not make yourself smarter. The world of tech has never moved faster. You need to stay up to date with the most thoughtful analysis. That is where @rodriscoll and @jasonlk come in! 😉 On the agenda this week: - NVIDIA Bonanza: Buys Poolside & Invests in Mercor and Perplexity - Anthropic's $30TRN Revenue Assumption & OpenAI Confirms IPO - Why Customer Service, Defence and Robotics are Overinflated My notes below 1. We Are So Much More Addicted to Tokens Than We Think As employees adopt 10 to 20 sub-agents running around the clock, businesses are facing unexpected $20,000 per-employee token bills. This is not a temporary trend. Token consumption is becoming an irreversible dependency, with top talent increasingly viewing continuous access to AI compute as essential to doing their jobs. 2. Why Customer Support Is Dead as a Category Standalone customer support software is collapsing as agentic interfaces merge siloed tools into unified workflows. Legacy CS and CX platforms risk becoming cheap commodities as support capabilities are absorbed into cross-functional sales, marketing, and operational agents. 3. Why AI Services Companies Are a Bullshit Category Funding traditional law or accounting firms rebranded as AI services companies relies on convoluted structures that look great on spreadsheets but break down in execution. Premium tech multiples cannot be created simply by wrapping overworked elite graduates in AI-powered agency business models. 4. $9 Billion Doesn’t Clear the Bar for Seed Investing in 2026 High seed valuations combined with massive dilution mean even multi-billion-dollar exits may no longer generate fund-returning outcomes. If effective entry pricing reaches $600 million after dilution, a $9 billion exit produces only a 15x return, far below the 50x outcomes that drive venture power laws. 5. If Danger Can Be Described as the Absence of Choice, They Were Now in Danger As Anthropic gains enterprise share and improves profitability, OpenAI risks losing control over its public-market timeline. Massive capital requirements and intensifying competition across every model tier could leave it with fewer strategic options, forcing it to accept whatever valuation public markets are willing to offer. 6. It’s All About Code. That’s the Only Sentence That Matters Coding is the highest-ROI, fastest-adapting, and most critical workload in AI. While consumer chat products struggle with low willingness to pay and heavy compute subsidies, developer workflows command massive enterprise budgets and offer a clearer path to durable software value creation. 7. Silicon Valley Forgets Every Three Years That the Average American Is Not Trying to Be Efficient Tech founders routinely overestimate demand for personal productivity apps by assuming everyone shares Silicon Valley’s obsession with optimization. Most consumers do not wake up wanting to grind down their inboxes or maximize output, making personal productivity a notoriously narrow and difficult category to scale.
#35 🔗 原推
2026-08-27 14:05:09
Harry Stebbings
I repeat, Paul is the most analytical VC in Europe and this is a must read 👇
中文: 我再说一遍，保罗是欧洲最具分析性的风险投资者，这是必读的👇
#34 🔗 原推
2026-08-27 12:26:21
Harry Stebbings
It is a travesty that sports stars are not educated more efficiently on how to manage their money in their playing careers. Clubs have a responsibility to their players to equip them for a career off the field also. https://www.cityam.com/jamie-carragher-hmrc-petitions-for-sky-sports-star-to-be-declared-bankrupt/
中文: 体育明星在如何管理自己职业生涯的资金方面，并没有得到更高效的教育，这是一种讽刺。 俱乐部有责任让球员也具备在场外职业发展的装备。
#33 🔗 原推
2026-08-27 11:02:06
Harry Stebbings
Lord help us. Scale up fund targeting science and we want these managers to run it… Should be interesting!
中文: 主啊，求你帮助我们。 扩大基金目标科学的规模，我们希望这些管理者能够管理它...... 应该很有趣!
#32 🔗 原推
2026-08-26 11:45:18
Harry Stebbings
Why will Triple Triple, double double come back? "I don't know if it will be three years, maybe five, but this will come back. Some are new markets, and some are replacement markets. In the case of a CRM company, they might be AI-native, but they're still having to replace a core system of record for a business. In a few years from now, most of the customers out there will have a solution and will hit a replacement market. But right now it's apples and oranges." @JulienBek Have core company scaling dynamics changed or do you think triple, triple, double, double comes back to being attractive in venture @mmurph @nchirls @bhalligan @km @nbt
🎬
视频
#31 🔗 原推
2026-08-26 01:08:00
Harry Stebbings
Which partner at Sequoia is the best at selecting companies? "That one's easy. @LucianaLix, my partner, actually brought me into Sequoia. We worked together at Accel before, so I've worked with Luciana for most of my career. When I met her, she had just invested in Deliveroo. Then she did Framer (@jornvandijk), Pennylane (@Ar_Waller), and Stark (@FlorianSeibelQS). It's just banger after banger. If you look at the pattern, there's no pattern. She's been able to reinvent herself across different categories, from consumer to software to physical AI and defense." @JulienBek
🎬
视频
#30 🔗 原推
2026-08-25 20:05:00
Harry Stebbings
Who is the best sourcer in Sequoia? "I will pick @DeanMeyerrr, my partner who sits in Tel Aviv but basically lives on a plane. He's just a phenomenal human being. He has the competitive juices of Messi, coupled with the technical depth of someone who's been working in tech his whole career, and that's a very dangerous combination. He's just amazing at reading people. He's got this ability to connect with founders, both the very young, spiky people and the guys who sold companies for billions of dollars." @JulienBek Single most impressive element about Dean @gradypb @shaunmmaguire @BogieBalkansky
🎬
视频
#29 🔗 原推
2026-08-25 17:38:00
Harry Stebbings
Why Sequoia is not less ownership-centric than ever "The outcomes are growing. It's also more capital intensive, but most importantly, it's your time. In my career, I can expect to be on the board of 20 companies. I'm not going to short myself. I'm going to work really hard for those founders. I'm basically their co-founder. They decide how to run the business, but I sit in the passenger seat and help them close their first customers and top hires." @JulienBek What is the single most needle-moving element of partnering with Sequoia @FDavidsonT @jamiecuffe @gorkem @EugenAlpeza @wwillsun
🎬
视频
#28 🔗 原推
2026-08-25 15:37:00
Harry Stebbings
Why investing in a neo lab now is like investing in Quora or StumbleUpon: "I think right now, if you're going to invest in the new Neo Lab, you're basically investing in Quora or StumbleUpon when Facebook and X came about." @JulienBek I would love to hear how you think about this @nikesharora @AnjneyMidha @deedydas
🎬
视频
#27 🔗 原推
2026-08-25 02:12:00
Harry Stebbings
Why Sequoia Invested $250M into Anthropic: "It's very important that you update your priors if the environment has changed. The human brain is just not very good at dealing with exponentials. We can think very well linearly, but not exponentially. In this case, I think we underestimated the company in the early days." @JulienBek I have to ask, what specifically did you not see that you wish you had seen @gradypb @Alfred_Lin @shaunmmaguire
🎬
视频
#26 🔗 原推
2026-08-25 00:38:00
Harry Stebbings
How Sequoia came to be the first ambassador in Citadel "@Konstantine helped us lead the investment in Citadel Securities, Ken Griffin's company. They had never taken outside capital. The reason we were able to invest is Konstantine built a relationship with Ken since he was a student. He had been his mentor for years and years. Konstantine never gave up and just kept asking, 'Can we invest?' Until Ken kindly said yes." @JulienBek
🎬
视频
#25 🔗 原推
2026-08-24 22:48:00
Harry Stebbings
What everyone thinks they know about Sequoia but actually gets wrong "Everyone thinks that we're just waiting for the phone to ring for the next Anthropic to call us to invest. That's completely false. Everyone at Sequoia is a hunter. It doesn't matter how long you've been here, everyone expects you to perform. It's very competitive out there. We think that people need to behave exceptionally well as individuals, but win as a team." @JulienBek What does everyone think they know about Sequoia that they actually get wrong @sonyatweetybird @Alfred_Lin @gradypb
🎬
视频
#24 🔗 原推
2026-08-24 20:25:00
Harry Stebbings
My first day at Sequoia and a lesson from Doug Leone "I show up to the office at maybe 4:30 AM. I felt very happy about myself hustling to the office that day. As I'm about to push the door, I see a man on the other side, and he looks at me and goes, 'What are you doing here so early?' I tell him, 'I'm here to take my first call. What are you doing here so early?' And he says, 'I've already taken my first call.'" @JulienBek One thing, single biggest lesson from Doug, what would it be @shaunmmaguire @giliraanan @DavidCahn6
🎬
视频
#23 🔗 原推
2026-08-24 17:40:00
Harry Stebbings
Paul is the smartest person in European venture. This is a must read!
#22 🔗 原推
2026-08-24 16:54:57
Harry Stebbings
Spotify 👉 https://open.spotify.com/episode/6xEx43wrdjryvXchA2yKhN Youtube 👉 https://youtu.be/N8CBejLRztg Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-inside-sequoias-investment-committee-lessons-from/id958230465?i=1000785432770
#21 🔗 原推
2026-08-24 14:35:29
Harry Stebbings
Warning: this is a slightly soppy post! I first met Julien Bek 8 years ago. It was my first week at Atomico and his first week at Accel. We instantly became friends. We both have parents with chronic illnesses and I think that bonded us early on. I saw what a truly good human he is. That was very clear. Over the next 8 years, he has become one of the best investors in Europe. He is a Partner at Sequoia. He has led early rounds in bangers like Rillet and Tacto. Despite all the success, he is one of the kindest people I know. He is an incredible son, and it makes me so proud and happy to see him become a Dad. This was one of the most special shows I have done, uncovering the magic behind what makes Sequoia one of the best firms in the world. Huge thanks to @DeanMeyerrr, @gradypb, @Konstantine, @shaunmmaguire, @dougleone, @nataliemiyake, @Bryce_Keane, @LucianaLix, @_georgerobson for helping to make this such a special one. My notes below with @JulienBek. 1. What Everyone Thinks They Know About Sequoia but Actually Gets Wrong Outsiders assume Sequoia sits back and waits for the hottest deals to come to them. In reality, every partner operates as a relentless hunter. Each person is expected to perform individually while working as a team to win the most competitive deals. 2. How Sequoia Came to Be the First Ambassador in Citadel Sequoia won Citadel Securities’ first outside capital round because partner Constantine built a relationship with Ken Griffin that began when Constantine was a student. Years of persistence, mentorship, and trust ultimately beat transactional dealmaking. 3. The Biggest Takeaway From Every Sequoia Offsite Decades of legendary returns show that financial engineering and ownership tweaks do not drive top-tier performance. The common thread behind Sequoia’s greatest investments is much simpler: a sponsoring partner with extraordinary conviction. 4. Why Sequoia Is Experimenting With Different Types of Decision-Making Live IC meetings are great for fast debate, but Sequoia is incorporating asynchronous written memos to encourage slower, more deliberate thinking. Combining documented reflection with live discussion helps expose blind spots and improve investment decisions. 5. Why Sequoia Is Not Less Ownership-Centric Than Ever Targeting high ownership reflects the scarcity of a partner’s time. An investor can realistically serve on only around 20 boards over a career. Deep, hands-on company building becomes impossible when attention is diluted across hundreds of tiny 2% positions. 6. Lesson From Don Valentine on Founder Selection Don Valentine’s matrix of “founders you like” versus “founders who make money” shows that likability does not determine returns. Even arrogance can be the byproduct of an exceptional strength. Investors should focus on whether that defining spike creates a genuine competitive advantage. 7. The Biggest Lesson From Doug Leone To uncover the truth in reference checks, use Doug Leone’s technique: ask for a founder’s best reference, then immediately ask, “Who would be your worst reference, and why?” Watching how their composure shifts can reveal far more about self-awareness and operating style. (links in comments)
#20 🔗 原推
2026-08-24 14:35:12
Harry Stebbings
I have interviewed 1,000 founders and had the fortune to invest in many of them. The top 5 founders that I have ever met (not in order): - @alanchanguk (Fuse Energy) - @awxjack (Airwallex) - @ceo_clickhouse (ClickHouse, hanging out today in pic!) - @lqiao (Fireworks) - @MaxJunestrand (Legora)
media 0 共 1 项
#19 🔗 原推
2026-08-23 16:43:56
Harry Stebbings
"OpenRouter is really strong in developer-type tools where you want a simple way to pick a model. It's really strong with chatbots where they don't have to be perfect. For frontier-esque models, people don't rotate through 11 models, and I don't think OpenRouter is the right product for that. The risk to Stripe is that they end up owning a successful niche product, and that's not their DNA." @jasonlk Love to hear your thoughts on this and the potential “niche” market @shensi @eglyman @awxjack @ThibaultJaigu @zachmoskow
🎬
视频
#18 🔗 原推
2026-08-21 22:56:00
Harry Stebbings
Why the OpenRouter deal does make sense "Stripe actually appears to be very good at acquisitions. It's how it accelerated into crypto and otherwise. They're good at it. If you're good at M&A, and this is 5% of your market cap plus cash, and you want it tomorrow, it makes sense. I love OpenRouter. I'm a customer, I'm a user. It was one of these pieces of software which is just instantly easier to deploy. It's just elegant." @jasonlk What is your bull case on where OpenRouter will be in 5 years, given the acquisition @deedydas @AnjneyMidha @davefontenot
🎬
视频
#17 🔗 原推
2026-08-21 20:09:00
Harry Stebbings
"I would much rather initially work for Elon than for Zuck. I tell founders to ignore the brand. Ignore what you think the job is today, because you have no idea in 24 months what the hell you're going to be doing. It is incredibly emotionally important to founders to land in something they want to land in. I would not want to land at Meta today." @jasonlk @carlrivera @glencoates @RamaswmySridhar @shishirmehrotra @michaelginzo what advice would you give to founders selling their company and choosing the acquirer?
🎬
视频
#16 🔗 原推
2026-08-21 17:28:00
Harry Stebbings
We wrote a $10M check into Fireworks in 10 mins. Two reasons: 1. Lin and Dmytro are literally best in the world for what they do. Top 0.000001%. 2. Companies of the future will have specialised intelligence built on their own models, with their own data. Fireworks will help them do so. Because of both of these, I wrote $500BN as company size in 5 years. @P_Bonnet said $300BN. I think we will both undershoot it massively...
#15 🔗 原推
2026-08-21 15:33:45
Harry Stebbings
Spotify 👉 https://open.spotify.com/episode/3yxdVI9XhzrgM6uputn6d2 Youtube 👉 https://youtu.be/y_3EjW0dyeI Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-spacex-buys-cursor-for-%2460bn-stripes-%248bn-openrouter/id958230465?i=1000784501578
#14 🔗 原推
2026-08-20 14:03:11
Harry Stebbings
If you are not staying up to date with the most pressing news, you are doing a disservice to yourself and your company. This is the only show you have to listen to every week and what a week’s worth of news it was… AGENDA: - SpaceX Buys Cursor for $60BN - Stripe's $8BN OpenRouter Bet - Anthropic's First Profit & The Math Behind Reaching $600BN in Revenue? - Lovable and Higgsfield Raise Mega Rounds My notes below with @jasonlk and @rodriscoll: 1. I Would Rather Be Acquired by @elonmusk Than Zuck Founders often prefer selling to an iconic, highly effective operator like Elon Musk over entering Meta’s corporate structure. Despite advice to ignore brand prestige, emotional alignment and shared vision frequently play a major role in determining the ultimate M&A destination. 2. What Buyout Financiers Should Look for in Companies Today Private equity buyers should target closed systems of record with near-zero churn and highly predictable cash flows. Closed architectures protect ecosystem budgets and create a defensible moat against disruption from third-party AI agents. 3. Why OpenRouter Is a Niche Product That Could Lead to a Bad Acquisition for Stripe Multi-model routers thrive in developer environments, but high-reasoning B2B workflows often standardize on specific models to prevent drift. Stripe risks acquiring a niche tool serving narrow developer use cases rather than a platform with broad enterprise transaction potential. 4. Why Revenue Is So Weird in M&A: The Tale of Two Worlds PE buyouts require precise accounting around existing revenue, while strategic acquirers can largely ignore it. Strategic M&A prices platforms on future expansion potential, sometimes abandoning legacy revenue streams entirely to unlock a much larger market opportunity. 5. Why the OpenRouter Deal Does Make Sense Paying a premium for an elegant, deployable product can be far more efficient than building the infrastructure internally. Even if OpenRouter remains a niche tool, an acquisition could provide immediate access to massive AI inference flows and create a critical second growth engine. 6. Why Optimism Beats Pessimism in Hyper-Growth Markets Fixating on early unit economics can make investors sound smart while causing them to miss massive market waves. In booming categories like AI coding, products scaling rapidly despite margin headwinds can still become category leaders and attract strategic buyers capable of absorbing those costs. 7. Why Legacy Software Roadmaps Can’t Survive the Agentic Era Rapidly improving AI capabilities have broken the traditional quarterly software roadmap. Engineering teams embracing agentic workflows can pull years of planned development forward, leaving slower legacy teams built around incremental release cycles increasingly behind. (links in comments)
#13 🔗 原推
2026-08-20 14:02:32
Harry Stebbings
"Travis used to have a saying: 'We need to raise more money than all our competitors in the world combined.' The basis for competition for rideshare was clear, so it was just a land grab at that point. Money helped you solve the land grab." @andrewgordonmac Do you believe we are in a similar market now, where capital truly is the moat for many businesses @travisk @altcap @Alfred_Lin @nikesharora
🎬
视频
#12 🔗 原推
2026-08-17 22:50:00
Harry Stebbings
Was Uber Right to Stop Investing in Autonomous "During Covid, our mobility business had lost 84% of our top line in three weeks. The company was burning billions annually, and we didn't have a core business producing cash. We did not believe we were leading in autonomy at the time. We were trailing, and Uber had a lot to prove that we could lead, win, and make money in our core business. We divested ATG. We turned the core businesses into cash-flowing machines, took the company public, and grew the business. Almost any metric you pick from that point in time is up and to the right." @andrewgordonmac Given the challenging time at the time, was Uber right to divest of ATG, in your mind @typesfast @marceloclaure @jason @DavidSacks @epaley
🎬
视频
#11 🔗 原推
2026-08-17 20:06:00
Harry Stebbings
What does Uber need to do to get to 500 million users? "Taking an UberX to and from work every day in New York City for $35 a direction, that's still a luxury product. If we want to get to 500M users, and go from using us six times a month to 25 times a month, the average cost of that transaction has to come down." @andrewgordonmac I would love to hear your thoughts @jason @cyantist @BillAckman @shervin other than price, what is the single biggest barrier to Uber hitting 500M users?
🎬
视频
#10 🔗 原推
2026-08-17 17:36:00
Harry Stebbings
@Uber Spotify 👉 https://open.spotify.com/episode/3pGXxX3zdAST7nTQshXIho Youtube 👉 https://youtu.be/sCa1MWB9Gcc Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-uber-president-on-the-untold-uber-stories-travis/id958230465?i=1000783757706
#9 🔗 原推
2026-08-17 14:41:38
Harry Stebbings
I have interviewed 1,000 of the best operators of the last 10 years. Ironically, one of the best I have ever interviewed isn’t actually a Founder. He is the longest serving employee at @Uber. Andrew MacDonald. He rarely does interviews and this was the most insane behind the scenes at one of the worlds largest companies: - Spending $52M per week in China - Lessons from Travis Kalanick - Why autonomy is existential - How AI helps and hurts Uber There was so much in this, I took notes and have added them below: 1. How Does Uber Decide What New Products to Do Versus Not Do? At nearly $250 billion in gross bookings, new mobility or delivery products must show a credible path to billions in transaction volume within a few years to justify organizational resources. Anything smaller risks being swallowed by the demands of the core business. 2. How Companies Need to Extract AI Efficiency Measuring AI ROI on a per-employee basis is nearly impossible because saved hours simply get absorbed by other work. Instead, leaders should capture efficiency through OpEx by capping or reducing headcount targets and requiring teams to produce more with existing resources. 3. Uber Have a SWAT Team of the Best AI Engineers Uber created a dedicated pod of 30 elite AI engineers paired directly with G&A teams and business process owners. By rebuilding workflows from the ground up, they cut weekly pricing processes from 15 hours to two and marketing QA from two weeks to two days. 4. Why We Were Right to Focus on Core Strategy and Divest the Autonomy Business at the Time When COVID erased 84% of Uber’s mobility revenue in three weeks, the company divested its ATG autonomous driving unit. Exiting a capital-intensive race where Uber trailed competitors freed the company to focus on turning its core marketplace into a cash-flowing machine. 5. Will Uber Have More or Fewer Employees in Five Years’ Time? While AI will create entirely new industries, execution-heavy enterprise functions will see significant headcount contraction. Roles centered on customer support, sales ops, content production, and reporting analytics will face substantial augmentation and partial replacement by AI agents. 6. Crazy Story Number One From Working in China At the peak of its market-share war with DiDi in China, Uber burned $52 million per week on price subsidies alone. Competing without access to WeChat made winning nearly impossible, but aggressive capital deployment gave Uber the leverage needed to negotiate a successful local exit. 7. Single Biggest Lesson From Travis Kalanick Travis Kalanick’s operational superpower was walking into a meeting, asking pointed questions, and advancing weeks of expert thinking in 15 minutes. Great leaders create leverage by constantly exercising this problem-solving muscle and teaching the principles behind their decisions. (links in comments) @andrewgordonmac
media 0 共 1 项
#8 🔗 原推
2026-08-17 14:41:25
Harry Stebbings
@Uber Spotify 👉 https://open.spotify.com/episode/3pGXxX3zdAST7nTQshXIho Youtube 👉 https://youtu.be/sCa1MWB9Gcc Apple Podcasts 👉 https://podcasts.apple.com/us/podcast/20vc-uber-president-on-the-untold-uber-stories-travis/id958230465?i=1000783757706
#7 🔗 原推
2026-08-17 14:19:52
Harry Stebbings
I have interviewed 1,000 of the best operators of the last 10 years. Ironically, one of the best I have ever interviewed isn’t actually a Founder. He is the longest serving employee at @Uber. Andrew MacDonald. He rarely does interviews and this was the most insane behind the scenes at one of the worlds largest companies: - Spending $52M per week in China - Lessons from Travis Kalanick - Why autonomy is existential - How AI helps and hurts Uber There was so much in this, I took notes and have added them below: 1. How Does Uber Decide What New Products to Do Versus Not Do? At nearly $250 billion in gross bookings, new mobility or delivery products must show a credible path to billions in transaction volume within a few years to justify organizational resources. Anything smaller risks being swallowed by the demands of the core business. 2. How Companies Need to Extract AI Efficiency Measuring AI ROI on a per-employee basis is nearly impossible because saved hours simply get absorbed by other work. Instead, leaders should capture efficiency through OpEx by capping or reducing headcount targets and requiring teams to produce more with existing resources. 3. Uber Have a SWAT Team of the Best AI Engineers Uber created a dedicated pod of 30 elite AI engineers paired directly with G&A teams and business process owners. By rebuilding workflows from the ground up, they cut weekly pricing processes from 15 hours to two and marketing QA from two weeks to two days. 4. Why We Were Right to Focus on Core Strategy and Divest the Autonomy Business at the Time When COVID erased 84% of Uber’s mobility revenue in three weeks, the company divested its ATG autonomous driving unit. Exiting a capital-intensive race where Uber trailed competitors freed the company to focus on turning its core marketplace into a cash-flowing machine. 5. Will Uber Have More or Fewer Employees in Five Years’ Time? While AI will create entirely new industries, execution-heavy enterprise functions will see significant headcount contraction. Roles centered on customer support, sales ops, content production, and reporting analytics will face substantial augmentation and partial replacement by AI agents. 6. Crazy Story Number One From Working in China At the peak of its market-share war with DiDi in China, Uber burned $52 million per week on price subsidies alone. Competing without access to WeChat made winning nearly impossible, but aggressive capital deployment gave Uber the leverage needed to negotiate a successful local exit. 7. Single Biggest Lesson From Travis Kalanick Travis Kalanick’s operational superpower was walking into a meeting, asking pointed questions, and advancing weeks of expert thinking in 15 minutes. Great leaders create leverage by constantly exercising this problem-solving muscle and teaching the principles behind their decisions. (links in comments) @andrewgordonmac
#6 🔗 原推
2026-08-17 14:19:39
Harry Stebbings
OpenRouter is one of the most insane stories in tech. Co-founded by OpenSea founder, Alex Atallah. It has become one of the most important companies in AI. Scaled to 100s of millions in revenue and wildly profitable. They process 25 TRILLION tokens every week and will do 1…
#5 🔗 原推
2026-08-10 15:29:44
Harry Stebbings
90% of the podcasts you hear on AI today are BS. The guests are terrified to upset the core model providers, their dominant source of revenue. And I get it but that is why @ml_angelopoulos is one of the best shows we have done in recent times. The most direct, no s**** given…
#4 🔗 原推
2026-08-03 14:40:14
Harry Stebbings
Everyone gets angry with me for saying triple, triple, double, double is dead. Fine, I do not really care. Venture is about investing in unbelievable outliers. Anomalies that own markets with generational founders. @FireworksAI_HQ is an example of this. They scaled to $1BN in…
#3 🔗 原推
2026-07-20 15:06:55
Harry Stebbings
Why Remote Work is White Collar Fraud. "I have a three-year-old and a five-year-old. The idea that I could do any work at my house is like a total fantasy. The kids come home at 3pm, your work day needs to keep going. I'm highly against it." @typesfast https://twitter.com/HarryStebbings/status/2068862030166282597/video/1
中文: 为何远程工作是白领舞弊。 我有一个三岁的孩子和一个五岁的孩子。我本可以在家里做任何工作的想法，就像一种完全的幻想。 孩子们下午3点回家，你的工作日需要继续。我对此非常反对。@typesfast
media 0 共 2 项
🎬
视频
#2 🔗 原推
2026-06-22 01:02:00
Harry Stebbings
I have never met a top-performing CEO who likes the role of HR. They are here to slow us down and instill meaningless process.
中文: 我从未遇到过一位对人力资源职位有过最佳表现的CEO。 他们在这里是为了延缓我们，并灌输毫无意义的过程。
#1 🔗 原推
2026-04-19 13:10:33