Create order amidst chaos| Drive lifelong learning with sustainable curiosity | cooking @AiSteamLabs

Joined November 2021
In the next ten years, two unmissable tickets: IONQ + RPRX (the only portfolio that can outperform BTC) 1/ Imagine you hold two keys in your hand: One that can unlock the secret room of tomorrow's computing power (IONQ); A dam (RPRX) that can connect to the cash flow of drugs. The former is a rocket, and the latter is a dam. One soars into the sky, the other holds back floods. This is what I call the "ten years without getting off" combination. 2/ IONQ = A cloud factory that turns atoms into musical instruments. Engineers press Enter in the cloud, and in a laboratory on the other side of the Earth, ions are suspended in electromagnetic fields, lasers strike quantum states like conducting wands, completing extreme calculations time and again. You don't need to understand vacuum chambers or laser physics; you can access it just like renting a server. This is not science fiction; it's a reality that can be accessed today on AWS / Azure / Google Cloud. 3/ Why should you care? Classic computers are like a straight highway, getting faster and faster; quantum computing is like branching off into parallel universes, where for certain tasks it's not just "10% faster," but "completely changing the map." What IONQ is doing is turning this door into a cloud service access card: enterprises can try it first, run it, and scale up, without waiting for a "perfect day" to get on board. 4/ The key evolution from experimental art to engineering product IONQ is integrating "laser shows" into more scalable control systems—by acquiring Oxford Ionics, moving towards "building quantum machines like servers." Transitioning from scientific experimentation to engineering is the true watershed in this track. 5/ RPRX = The "toll operator" in the pharmaceutical industry. It doesn't produce medicine or advertise—what it buys are the tolls for passing through: for each bottle or injection of a therapy sold, it takes a small "toll." Scientists get early funding to accelerate research and development, while RPRX secures long-term, replicable cash flow. This is not mysticism; it's its core business: "the world's largest buyer of biotech and pharmaceutical royalties." 6/ How solid is this model? Back then, the Cystic Fibrosis Foundation sold the future royalties of a blockbuster drug to RPRX in a lump sum—immediately providing funds to expand patient support and research; as the drug goes global, royalties "ka-ching" back steadily. This is an example of "turning the future into today's fuel." 7/ RPRX is still adding new inlets to the dam. For example, it recently acquired a portion of the global sales royalties for Imdelltra (small cell lung cancer therapy)—not by selling the drug, but by capturing the rights to the cash flow. That’s its skill: choosing the right drug, signing good contracts, and sitting on the dam collecting “tolls.” 8/ Why are IONQ and RPRX linked together? Because true long-term navigation requires a propeller and a stabilizer. - IONQ is a thruster: if the "quantum crossroads" are opened year by year, the company with the cloud entrance will be the first to enjoy the dividends. - RPRX is a stabilizer: for every additional reliable pipeline and approved indication in global pharmaceuticals, a thicker layer of cash bread accumulates in the dam. 9/ A quick label for those who have never interacted with these two companies - IONQ: The quantum orchestra in the cloud — what you are buying is a "ticket to future computing power." - RPRX: The toll booth in the pharmaceutical world — what you are buying is a "toll for innovative drug passage." 10/ Ten-year perspective mindset (simplified version) Don't ask about today's fluctuations, ask about the scene ten years from now: - At the quantum gate, will it become a "basic infrastructure that everyone has to queue for," like GPUs? (If so, the entrance is in the cloud, with IONQ at the gate.) - Will medical innovation become faster, broader, and more enduring? (If so, the cash flow royalties dam will be fuller, and RPRX will be more "cyclical-resistant.") 11/ Summarize in one sentence IONQ is the "key to bringing future computing power into the cloud," RPRX is the "gateway turning innovation into a cash waterfall." Rocket + dam, attack when advancing, defend when retreating. Ten years, never get off the bus.
1
389
People who cured / are curing their own cancer on X: 1/ @sytses - bone cancer, billionaire founder of Gitlab, developed 10+ personalized therapies for himself. 2/ @paul_conyngham - shrank his dog Rosie's mast cell cancer tumors after chemo and surgery weren't enough. Now backed by YC and Founders Fund 3/ @miriamgonp - stage 4 breast cancer, tumor sequenced, searching for vaccine manufacturer now 4/ @kevinace - stage 4 lung cancer, angel investor in 300+ companies, first drug stopped working after 18 months. Vaccine designed, now paying $300k out of pocket for the vaccine 5/ @masonbgrant: going founder mode on his his 32-year-old wife's rare liver cancer, sequencing her tumor with AI agents 6/ @ArunRLverma: former product leader at Atoms and Shopify, brain tumor, built a vaccine from his tumor. Now building an AI health agent startup based on what he went through We're going to make it
107
629
112
4,872
454,385
Introducing Gemini 3.5 Transcribe 🚀 Our most precise speech to text model, that can handle over 85+ languages, has smart correction, and custom vocabulary. I vibe coded a Wispr Flow like app powered by the model. Demo + open sourcing below!
82
69
32
999
220,935
Socratic prompt:若要探求此言之真意,我们是否应先自问:何为“翻译”?当我们将一种语言的辞章转为另一种语言时,我们是在捕捉其表象的字句,还是在追寻其背后那不可磨灭的真理?若要求“一次不差”,这是否意味着我们不仅要保留其辞藻,更要保留那潜藏于字里行间的灵魂?若翻译是一场对话,那么我们又该如何确保那原始的意旨,在跨越语言的鸿沟后,依然保持其本初的纯粹? 2026年8月19日 致各位投资者: 我们致信各位,是为了宣布 OpenRouter 将加入 Stripe,这也是我们有史以来规模最大的一笔收购。在此之前,我们近期还先后收购了 Bridge、Privy 和 Metronome。我们认为,借此时机向大家分享我们如何在 Stripe 的战略框架下审视这些业务,将大有裨益。 Stripe 的宗旨是提升互联网的 GDP。当我们畅想一个繁荣的世界时,目光总会被那些让一切成为可能的底层支柱所吸引:诸如货币、信贷、通货、法律架构以及风险管理等运行机制。我们坚信,世界可以且理应比今天更加宏大与繁荣,而更优质的经济基础设施正是促成这一愿景的关键所在。 奇点 “奇点”(The singularity)是一个内涵模糊且或许已被过度滥用的词汇,但我们认定今年 1 月 1 日标志着奇点的开端,并自那时起便以此为基准展开运营。谈及奇点,人们往往联想到千禧年主义式的宏大预言;但就我们而言,我们只是观察到了长期趋势中的巨大拐点(例如新创企业成立速率的大幅飙升),并决定严肃对待这一阶段性的相变。 事实证明,Stripe 自创立伊始就坚持的“为开发者优化”,在诸多方面与“为代码测试架构(coding harnesses)和智能体(agents)优化”殊途同归——因为它们同样追求可编程性与零摩擦的部署体验。幸运的是,全球众多 AI 企业因此纷纷接入了 Stripe;我们也愈发清晰地认识到:构建互联网的经济基础设施,在很大程度上等同于构建 AI 的经济基础设施。 尽管显而易见变革将是极其深远的,但没有人能确切知晓 AI 将如何重塑我们的世界。许多智者的预言已然被现实彻底证伪。怀着对未知与不确定性的谦逊敬畏,我们确立了两个核心目标: 第一,我们希望加速 AI 在整体经济中的渗透与扩散。随着 AI 不断拓宽可能性的边界,我们目睹了大量令人欣喜的新产品与新服务如雨后春笋般涌现(相较于未来的惊人创造,这无疑只是序幕),而这些都需要不同且更契合的金融工具。AI 在经济领域的兴起也带来了新的挑战,例如新型的盗窃与欺诈,这需要依托前沿的尖端技术来进行有效防范与化解。总体而言,人们对 AI 的期许与共同夙愿,在于它能够带来更大的物质繁荣与丰沛,我们希望全力促成其实现。 第二,人们普遍担忧 AI 可能导致失业或权力集中,抑或二者兼有。我们认为,AI 的部署必须增强人的能动性(human agency),并且希望 Stripe 能够在捍卫经济自主权方面发挥作用——这是自由社会不可或缺的基石。伴随 AI 的演进,我们期盼能有更多新公司应运而生,并能以更高的效率与老牌巨头并肩运作、同台竞技。尽管早期数据印证了这一点(正如我们在 Stripe Economics Substack 及其他渠道所记录的那样),但没有任何结果是命中注定的。作为“小微力量”的坚定支持者,我们将竭尽全力保持前路畅通无阻。 我们始终坚信,全球经济的规模没有天花板(如今略超 100 万亿美元)。尽管乍听之下可能匪夷所思,但我们认为,思考一个规模达“千万亿美元(quadrillion-dollar)”的世界并逐一排查实现它所面临的瓶颈,是极具价值的。(如果全球人均 GDP 能够达到爱尔兰人均水平——约 10 万美元——我们就已经走完了 80% 的历程。) 如今,已有超过 500 万家企业接入 Stripe,经由我们的资金流占据了全球 GDP 的近 2%。我们很高兴拥有一个良好的开局,但相比未来几年可能实现的广阔图景,这些数字依然微不足道。 我们的构筑 宏观而言,全球经济必将迎来长足的增长;微观来看,商业的运转模式正在迅速更迭。现有企业正在调整其商业模式(计量计费模式迅速崛起,而许多传统模式日渐式微),并快速调动资源以推出新产品与新服务(速度与灵活性已成为当务之急)。企业创设正在加速。智能体正处于成为独立经济参与者的临界点。稳定币正在被快速采纳,且随着其成为 AI 经济的原生货币,这一势头可能还将进一步加速。鉴于代币(tokens)能够轻易实现跨境流转,全球覆盖能力变得比以往任何时候都更加关键。 我们正在竭尽全力构建这个时代所需的经济工具。原生稳定币支持与智能体可访问性的结合,正在催生一组全新的底层原语(primitives): 发现与入驻(Discovery + onboarding):Stripe Projects(使智能体能够注册第三方服务)、Stripe Directory(为智能体提供产品与服务检索发现)、Provisioning API(为智能体提供内嵌式注册)。 用量管理(Usage management):Metronome。 支付结算(Payment):Bridge(稳定币编排与调度)、Stripe's Agentic Commerce Suite(智能体商业套件)、Tempo(专为智能体打造的区块链)、MPP(机器支付协议 / Machine Payments Protocol)。 资金存储(Fund storage):Privy(加密/稳定币钱包)、Open Standard(一种新型稳定币)。 随着时间的推移,我们预计将发生结构性转变:相较于前 AI 时代构建的架构,“AI 经济”技术栈所占的份额将持续提升。这些产品的普及不一定会以惊天动地的方式呈现——我们正将它们深度集成到 Stripe 现有的产品与平台中,以确保任何企业都能轻松接入与使用。 为了确保 Stripe 在这一新阶段能发挥最大效用,我们密切关注全球成长最快、最具影响力的新型企业对我们的采纳情况。如今,《福布斯》AI 50 强企业中已有 88% 构建在 Stripe 之上(包括 OpenAI 和 Anthropic,剩余的 12% 大多数尚处于商业化前阶段);最新发布的 Brex 最快增长初创企业榜单中,这一比例更是达到了 100%。这些公司中绝大多数使用了超过 10 款 Stripe 产品,且 Stripe 来自 AI 公司与加密领域的收入占比均实现同比翻倍以上。我们希望随着时间的推移,Stripe 的成长轨迹能与整个 AI 部署的增长保持同步。 OpenRouter 纵览全局,我们看到“资本”与“智能”正在成为支撑每家企业运转的两大数字化流动要素。在此之前,每位开发者都需要一种直接、可靠的方式来管理其收入管道(revenue pipeline),对这一需求的服务孕育了 Stripe。而展望未来,每位开发者同样需要一种直接、可靠的方式来管理其智能管道(intelligence pipeline)。 正是这一洞察引领我们关注到了 OpenRouter。OpenRouter 构建了全球规模最大、最受信赖的 Token 路由引擎,不仅支持所有主流模型与供应商,且深受客户青睐。得益于其产品的实用性、创始人及团队的卓越能力,以及每周层出不穷的新模型发布,他们的业务正以惊人的速度增长(即便以 AI 行业的标准衡量亦属迅猛),今年迄今 Token 消耗量每周复合增长达 9%。 OpenRouter 对任何开发者而言都极为实用,而 Stripe 拥有全球最大的开发者平台之一。因此,将这两项核心需求结合在一起,必将带来巨大的协同效益与效率提升。 然而,促成此次整合的深层考量远不止于便利性。在与客户的深度合作中,我们意识到“智能”具有其特殊性:它昂贵、异构且瞬息万变。如同对待金融资本一样,企业必须以严谨且精细的方式权衡每一个单元的成本与回报。这项任务有多大价值?用哪个模型处理最为适宜?谁来付款、何时付款、这一延迟的时间价值几何? 我们在自身的产品中直接看到了“管理智能”与“管理资本”之间的异曲同工之处。例如,最初专为防范金融欺诈而设计的 Radar,如今在全球许多顶尖 AI 公司中,在抵御 Token 欺诈方面被证明极其有效。被 Anthropic、Nvidia 等行业领军者广泛使用的 Metronome 也表明,在 AI 语境下,计量计费与 Token 的分发供应及消耗本身是密不可分的。 我们预计该交易将在未来数周内完成交割。能够将 Stripe 的金融能力扩展到这一全新领域,并协助企业高效配置“智能资本”这一新型资产货币,令我们倍感振奋。 Stripe 业务表现 在过去的几个小时里,就像往常的每一个清晨一样,又有数千家新企业在 Stripe 上扬帆起航。从面向政府的结构化数据平台,到为园艺师量身打造的景观管理系统,如今的新创企业已几乎横跨经济的各个领域。创新前沿正展现出蓬勃的生机。 奇点似乎正在加速推动我们的核心业务。Stripe 上半年(H1)净收入同比增长 41%,上半年自由现金流同比增长 43%。到今年年底,我们旗下将有超过 10 款产品创造逾 1 亿美元的净收入,且其中许多产品在规模化扩张的同时依然保持高速增长。上半年,Stripe Billing 同比增长 71%;我们通过 Atlas 设立了第 100,000 家企业(目前特拉华州超过四分之一的新注册企业均通过 Atlas 完成);Stripe Capital 的活跃贷款数量突破 100,000 笔(同比增长 49%)。通过 Stripe Connect,新平台在 Stripe 上的激活速率较一年前翻了一倍有余。总体而言,Stripe 平台上的企业增长显著快于整体经济增速,这得益于选择效应(创新型企业更倾向于选择 Stripe),以及我们每年为促进客户收入增长所做的数千项细微改进的累积效应。 我们正在大力投资以拓展多条新产品线。例如,Stripe Treasury 是我们有史以来推出增长最快的产品之一,未来一年它将获得大量新功能并拓展全球覆盖范围。作为最便捷的在线支付方式,Link 的用户量刚刚突破 3 亿大关,其产品路线图上也排满了待推出的优化升级。对于 Stripe 的建设者而言,这是一个极具成就感的时代。 在推进扩张的同时,我们高度重视股东回报。Stripe 核心支付引擎的盈利能力,使我们能够在不显著稀释现有股东权益的前提下完成此类收购。即使在进行大规模的组织投资与并购的情况下,Stripe 当前的股份总数仍低于三年前。自 10 年前完成 D 轮融资以来,Stripe 的股价年化复合增长率达 31%,而同期标普 500 指数和纳斯达克指数分别为 14% 和 18%。我们对公司未来的发展前景比以往任何时候都充满信心。 当然,Stripe 如今依然是一家私有公司。在我们迈向奇点的诸多变迁与激流中,我们认为这构成了日益凸显的竞争优势。世界正变得愈发难以预测,每家企业都亟需从容笃定、敏锐娴熟的掌舵领航能力。我们深感庆幸能够拥有这样一个有助于我们坚守正途、着眼长远的治理架构。 感谢各位的信任与投资。我们将倾注全部热情与精力,确保 Stripe 不负所期、尽展潜能。 —— Patrick、John 与 Will
Stripe 写给投资人的信泄露出来了,清楚的说明了 Stripe 在过去这段时间激进的收购(Bridge、Privy、Metronome、Openrouter)背后的意图: 通过原生稳定币与 agent 经济的结合,搭建一套全新的 AI 时代的基础设施原语(primitives)。 具体来说,之前这系列收购和产品发布,战略框架是: - 发现与入驻 (Discovery + onboarding):Stripe Projects(让 agent 能够自主注册第三方服务)、Stripe Directory(供 agent 使用的产品及服务发现功能)、Provisioning API(为 agent 提供嵌入式注册接口)。 - 使用量管理 (Usage management):Metronome。 - 支付 (Payment):Bridge(稳定币发行)、Stripe 的 ACS 套件(Agentic Commerce Suite)、Tempo(稳定币公链)、MPP(机器支付协议)。 - 资金存储 (Fund storage):Privy(用户友好钱包)、Open Standard(豪华稳定币联盟)。 而对于 Openrouter 的收购,宏大的故事是: 资本与智能(intelligence)正成为支撑每一家企业运转的两大数字流,Openrouter 目前已经是全球规模最大且最受信任的 token 路由引擎,而未来,Openrouter 和之前这些收购搭建的基础设施结合在一起,可以让 Stripe 在 AI 时代,帮助企业高效分配“智能资本”这一新型货币。 注意:支付公司之前处理的是“钱”, AI 时代,他们想处理钱和 token——Stripe 说,“智能资本”将是一种新型货币——这将是金融服务能力的延伸。
1
3
1,404
The boundaries of imagination are the boundaries of magic, and AI is systematically expanding them, continuously building on that capability.
您不相信是因为您的技术深度不足以想象更极致的软件形态该如何被建模。 芙丽莲里有句话,想象力的边界就是魔法的边界。Fable 让人人都能当魔法使,这时产品经理有自己擅长的想象领域,框架设计者也有。AI 是在定向扩张我们各自的想象空间。 纯从框架设计的角度我很喜欢 DSH,主要是给了我很多启发。
1
77
We can finally talk about it: We found a way to extract hidden reasoning of frontier models using a vulnerability in the APIs of every frontier AI company. We verified that our reasoning token count matches billed API thinking tokens 1:1 for most of the prompts we queried.
380
1,960
620
13,516
3,543,088
new post on harness engineering for AI self-improvement: lilianweng.github.io/posts/2… It is hard to forecast how much the future of RSI will rely on harnesses. Likely harness engineering will evolve in the direction of self-improvement and enable auto-research, and, in turn, smarter models keeps harnesses simple. Even when many harness improvement get eventually internalized into core model, the need to specify goals and context will not disappear.
141
797
135
5,477
925,300
这老外比中国人还了解这些十八线城市…很有意思
I am pleased to introduce the Chinese Invisible Cities Index (CICI) - some Friday fun. 😏 These aren't the smallest or poorest cities. These are the cities with the least aura vs. their population. They make you say: "wait, I've never heard of it and it has HOW many people?"🧵
15
38
1
309
152,059
Personally, my intuition is that DeepMind’s culture might be more inclined towards playing infinite games, such as eradicating all human diseases. The current model competition, being a finite game, might not provide the team with enough stimulating feedback for an intellectual "brain orgasm."
1
1
88
Announcing a new division of Midjourney called "Midjourney Medical"
2,474
5,015
4,089
39,429
18,924,018
This take from the @PirateWires Daily 👇 https://xitter.cf/t.co/MDWks6q42g 🤣
1
2
107
RSI, and whether “SOTA rotation” is about to break. We’re used to SOTA rotating every few months. One lab pulls ahead, another catches up. Gemini was SOTA literally two quarters ago - feels like ages. But RSI - recursive self-improvement - could change the competitive dynamics. Dario on Dwarkesh’s podcast back in Feb: Dwarkesh challenged him - if recursive improvement is real, why does SOTA still keep rotating between labs? Dario’s answer was that, until very recently, the compounding advantage from AI-assisted AI research was still too small to really matter (but it's changing). Since Jan/Feb, I’ve heard more and more researchers talk about RSI. The idea is simple: you build a better AI, that AI helps you build the next better AI, and the loop starts compounding. It also explains part of the current “token-maxxing” dynamic - big co. spending $ billions on Claude Code just to keep up. And it helps explain the rumors that $OpenAI / $Google are reorganized around coding as the top priority. Once that loop crosses a certain threshold, it starts to look like a true industrial revolution. Horses were never going to catch up with cars.
7
11
1
66
11,700
masters through AI sickness to AI Zen.
51
We’re releasing TranslateGemma, a new family of open translation models with support for 55 languages. 🌐 Available in 4B, 12B, and 27B parameter sizes – they’re designed for efficiency without sacrificing quality.
187
908
304
6,761
2,192,371
分享一个具有科技美感的私钥保持方式,keepassx,你的key密码设置为32位以上,然后256位加密,可以任意存在任何一个云盘之上且无需担心破解。 ps:Bitfinex 的股东也在使用哦!/摆摆手。
分享一个保存私钥的方式,不少朋友担心直接把助记词写下来,就算藏到床底下也还是有被偷的可能,你可以把助记词本身再进行一次专属于你自己的二次加密。首先你找一本相对印刷量比较大的书,比如圣经和莎士比亚之类的。 然后你在这本书里依次找到你助记词的12个单词,这时候拿张纸,写下每个单词对应在哪里的“第几页”、“第几行”、“第几个”词。 于是你就会得到3*12总共36个数字,把这些数字写下来,并且务必记住这本书的名字和印刷版本。 写下这些数字的纸条丢了?不用怕,人家根本不知道这些数字意味着什么。 就算知道这些数字对应到某本书的助记词,人家怎么知道你用的是哪本书? 那本书被你自己搞丢了?也不用怕,再买一本同样印刷版本的就好。 你学会了吗?
109
OpenAI aren't talking about it yet, but it turns out they've adopted Anthropic's brilliant "skills" mechanism in a big way Skills are now live in both ChatGPT and their Codex CLI tool, I wrote up some detailed notes on how they work so far here: simonwillison.net/2025/Dec/1…
76
196
41
2,080
441,361
👏
Replying to @DefiLlama
LlamaAI now has the ability to search on 𝕏. Responses will incorporate community sentiment and other context from Crypto Twitter to add additional insights to DefiLlama data. Citations link back to their source on 𝕏 so you can read what was originally said.
31
产生收益很棒,但以某种方式绑定用户开放自行创建交易事件会更棒,pm 与 kalshi 当前最大的问题是封闭,利益也不与创作者相关。@cz_binance
Welcome a new prediction market on @BNBChain. When you make a prediction, you funds don't sit idle, they generate yield. Disclaimer: Founder is ex-Binance (a few years ago). Incubated/invested by YZiLabs. This tweet is not endorsement. 🙏
56