🌐 双语
Archive

AI Builders
Digest

2026-11-08 10 builders · 19 tweets · 1 podcasts · 0 blogs

🔥 热点话题

Robinhood CEO Vlad Tenev:所有权、Tokenization 与 Agentic TradingRobinhood CEO Vlad Tenev on Ownership, Tokenization and Agentic Trading

The Takeaway:真正的所有权民主化——让普通人从出生起就能持有美国经济,并通过 agent 和 tokenization 把对冲基金级能力装进口袋——才是对抗 AI 集中化冲击的最佳路径。

Robinhood 创始人兼 CEO Vlad Tenev 把公司定位为「所有权公司」。从东欧童年经历高通胀与货币崩溃的他,坚信广泛资产所有权是个人财务自由与社会稳定的基石。Robinhood 已从零佣金股票 App 进化为覆盖 120 多国的金融综合体,核心产品包括 Trump Accounts(为新生儿注入 1000 美元并投资标普 500 ETF)、Robinhood Ventures(让零售投资者参与 OpenAI 等晚期私有公司与 YC 早期项目)、以及完全 DeFi 可用的股票代币与 Robinhood Chain。

Tokenization 把美股、私有公司份额甚至实物资产「装进盒子」并铸造可在链上自由流通的代币,让没有成熟资本市场的国家用户也能低成本接触美国资产。与此同时,Agentic Trading 已验证强产品市场契合:上线数月即有超 15 万用户开通账户,日工具调用量达 3000 万次。用户可在沙盒账户中运行自主策略、组合再平衡,甚至接入卫星数据做车位计数策略。Tenev 强调,目标是把对冲基金与量化公司的能力下放到普通人手机里,就像 coding agents 让更多人成为软件工程师一样。

对 AI 数学革命,Tenev 既兴奋又警惕。他联合创办的 Harmonic 去年用 Aristotle 模型拿下 IMO 金牌,并让用户用它解决未解的 Erdős 问题。但他反对「大实验室用秘密模型与巨型数据中心垄断所有科学发现」的中心化路径,主张把工具交到个人手中,让随机普通人也能做出被庆祝的突破。他直言:「如果所有发现都由 AI 实验室垂直拥有并售卖,那将是一种反乌托邦结局。」

在监管与基础设施上,Tenev 支持知识型认证考试替代纯资产门槛,认为信息爆炸与 AI 合成能力已让旧有 accredited investor 规则过时。他相信,只要把风险讲清楚,人们有权自己决定如何投资。最终,他把厚脸皮视为创业者超能力——童年挑战让他习惯了压力,也让他能在市场与公关的剧烈波动中保持冷静。
The Takeaway: True democratization of ownership—giving every person a stake in the U.S. economy from birth and putting hedge-fund-grade capabilities into ordinary pockets via agents and tokenization—is the strongest antidote to AI’s centralizing force.

Robinhood founder and CEO Vlad Tenev frames the company as “the ownership company.” Shaped by an Eastern European childhood of hyperinflation and currency collapse, he sees broad asset ownership as both the personal route to financial independence and the collective stabilizer of society. From a zero-commission stock app, Robinhood has become a global financial conglomerate spanning 120+ countries. Flagship efforts include Trump Accounts (seeding every newborn with $1,000 in a low-cost S&P 500 ETF), Robinhood Ventures (giving retail investors exposure to late-stage privates like OpenAI and YC seed/Series A companies via publicly traded funds), and fully DeFi-enabled stock tokens running on Robinhood Chain.

Tokenization wraps real-world assets—U.S. stocks, private-company SPVs, art, real estate—into on-chain tokens that move freely, trade fractionally, and compose with smart contracts. For users in countries lacking functional capital markets, this becomes the default way to own U.S. equities, just as stablecoins became the default way to hold dollars. Agentic Trading has already shown clear product-market fit: more than 150,000 customers opened accounts in a few months, generating 30 million tool calls per day. Users run strategies inside sandboxed accounts, rebalance portfolios, and plug in external data (satellite car-counts, etc.). Tenev’s north star is putting the capabilities of a sophisticated quant firm into any smartphone, the same way coding agents turned more people into software engineers.

On the AI math revolution, Tenev is both excited and wary. His co-founded startup Harmonic produced Aristotle, one of three models to earn a gold medal at the International Math Olympiad; users have already applied it to unsolved Erdős problems. Yet he rejects the centralized model in which a handful of labs with secret models and giant data centers produce every discovery. He prefers tools that put power in individual hands so that “a random lay person” can make breakthroughs that are celebrated rather than resented. “If all of that is happening and it’s owned by the AI labs themselves, I think that’s a bit of a dystopian outcome.”

On regulation he backs knowledge-based accreditation exams over pure asset tests, arguing the internet and AI synthesis have rendered 1980s rules obsolete. Customers, he notes, overwhelmingly agree with the statement “I should be in charge of my money.” His personal advice to leaders: grow thick skin. Childhood adversity taught him that most business crises pale next to real existential risk, and that calm under fire is a trainable superpower.
查看原文 →

Aaron Levie:AI 带来数量级生产力跃升,且差距仍在扩大Aaron Levie: Order-of-Magnitude Productivity Gains from AI Are Already Real

Box CEO Aaron Levie 指出,在正确 harness、领域深度理解、较新代码库、多 agent 并行流畅度以及低摩擦测试部署的前提下,软件工程已出现比 AI 前高出一到两个数量级的产出提升。更关键的是,这种差距并未随时间收敛——先前深度与现在愿意采用全新工作方式的人反而获得了不断复利的杠杆。他预测同样模式将很快出现在科学、AEC、金融、网络安全等所有对计算机操作敏感的领域。

在另一条推文中,Levie 分析了长时运行 agent 对数据中心选址的影响:AI 推理对用户延迟并不极端敏感,任务经济价值远高于多一秒返回的成本,因此推理可在远程「AI 工厂」完成,数据加密进出后即弃,数据驻留约束也随之放松。这为愿意建设算力的州或国家创造了巨大经济机会,也让拒绝机会的地区付出代价。
Box CEO Aaron Levie observes that, with the right harness, deep domain understanding, relatively new codebases, fluency with multiple agents, and low-friction test-and-deploy pipelines, software engineering is already delivering productivity gains one or two orders of magnitude above pre-AI baselines. Crucially, the gap is not closing; prior depth plus willingness to adopt a radically different workflow continues to compound leverage. He expects the same pattern to appear soon in other computer-use-intensive domains—scientific research, AEC, finance, cybersecurity and beyond.

In a second post Levie notes that long-running agents change data-center geography. Inference is relatively latency-insensitive; the economic value of a completed task dwarfs an extra second of return latency. Work can therefore be shipped to remote “AI factories,” data enters encrypted and is discarded after use, relaxing traditional data-residency constraints. States or countries that seize the opportunity gain a major economic edge; those that forgo it lose.
查看原文 →查看原文 →

Garry Tan:超级智能下的「丰裕还是黑暗」将成为未来十年定义性政治议题Garry Tan: Abundance vs Darkness Will Be the Defining Political Issue of the Next Decade

Y Combinator 总裁兼 CEO Garry Tan 断言,我们本可以拥有丰裕。未来十年的核心政治问题将是:人类是否愿意通过超级智能在科学、医学与工程上取得进步,还是选择黑暗。他同时批评 WIRED 已沦为昔日荣光的空壳,并调侃某些反增长立场「千万别让 NIMBY 看到这篇论文」。
Y Combinator President & CEO Garry Tan declares that abundance is within reach. The defining political question of the next decade will be whether humanity chooses progress in science, medicine and engineering through superintelligence—or chooses the darkness. He also calls WIRED a shell of its former self and jokes that certain anti-growth arguments should never be shown to NIMBYs.
查看原文 →查看原文 →查看原文 →

🛠️ 开发者工具与技巧

Andrej Karpathy:三分钟用 Claude Code 一键生成 TristanTubeAndrej Karpathy: One-Shotted TristanTube with Claude Code in Three Minutes

AI 研究者 Andrej Karpathy 看到 Tristan 发布的顶级 YouTube 频道列表后,用 Claude Code 在三分钟内一次性生成了 TristanTube——一个包含 4.7 万条超硬核视频并带 TV 模式的观看平台,方便轻松刷完所有内容。
AI researcher Andrej Karpathy took Tristan’s curated list of top YouTube channels and, in a three-minute one-shot with Claude Code, built TristanTube: 47,000 super-nerdy videos plus a TV mode for laid-back binge watching.
查看原文 →

Thibault Sottiaux:澄清 ChatGPT / Codex Pro 套餐用量倍数Thibault Sottiaux Clarifies ChatGPT / Codex Pro Plan Multipliers

OpenAI Codex & ChatGPT 团队成员 Thibault Sottiaux 澄清:所有 Pro 套餐均无 5 小时限制。Pro 100 = 10 倍 Plus 用量,Pro 200 = 20 倍,Pro 500 = 50 倍并解锁 ultrafast。他同时公开征求社区对「Pro 50」档位的意见。
OpenAI Codex & ChatGPT team member Thibault Sottiaux clarifies that every Pro plan has no 5-hour limits. Pro 100 delivers 10× Plus usage, Pro 200 delivers 20×, and Pro 500 delivers 50× plus ultrafast access. He also asks the community whether a Pro 50 tier would make sense.
查看原文 →查看原文 →

Peter Steinberger:Asahi 项目不应反向增加限制Peter Steinberger: Asahi Should Not Add New Restrictions

OpenClaw 与 OpenAI 相关开发者 Peter Steinberger 批评 Asahi 项目方向:该项目本意是打开封闭受限的系统,现在却增加新限制并排斥 AI,这与开源精神背道而驰,也让其他项目更难在其基础上构建。
OpenClaw and OpenAI-adjacent developer Peter Steinberger criticizes the Asahi project’s recent direction. The original mission was to open a closed, restricted system; adding new restrictions and rejecting AI runs counter to that spirit and makes it harder for other projects to build on the work.
查看原文 →

💰 创业成功案例

Zara Zhang:创业命运是「做」出来的,而非命中注定Zara Zhang: Startup Fates Are Made, Not Predetermined

Builder Zara Zhang 反驳「某个点子注定失败、某个市场太小众」的宿命论。她指出,给定同一个想法,有上千种执行路径;一百个人会走出一百条完全不同的路。真正出色的人与出色的执行,往往能在一片空白中创造价值,或让曾经的「不可能」在事后看起来理所当然。
Builder Zara Zhang pushes back against fatalistic startup talk—“this idea will never work,” “that market is too niche.” Given one idea there are a thousand ways to execute it; a hundred people will take a hundred different approaches. Brilliant people with brilliant execution routinely create something from nothing or make the previously impossible look obvious in hindsight.
查看原文 →

Amjad Masad:哈萨克斯坦政府高度 AI 友好,Network School 落地Amjad Masad: Kazakhstan’s AI-Pilled Government and the Network School

Replit CEO Amjad Masad 分享对哈萨克斯坦的印象:政府极度 AI 友好,是 Network School 的绝佳落地选择。他同时推荐了「The Better Lesson」相关内容。
Replit CEO Amjad Masad reports that Kazakhstan’s government is highly AI-pilled and an excellent home for the Network School. He also highlights “The Better Lesson.”
查看原文 →查看原文 →

🌍 其他动态

Aditya Agarwal:数据中心建设受阻将利好现有算力持有者Aditya Agarwal: Harder Datacenter Builds Favor Existing Capacity Holders

SPC 合伙人 Aditya Agarwal 指出,如果新建数据中心变得更难,拥有现有产能的超级云计算厂商将明显受益——算力需求只会继续上升。代价是新模型训练节奏可能放缓,前沿进展将被「踩刹车」。
SPC General Partner Aditya Agarwal argues that if building new datacenters becomes harder, existing capacity holders (the hyperscalers) win big—demand for compute only keeps rising. What suffers is the pace of new model training; the frontier will be paced after all.
查看原文 →

Nikunj Kothari:讽刺「如何制造 Series A FOMO」的完整剧本Nikunj Kothari’s Satirical Playbook for Manufacturing Series A FOMO

FPV Ventures 合伙人 Nikunj Kothari 用一整段讽刺文完整复盘了当前制造 Series A FOMO 的标准操作:模糊发布收入曲线、多轮功能发布视频、找趋势公司做案例、买通「顶级」媒体、堆砌「long-horizon agent swarms」等咒语,最后假装「我们其实没在融资」。他标注整段是讽刺,但每一句单独拿出来都是真的。同时预告下周与 Arcanum Labs 联合创始人的闭门数据主题讨论。
FPV Ventures partner Nikunj Kothari delivers a full satirical playbook for manufacturing Series A FOMO: vague-post revenue charts, multiple feature launch videos, trendsetter case studies, paid “premier” coverage, magic phrases about long-horizon agent swarms, and the classic “we’re not actually raising.” He notes the whole thing is satire—yet every individual line is true. He also previews an off-the-record data-session with Arcanum Labs co-founder next week.
查看原文 →查看原文 →

Swyx:最新 swag 与贴纸到货Swyx: Fresh Swag and Stickers Arrive

smol.ai / Cognition 相关 builder Swyx 展示了新到的高质量 swag 与贴纸,并表示「你们下周有福了」。
smol.ai / Cognition builder Swyx shows off newly arrived high-quality swag and stickers, teasing that “you guys are in for a treat next week.”
查看原文 →查看原文 →