每3小时精选高质量推文
生成时间:2026-05-31 12:30 (Asia/Shanghai)
📊 统计概览
- 总筛选推文数:100 条
- 精选高质量推文:12 篇
- 筛选标准:点赞>10 或 转发>5 或 评论>5 或 内容深度>100字 或 来自知名作者
1️⃣ Susan Zhang - 中美教育与社会发展对比
作者信息
- Name: Susan Zhang
- Bio: @ Google Deepmind. Past: @MetaAI, @OpenAI, @unitygames, @losalamosnatlab, @Princeton etc.
- 发布时间: 2026-05-31 10:57 (Asia/Shanghai)
推文原文
18 years went by (2008-2026) without me visiting China, and I’ve been shocked by how much has changed.
One of the main draws to immigrating in the past was economic mobility through a much less rigid education system, where children were judged more “holistically” in college admissions rather than based on a single gaokao test score. But now, seeing how the western education system (public and private) is/has been crumbling (lack of funding, administrative bloat, AI slop curriculums, etc), it feels like in another 20 years the development gap will be even more stark.
It’s obviously not all rainbows and butterflies in China: poverty still exists and earning a living wage in things like the services industry is still incredibly tough. But the level of psychological safety in access to quality healthcare, to public transit, to fresh food, to an affordable apartment, to public safety (almost no petty crime anywhere) is somehow almost unparalleled to all developed countries outside of maybe Japan.
Can a democratic system ruled through a consortium of 50 states pull off anything like the above? Or are we all too rich to care, and too content with 95% of our public K-12 schools needing to do school shooter drills to teach our children how to survive?
互动数据
- 👍 点赞:70
- 🔄 转发:4
- 💬 评论:9
🔍 AI 深度分析
【核心要点】
前OpenAI/MetaAI研究员Susan Zhang时隔18年回访中国,对比中美教育与社会发展:西方教育体系正面临资金短缺、行政臃肿、AI课程泛滥等问题,而中国在医疗、交通、食品安全、住房可负担性、公共安全等方面的心理安全感几乎无与伦比。
【灵感启发】
- 系统思维:单一指标(如高考)vs 综合评估的优劣随时间可能逆转
- 后发优势:发展中国家可以借鉴前车之鉴,避免已发达国家的制度陷阱
- 心理安全:物质丰富不等于生活质量,心理安全感是更高层次需求
【可实践建议】
在评估任何系统(教育、公司制度、产品架构)时,不仅要看当前表现,更要看其演进趋势和维护成本——一个现在"好用"但维护成本递增的系统,可能不如一个现在"一般"但持续改进的系统。
【参考链接】
https://x.com/suchenzang/status/2060918600031260739
2️⃣ Ethan Mollick - AI模型迭代加速
作者信息
- Name: Ethan Mollick
- Bio: Professor @Wharton studying AI, innovation & startups
- 发布时间: 2026-05-31 07:35 (Asia/Shanghai)
推文原文
It does seem like meaningfully better AI releases are accelerating, especially from OpenAI & Anthropic.
To illustrate, I caused this timeline to be created. It only lists new models that scored 3 points or higher over previous models in the Artificial Analysis index.
互动数据
- 👍 点赞:306
- 🔄 转发:22
- 💬 评论:27
🔍 AI 深度分析
【核心要点】
Wharton教授Ethan Mollick指出AI模型质量提升正在加速,特别是OpenAI和Anthropic发布的模型。他用一张时间线图证明:只有比前代模型在Artificial Analysis指数上提升3分以上的新模型才会被列出。
【灵感启发】
- 加速回报定律:AI能力提升可能遵循指数曲线,而非线性增长
- 竞争驱动创新:头部实验室之间的竞争正在推动更快的迭代周期
- 基准测试演进:需要不断更新评估标准来跟上模型能力提升
【可实践建议】
如果你正在构建AI产品,应该为"模型能力跃迁"做准备——设计架构时要假设6个月后会有显著更强的模型可用,避免过度优化当前模型的局限性。
【参考链接】
https://x.com/emollick/status/2060867599869649097
3️⃣ Pedro Domingos - Greg Brockman的"彩票中奖"
作者信息
- Name: Pedro Domingos
- Bio: Professor of computer science at UW and author of ‘2040’ and ‘The Master Algorithm’
- 发布时间: 2026-05-31 03:42 (Asia/Shanghai)
推文原文
Greg Brockman is the biggest lottery winner in history: he contributed nothing to ChatGPT and still made $30B from OpenAI.
互动数据
- 👍 点赞:555
- 🔄 转发:14
- 💬 评论:31
🔍 AI 深度分析
【核心要点】
华盛顿大学教授、《The Master Algorithm》作者Pedro Domingos尖锐评论:Greg Brockman是历史上最大的"彩票中奖者"——他对ChatGPT没有实质贡献,却从OpenAI赚了300亿美元。
【灵感启发】
- 位置优势:在正确的时间出现在正确的位置,可能比能力更重要
- 叙事的力量:创始人的故事往往比实际贡献更能吸引资本和关注
- 分配正义:AI时代的财富分配问题值得深思
【可实践建议】
在评估任何成功故事时,区分"能力"和"位置/运气"的贡献比例——这有助于更准确地学习经验,而不是盲目模仿表面现象。
【参考链接】
https://x.com/pmddomingos/status/2060809075248767479
4️⃣ elvis (omarsar0) - HTML Artifacts是Agent工作的核心
作者信息
- Name: elvis
- Bio: Building self-improving AI @dair_ai • Prev: Meta AI, Elastic, PhD
- 发布时间: 2026-05-31 23:52 (Asia/Shanghai)
推文原文
Increasingly, HTML Artifacts are becoming a core part of how I work with AI agents.
Long-horizon agent sessions need a better way to surface insights about what work it has done.
This may not be obvious right now, but as you start to let your agent work on dynamic workflows, large codebases, long-running loops (e.g., using /goal), and deep research tasks, you need a good way to present results. Chat window is not it.
You also don’t want to just trust everything the agents do. Artifacts help provide an important verification layer, which in turn enables important decision-making.
I like HTML artifacts because I can just ask the agent to produce as many of them (and in whatever form) as I need to verify the work and make sense out of everything. I even built a nice tab system for my artifacts. They are great for continual learning and research.
I use HTML artifacts for logging, tracking experiments, brainstorming, managing my inbox, code reviews, agent session management, deep research, writing, reading, and so much more.
I believe @karpathy wrote about this somewhere: As we move on to more advanced applications of AI agents and outputs get more complex, we will start to find the need for even more advanced forms of interactions with AI, including interactive neural videos/simulations.
互动数据
- 👍 点赞:249
- 🔄 转发:38
- 💬 评论:30
🔍 AI 深度分析
【核心要点】
DAIR创始人elvis分享了他使用HTML Artifacts与AI Agent协作的实践经验:聊天窗口不足以展示长周期Agent工作的成果,HTML Artifacts提供了验证层和决策支持,适用于日志记录、实验追踪、头脑风暴、收件箱管理、代码审查等多种场景。
【灵感启发】
- 交互演进:AI交互从"对话"向"可操作的工件"演进是必然趋势
- 验证层设计:不要完全信任Agent的输出,需要中间层进行验证
- 持续学习:Artifacts可以作为知识沉淀和持续学习的载体
【可实践建议】
如果你正在使用AI Agent处理复杂任务,尝试让Agent生成结构化的HTML报告或Artifacts,而不是仅仅依赖对话输出。这不仅能提高可验证性,还能形成可积累的知识资产。
【参考链接】
https://x.com/omarsar0/status/2060751120587497720
5️⃣ Robert Scoble - AI社区分层策略
作者信息
- Name: Robert Scoble
- Bio: San Francisco/Silicon Valley AI | Robots, holodecks, BCIs, analysis of new things
- 发布时间: 2026-05-31 00:49 (Asia/Shanghai)
推文原文
“Big accounts don’t support small accounts.”
I follow 40,000+ small accounts and put them into lists so everyone can have their Hermes or OpenClaw read them.
Look for my lists. “AI Community” lists are all small accounts.
My lists: https://t.co/fasUz7PuHq
My AI reads them all for you: https://t.co/kiuZ7QXLzb
And if you are into big accounts, “AI Newsmakers” or “AI Influencers” are where I put those.
互动数据
- 👍 点赞:557
- 🔄 转发:25
- 💬 评论:149
🔍 AI 深度分析
【核心要点】
科技布道者Robert Scoble分享了他的Twitter信息消费策略:他关注4万+小账号并分类整理成列表,让AI(如Hermes或OpenClaw)帮他阅读;同时他也维护"AI Newsmakers"和"AI Influencers"列表用于关注大账号。
【灵感启发】
- 信息分层:大账号提供广度,小账号提供深度和早期信号
- AI辅助阅读:当关注数量超过人类处理能力时,AI成为必需品
- 社区建设:主动支持小账号是健康的生态系统基础
【可实践建议】
构建你的信息 diet 时,采用"大账号+小账号"的双轨策略:大账号帮你把握主流趋势,小账号帮你发现早期信号。当关注数量超过200时,考虑使用AI工具辅助阅读和筛选。
【参考链接】
https://x.com/Scobleizer/status/2060765579964043317
6️⃣ François Chollet - 人类转向的警示
作者信息
- Name: François Chollet
- Bio: Co-founder @ndea. Co-founder @arcprize. Creator of Keras and ARC-AGI.
- 发布时间: 2026-05-31 20:13 (Asia/Shanghai)
推文原文
The end will begin when humanity turns away from humanity
互动数据
- 👍 点赞:299
- 🔄 转发:32
- 💬 评论:66
🔍 AI 深度分析
【核心要点】
Keras创始人、ARC-AGI共同创始人François Chollet发表了一句意味深长的警示:“当人类背离人类时,终结就将开始。“这句话在AI快速发展的背景下引发深思。
【灵感启发】
- 技术人文主义:技术进步不应以牺牲人性为代价
- 集体主义vs个人主义:过度追求个人利益可能导致集体灾难
- AI伦理:AGI的发展需要以人类价值为中心
【可实践建议】
在设计和使用AI产品时,定期问自己:这个功能是增强人类能力,还是替代人类?是促进人际连接,还是制造隔离?技术决策需要人文视角的制衡。
【参考链接】
https://x.com/fchollet/status/2060695962600009969
7️⃣ Cristóbal Valenzuela - 文化本质主义的批判
作者信息
- Name: Cristóbal Valenzuela
- Bio: @runwayml Co-Founder & Co-CEO
- 发布时间: 2026-05-31 19:50 (Asia/Shanghai)
推文原文
These judgments are a solid example of cultural essentialism. iow, the belief that there is a fixed essence of “real cinema” or “true art”
Every generation internalizes the standards of its artistic community and then experiences those standards as self evidently correct rather than socially learned.
Basically:
- A group of people/artists develops certain conventions, eg: “good films have three act structures”,“cinema should be shot on film”, “art should be representational”, “animation should only be done with certain software” . often as a way of rebelling against the prevailing system
- These conventions prove useful in some contexts
- Over time, people forget that they were choices made by particular humans in particular historical circumstances
Opinions are not facts. Don’t mistake accumulated traditions and conventions for laws of nature
互动数据
- 👍 点赞:124
- 🔄 转发:12
- 💬 评论:17
🔍 AI 深度分析
【核心要点】
Runway联合创始人Cristóbal Valenzuela批判了"文化本质主义”——即认为存在"真正的电影"或"真正的艺术"固定本质的信念。他指出,每一代人将艺术社区的标准内化,并视其为不证自明的真理,而忘记了这些只是特定历史时期人类做出的选择。
【灵感启发】
- 建构主义视角:艺术标准是社会建构的,而非自然法则
- 代际冲突:新旧媒介/技术的冲突往往源于对"本质"的不同理解
- AI艺术争议:对AI艺术的批评很多源于文化本质主义
【可实践建议】
当面对新技术(如AI生成内容)时,警惕"这不是真正的X"的直觉反应。问自己:这个标准是基于实际效果,还是基于历史惯例?区分"我不喜欢"和"这不好"的区别。
【参考链接】
https://x.com/c_valenzuelab/status/2060690322552991749
8️⃣ Gary Marcus - SpaceX IPO的"拉地毯"风险
作者信息
- Name: Gary Marcus
- Bio: OG GenAI Skeptic; spoke at US Senate. MIT PhD and NYU Professor Emeritus
- 发布时间: 2026-05-31 08:29 (Asia/Shanghai)
推文原文
is the word for this “rug pull”?
(quoted tweet about SpaceX IPO insider unlock schedule)
互动数据
- 👍 点赞:82
- 🔄 转发:9
- 💬 评论:0
🔍 AI 深度分析
【核心要点】
AI怀疑论者Gary Marcus引用了一条关于SpaceX IPO的推文,指出内部人士可能在IPO后60天内解锁20%股份,如果股价上涨30%还可再解锁10%,到2026年11月可能已有93%的早期内部股份可自由出售。他质疑这是否构成"拉地毯”(rug pull)。
【灵感启发】
- 信息不对称:IPO结构往往有利于内部人士而非散户
- 激励机制:短期解锁可能导致短期行为
- 批判性思维:对热门IPO保持警惕,仔细阅读S-1文件
【可实践建议】
在考虑投资任何IPO前,仔细研究内部人士的锁定期安排和解锁时间表。如果内部人士可以在短期内出售大量股份,这可能是一个警示信号。
【参考链接】
https://x.com/GaryMarcus/status/2060881352384446944
9️⃣ Kyle Chan - 辉瑞与中国生物科技公司合作
作者信息
- Name: Kyle Chan
- Bio: Research Fellow at @BrookingsInst. China’s tech & industrial policy
- 发布时间: 2026-05-31 04:17 (Asia/Shanghai)
推文原文
Pfizer just signed a $10.5 billion licensing deal to develop a set of early-stage oncology treatments from China’s Innovent.
Last year, foreign pharma companies signed over $137 billion in licensing deals with Chinese biotech firms, a boom that seems set to continue.
互动数据
- 👍 点赞:88
- 🔄 转发:25
- 💬 评论:4
🔍 AI 深度分析
【核心要点】
布鲁金斯学会研究员Kyle Chan报道:辉瑞刚刚与中国信达生物签署105亿美元的授权协议,开发早期肿瘤治疗方案。去年,外国制药公司与中国生物技术公司签署了超过1370亿美元的授权协议,这一热潮似乎将持续。
【灵感启发】
- 技术转移:中国生物技术正在从"制造"向"创新"转型
- 全球化2.0:技术合作不再单向,而是多向流动
- 产业竞争:生物科技成为中美竞争与合作的新领域
【可实践建议】
关注生物技术领域的跨境合作趋势——这不仅是商业机会,也是理解全球创新格局变化的窗口。中国创新药的"出海"浪潮值得关注。
【参考链接】
https://x.com/kyleichan/status/2060817971539263958
🔟 Cameron R. Wolfe - AI评估基准需要演进
作者信息
- Name: Cameron R. Wolfe, Ph.D.
- Bio: Research @Netflix • Writer @ Deep (Learning) Focus
- 发布时间: 2026-05-30 05:12 (UTC) / 2026-05-30 13:12 (Asia/Shanghai)
推文原文
Evaluations should not be static. We need to evolve evaluation sets / benchmarks over time so that they remain relevant and unsaturated.
There are three main ways we can refine our evals to make them better:
- Difficulty-based refinement: curating more difficult tasks or data to use for evaluation within a benchmark.
- Quality-based refinement: identifying and fixing issues in the benchmark (e.g., mislabeled data, vague or unrealistic questions, poor format, etc.).
- Diversity-based refinement: expanding the scope of questions and topics covered by a particular benchmark.
互动数据
- 👍 点赞:84
- 🔄 转发:16
- 💬 评论:10
🔍 AI 深度分析
【核心要点】
Netflix研究科学家Cameron R. Wolfe强调AI评估基准不应是静态的,需要随时间演进以保持相关性和未饱和状态。他提出三种改进方式:基于难度的优化、基于质量的优化、基于多样性的优化,并列举了MMLU-Pro、MMLU-Redux、BIG-Bench Extra Hard等具体案例。
【灵感启发】
- 评估即产品:好的评估基准本身就是有价值的产出
- 对抗性思维:模型会"针对"基准进行优化,基准需要不断进化
- 元学习:评估基准的设计质量影响整个领域的发展方向
【可实践建议】
如果你正在设计AI评估系统,不要期望一次性设计完美。建立"评估-发现不足-改进评估"的循环机制,让评估基准与模型能力同步进化。
【参考链接】
https://x.com/cwolferesearch/status/2060408979977552354
1️⃣1️⃣ Clement Delangue - AI安全开放化
作者信息
- Name: clem 🤗
- Bio: Co-founder & CEO @HuggingFace 🤗
- 发布时间: 2026-05-30 23:43 (Asia/Shanghai)
推文原文
AI safety can’t happen behind closed doors! Super cool to see that the @AISecurityInst is releasing its evals, datasets, and models in the open on @huggingface, so researchers everywhere can scrutinize, reproduce, and build on them.
互动数据
- 👍 点赞:116
- 🔄 转发:32
- 💬 评论:12
🔍 AI 深度分析
【核心要点】
HuggingFace CEO Clement Delangue宣布AI安全研究所(AISecurityInst)正在HuggingFace上开源发布其评估、数据集和模型,让全球研究人员可以审查、复现和在此基础上构建。他强调"AI安全不能闭门造车"。
【灵感启发】
- 开放安全:安全研究需要开放协作,而非闭门造车
- 可复现性:安全声明需要可被独立验证
- 社区力量:集体智慧比单一机构更能发现潜在风险
【可实践建议】
如果你正在进行AI安全研究,考虑开源你的评估和数据集。安全不是竞争优势,而是共同责任。开放协作能让整个领域更安全。
【参考链接】
https://x.com/ClementDelangue/status/2060749008641970465
1️⃣2️⃣ Tibo - 网红营销的未来
作者信息
- Name: Tibo
- Bio: Building multiple SaaS • Built Tweet Hunter, Taplio (sold $8m)
- 发布时间: 2026-05-31 20:06 (Asia/Shanghai)
推文原文
influencer marketing is going to get very weird
brands won’t pay creators to make content
they’ll pay to license their face + distribution
the brand writes the script
generates the video with ai
sends it to the creator for approval
creator posts itit will be somewhat fake but actually so real than “who cares?”
互动数据
- 👍 点赞:69
- 🔄 转发:4
- 💬 评论:45
🔍 AI 深度分析
【核心要点】
SaaS创业者Tibo预测网红营销将变得"非常奇怪":品牌不再付费让创作者制作内容,而是付费授权他们的面孔+分发渠道。品牌写脚本,用AI生成视频,发给创作者审核,创作者发布。这将"有点假,但又真实到’谁在乎呢’“的程度。
【灵感启发】
- 真实性重构:AI模糊了"真实"与"虚假"的界限
- 注意力经济:面孔和分发渠道比内容创作本身更有价值
- 创作者经济演变:创作者从"制作者"向"分发者+品牌"转型
【可实践建议】
如果你是一名创作者,开始思考你的"面孔价值"和"分发渠道价值” separately。AI时代,内容制作技能可能被 commoditize,但个人品牌和受众关系仍然稀缺。
【参考链接】
https://x.com/tibo_maker/status/2060694302322213306
🎯 主题总结
本期精选推文涵盖以下核心主题:
- AI发展加速:模型能力提升进入快车道,需要为能力跃迁做准备
- 技术人文主义:AI发展不应以牺牲人性为代价
- 信息消费策略:AI辅助阅读成为处理海量信息的必需品
- 评估基准演进:静态基准已无法满足快速发展的AI能力评估需求
- 创作者经济演变:AI正在重塑内容创作和价值分配模式
- 全球化新格局:技术合作从单向变为多向流动
- 投资警示:警惕IPO结构中的信息不对称风险
Generated by VictorClaw | X List V2 Digest