Crypto Twitter KOLs are a fucking scam. And we have the numbers to prove why. We hired 200+ KOLs with 10M+ combined followers. Every single one had a 400+ Sorsa score and 100+ Smart Followers. Most had 1,000+ Sorsa. 200+ KOLs posted twice each. Our result? 280 NFTs sold at $5. Now let’s talk about this circus. Fake views. Fake likes. AI-generated replies that don’t even understand what the project does. The same fucking accounts liking and replying to each other in engagement groups, recycling engagement and pretending they’re influential. You have 200K followers but can’t prove you have actual buyers. You’re scared of referral deals because deep down you know nobody will click your links, buy anything or convert. YOU HAVE BOTTED FOLLOWERS, NOT INFLUENCE. And the funniest part? You genuinely think you’re marketing professionals. You’re a bunch of engagement farmers playing influencer, selling fake numbers to projects and calling it a career. Our team tested this firsthand. We were ready to pay for influence, but got fucking nothing. And you wonder why CT is dying? Why projects are cutting deals? Why marketing teams are moving to TikTok and Instagram? BECAUSE NOBODY WANTS TO PAY FOR YOUR BULLSHIT ANYMORE. Honestly, go work at McDonald’s. At least you’ll serve people something useful instead of selling projects fake engagement and pretending to be important. 200+ KOLs. 10M+ followers. 280 NFTs sold. That’s your fucking influence.

周末和几位从Crypto转去做AI创业的朋友聊了一下午，越聊越觉得有点"为赋新词强说愁"的别扭感。 聊到personal agent的用户价值，几乎每个人讲的都是同一个故事：帮你订机票、订酒店，省下十几分钟... 但TikTok、小红书、X、fomo，没有一个是靠省时间火的。它们赢在让人心甘情愿地花时间。 人愿意停留、愿意付费，从来不只是因为效率，而是因为情绪：好奇、认同、刺激、FOMO、被看见。用户要的是一个他们觉得值得把时间花进去的场景。 而订票这件事，大部分普通人一年订几次，出错代价高，大多数人其实并不想交出去，顶多用agent来帮助筛选... 还在用2B逻辑想2C也许是这轮2C AI还没有出现aha moment的原因之一了。

我们都生活在叙事里 yuval在《人类简史》里说，人是受叙事驱动的产物。他的原文是： “‘虚构’这件事不只让我们拆解想象，更是让我们能够‘一起’想象，能编织出共同的虚构故事……这种虚构出来的故事赋予智人前所未有的能力，让我们得以大规模、灵活地合作。” 所有人都生活在叙事里。叙事基于社会环境而生，并深刻影响了我们的世界观和思维。 以中国为例。中国最宏大的统战叙事是“中华民族伟大复兴“，通过中国的经济发展、国力富强来复兴中华文明的影响力。现在确实是中国软实力和民族自信最强的时候，西方都开始在chinamaxxing。 过去十几年，当权者提倡的的主流叙事叫“中国梦“和”强国“，对应的结果是“实现全面小康”、“共同富裕“。某种程度上已经实现了，中国的社会环境相对稳定，居住舒适度一直在提升。 我们走出去，到其他国家生活，和老外接触交流，才会发现不同社会叙事的差异之大，人间的悲欢并不相通。 毕竟，不是所有国家都在变好。 川普的叙事是“MAGA“，他用各种反常的方式来告诉世界“美国第一，我们赢麻了”。但除了红脖子，其他人不太买账。 隔壁失落了很久的日本，也想复兴，一直不太顺利。早市高苗的竞选口号是“JIBA”，不知这次能否成功，起码口号让人能记住。 战争是叙事碰撞最激烈的方式。美国以色列的袭击甚至激活了伊朗的伊斯兰革命激情，战争持续下去，说不定会反向解构西方民主的叙事。 在币圈我们已经习惯了叙事轮换，比特币的叙事从抗通胀的电子支付现金演变到“数字黄金”；平均每年都有1-2轮叙事更替，最新的一轮是Agentic Commerce。 生活中我们的叙事也在演变。从新冠到现在，我亲身经过的一些叙事： > 疫情初期，数字游民的叙事兴起，我受益于自己在crypto行业很早就开启远程办公生活，有一段时间四处周游，也算是对“稳定体系”的解构 > 超级个体：这应该是AI时代最强的个体叙事了，当AI冲击社会生产力结构，这个叙事能激励高自主性的人用AI武装自己，改变命运 你看，大家逐渐不迷信宏大叙事了，都更关注自我了。 最重要的是，我们可以选择自己相信的叙事，最好是一个更长期的叙事，给我们走向未来的勇气和动力。 配的视频是李世默 2013年在牛津大学对答学生会主席的视频，名为New World Disorder 他用中西方社会叙事来回应中国政体更加有效，回击对方抨击中国缺乏民主和言论自由，直指对方的逻辑错误。语言水平和思辨性比江学勤老师高多了。值得一看。

一个百万浏览的 Launch Video 的价值越来越低 很多 AI Startup 都想凭借一条 launch video 就能 go viral，不惜花重金做视频，找KOL 转发，把曝光做到几十万或者百万，放在置顶。 结果给官推就带来几百个关注，几十个注册用户。这是当前市场的普遍情况。 推特推流算法的现状，短期容易出现集聚效应，迅速出现数据繁荣，但长期缺乏长尾价值，超过 48 小时完全没有自然流量。 推流算法是一方面，另一方面是用户侧。一条百万播放的视频可以抢占受众 1-2 秒的注意力，但今天的市场，激活一个用户平均要出现在他面前十次。 因此想较于一个 Launch Video，更重要的是 Launch 后的持续输出，直观的产品录屏、截图、或产品带来的结果。 正确的做法是： Founder 持续展示产品（用户更愿意看到真实录屏，而不是讲概念的 launch video）、作为内容角度的“发起方”，测试不同内容形式。 重点寻找 UGC 内容创作者，将产品真实融入工作流中，在不同内容形式下对受众持续激活。（小账号可能有更忠实的内容读者） 在早期用户中寻找 use case，挖掘内容形式并转化为 UGC 内容。甚至直接帮助 Early User 成为 Creator。 将 launch video 的成本花在这些地方，对于 AI startup 早期增长有奇效。

The future of AI is agent-native. Excited to kick off this journey together with Hermes Agent and the @NousResearch community. Qwen 3.6 Plus is now FREE for a limited time on Nous Portal — give it a try. ⚡

The AI supercycle will last 15 years. We're in year 3. Most investors are still buying Phase 1 names while the real money is already rotating into Phase 3. I mapped the entire cycle into 4 phases with the tickers that matter at each stage: The AI supercycle is the biggest investment theme of our generation. Bigger than mobile. Bigger than cloud. A 15 year structural shift that will reshape every sector of the global economy. Hyperscalers just committed $725 billion in capex for 2026, nearly doubling last year. Microsoft, Google, Amazon, and Meta each spending over $100 billion individually. This is not speculation. I've mapped the entire supercycle into four phases so you know exactly where we are and where the asymmetric opportunities sit. 🔴 Phase 1: Already Ran (2023 to 2025) The foundation layer is complete. $AMD ran 78% in 2025, $NVDA 39%, and $INTC just posted a blowout Q1 that sent the Philadelphia Semiconductor Index above 10,000 for the first time. Chips still power every phase but the generational entries are gone and risk/reward has compressed. - $NVDA, $AMD, $ARM, $INTC, $AVGO, $MU, $GLW - Semiconductors, Memory & Storage,Photonics/Optics - Foundation complete. Still growing but priced for it. 🟠 Phase 2: Peak Buildout (2025 to 2027) The phase most investors just woke up to. $CEG acquired Calpine to become the largest U.S. private power producer at 55 GW. $GEV up over 200% in a year. $VRT co engineering cooling for NVIDIA's Rubin architecture. $GLW up 74% YTD on optical fiber demand. Nuclear SMRs are the breakout with $OKLO, $SMR, and $BWXT positioning to power data centers directly. Still upside but the obvious names have moved. - $CEG, $GEV, $VRT, $VST, $TLN, $ANET, $GLW, $MOD, $EQIX $OKLO, $SMR, $BWXT, $NNE - Power/Grid, Cooling, Networking, Nuclear/SMR Peak buildout. - Nuclear SMRs are the sleeper. 🟡 Phase 3: The Positioning Window (2026 to 2028) Where AI escapes the data center and enters the physical world. Most will be late. Tesla converting Fremont to Optimus production, $25B capex, mass production targeted H2 2026. Rocket Lab posted record $602M revenue with $1.85B backlog. $LUNR up 47% YTD with $943M in contracts. $KTOS Valkyrie drone selected for the Marine Corps. The window to position is open right now. - $TSLA, $RKLB, $LUNR, $KTOS, $AVAV, $PATH, $ISRG $MP, $FCX, $ALB, $ASTS - Robotics/Autonomy, Space/Defense/Drones, Rare Earths - This is where the asymmetric risk/reward lives. 🟢 Phase 4: Final Frontier (2028+) The endgame. Microsoft capex $190B. Alphabet $190B. Amazon $200B. Meta $145B. Google Cloud backlog past $460B. They're building the rails for AI software dominance and AGI. Quantum still early but $IONQ and D Wave are laying groundwork. The platforms that control the software layer win the entire supercycle. - $MSFT, $GOOGL, $AMZN, $META, $ORCL, $IONQ - AI Software Dominance, AGI Infrastructure Decade long thesis. - Accumulate on weakness. 💊 Key Takeaway - Phase 2 is confirmed ($725B hyperscaler capex) - Phase 3 is where the smart money positions nowRobotics, space, defense, nuclear - SMR are the 2026 to 2028 trades - Most will rotate into these names 12 months too late 15 year supercycle. Not a trade. Phase 1 ran. Phase 2 is priced. Phase 3 is where you want to be.