Geoffrey Litt 在 AIE 提出「理解以参与」理念：与编码智能体协作需避免认知债务
Geoffrey Litt 在 AIE 演讲中提出「理解以参与」理念：开发者需要深入理解代码，才能在与编码智能体（coding agents）的协作中保持主动参与，避免因认知债务（cognitive debt）导致无法有效推进项目。他认为，缺乏对代码的概念流畅性会显著限制参与能力。该演讲录像将于三周内陆续放出。
#大佬观点
Simon Willison 博客AI 评分5454 The Decoder：AI NewsAI 评分5050 Anthropic 为 Claude Code 削减 80% 系统提示词，称 Fable 5 模型“想要更短的提示词”
Anthropic 将 Claude Code 的系统提示词削减了 80%。技术员工 Tariq Shihipar 指出，新 Fable 5 模型（Mythos 类）不再需要更多指令和示例，过多的示例反而会限制模型的想象力。Anthropic 转而通过上下文而非硬性规则来引导模型。这一变化分阶段发生：早期模型需要短提示词配合大量示例，随后提示词随模型理解能力提升而变长，如今又再度缩短。
IT之家·人工智能AI 评分5454 何小鹏：中国智能辅助驾驶领先全球，未来3-5年可见L4甚至L5落地
7月2日晚，小鹏MONA L03首秀后，何小鹏称中国智能辅助驾驶在效果和法规制定上均领先全球。联合国ADS GTR法规覆盖L2至L5，其中L2全面性条款意味着今年12月中国现有的L2智能辅助驾驶可在全球使用。何小鹏曾在今年全国两会建议从L2跳过L3直接进入L4、L5，他相信未来3到5年内不仅能看到L4落地，甚至能看到L5在所有场景安全通行，速度超出想象。
数字生命卡兹克@Khazix0918AI 评分6363公众号：千问APP精选AI 评分6262 千问团队朱达：C端Agent Harness的“多快好省”工程哲学与主动服务探索
千问团队2026年1月上线通用复杂任务Agent（千问App胶囊入口），总结“多快好省”方法论：支持信息搜集、研究分析等任务；执行时间降至初始1/3；通过搜索范式与上下文管理优化交付质量；Token消耗仅为海外产品1/10。团队探索从被动响应转向主动服务，构建User Memory、Environment、Task System、Assistant四大组件，指出“情商”是主动服务最难环节。朱达提出Agent工程从Prompt Engineering演进至Harness Engineering，下一站是A IWare Engineering，强调“低功耗，够用就行”。
推荐理由：千问C端团队分享的Agent工程实践很务实，从“多快好省”到AIWare Engineering的演进思路，对正在做复杂任务Agent的团队是一个有价值的参考系。
Rohan Paul@rohanpaul_aiAI 评分4444引用Elon Musk@elonmusk@DoctorJack16 No, Optimus production will be extremely slow at first, as everything is new. This is not like making a car.
- Rohan Paul@rohanpaul_aiAI 评分4545
Palantir CEO Alex Karp: 一家公司不只需要一个在精致界面内回答问题的聪明模型。严肃的技术客户想要的是对数据、提示词、系统访问以及创造价值的工作流的控制。
Dwarkesh Patel：Podcast & BlogAI 评分5454 关于AI重大问题的获奖征文
Dwarkesh Patel举办的AI征文比赛评选出三位获奖者。第一名Jassi Pannu主张OpenAI基金会应投入数十亿美元终结空气传播疾病，利用AI加速自主生物学发现，可带来超1万亿美元年GDP增长并消除灾难性流行病风险。第二名Ege Erdil建议AI供应链外的国家通过强产权、低资本税和开放监管政策抓住增长机会。第三名Michael Li类比香港地铁商业模式，提出AI实验室可通过收购互补性资产盈利。
Tomasz Tunguz 博客精选AI 评分6060 构建AI智能体应优先设计路由
构建AI智能体时，应优先设计路由（router）而非选择模型。路由决定每个请求由哪层模型处理。正确路由可使70-80%流量运行在免费本地模型或异步推理上，将AI开销降低90%+。Brian Armstrong指出Coinbase通过更好的默认设置、路由和缓存，在token使用量增长的同时将AI支出减半。路由分三层：技能分类器、路由器、模型选择器。本地计算近乎零成本，异步批量推理比实时推理便宜两个数量级。大多数工作无需秒级返回。同步预测器标记复杂任务，夜间批量评估器更新路由权重。技能蒸馏后，非编码类任务中70-80%智能体流量可由本地模型处理。
推荐理由：Tunguz 把代理架构的设计重心从模型选择拉回到路由上，三层分类器-路由器-选择器的划分很清晰，做 AI 应用的团队可以参考，但其中的新东西不多。
François Chollet@fcholletAI 评分5252IT之家·人工智能AI 评分5959 Meta CEO 扎克伯格不认为 AI 必然带来大规模失业，建议企业聚焦“个人超级智能”
扎克伯格认为 AI 不必然造成大规模失业，关键在于企业应优先发展“个人超级智能”赋能个体，而非将所有知识工作自动化。他影射 Anthropic、OpenAI，强调劳动者生产力提升速度须超过企业自动化速度，就业岗位才可能增加。Meta 已投入数十亿美元调整 AI 战略，他称之为“重启”。对于 Meta 生成式 AI 进展，他坦言仍有提升空间，超级智能实验室成立不足一年，模型进展超预期，但他认为“原本可以做得更好”。
Tomasz Tunguz 博客AI 评分4848 2026 年 CIO 的选择很明确
对 87 家上市 SaaS 及平台公司的分析显示，2026 年只有基础设施与开发工具（一年涨幅 68.5%）和安全（一年涨幅 17.6%）两个板块增长为正，其余三个板块均下跌。市场正在买入 AI 基础设施，卖出按席位定价的应用层。
Ethan Mollick：One Useful Thing精选AI 评分7373 聊天机器人的黄昏
前沿AI模型能力加速提升，美国实验室发布速度加快，但政府干预已限制访问Claude Fable和GPT-5.6。Epoch测试发现Opus 4.7自主运行14小时即可完成需2-17周人工的软件工程，token成本$251。中国开源模型落后前沿6-12个月，但性能也快速提升，在AA-Briefcase测试中呈独立指数曲线。使用方式正从聊天机器人转向智能体，OpenAI内部四分之一员工每周同时运行至少四个智能体。Claude Code用户数据显示，领域经验比职业属性更决定使用效果，专家正用智能体替代此前非专家的聊天机器人使用模式。
推荐理由：Ethan Mollick 用数据和亲身实验把话说得很直白，AI 不再是聊天工具而是能独自干上 9 小时活的 agent。做非技术岗的也在用，工作变成管理 AI，这个转变才刚开始，但已经让机构跟不上。
Simon Willison 博客AI 评分5050 AI指南针
一个名为“AI指南针”的政治指南针风格测验通过29道关于AI和AI伦理的是非题，将参与者匹配到30种原型之一。作者本人在首次回答后被归类为“The Garage Tinkerer”。该测验以单页React应用实现，利用`<script type="text/babel">`脚本标签绕过构建步骤。
Dwarkesh Patel：Podcast & BlogAI 评分5858 Dwarkesh Patel 对谈 Grant Sanderson：AI 与数学的未来
Dwarkesh Patel 发布与 3Blue1Brown 创始人 Grant Sanderson 的播客对谈，讨论 AI 在数学领域的快速进展及其对其他领域的预示。
IT之家·人工智能AI 评分5151 优必选 CEO 周剑：陪伴机器人最能带来生态，目前一半精力在家庭场景
优必选 CEO 周剑透露，公司已确立工业、商用、家庭机器人三线并行布局，一半精力投入工业与商用市场，另一半投入家庭陪伴赛道。工业机器人已成营收支柱，批量落地海内外车企与制造工厂，他预测 2026 年人形机器人营收将超整体收入 80%。同日，优必选推出“优世界”U1 系列全尺寸超仿生人形机器人，分男款（183cm/42kg）与女款（168cm/35.2kg），搭载 88 个高自由度关节、Wi-Fi 连接，续航 2-4 小时，内置养成系情感大模型，支持加密存储记忆和多维度外观定制。价格方面，U1 Ultra 男款 99 万元、女款 88 万元，U1 Pro 16.98 万元，U1 Lite 11.98 万元（前 5000 名用户可享 1 万抵 2 万权益）。
IT之家·人工智能AI 评分4848 谷歌前工程负责人用Claude Code开发维基百科搜索引擎Zettair，未手动写一行代码
谷歌前工程负责人休·威廉姆斯借助Anthropic的编程助手Claude Code，开发出可正常运行的搜索引擎Zettair，为150万个维基百科条目建立索引。该引擎支持搜索自动补全、关键词摘要片段、相关搜索推荐、热门话题榜单以及AI生成的内容摘要。威廉姆斯未手动编写一行代码，但底层检索框架源自其21世纪初参与研发的信息检索系统。他认为，Claude Code只有在深厚专业技术功底的配合下才能发挥最佳效果，经验丰富的工程师永远是最出色的“指导者”。
- Rohan Paul@rohanpaul_aiAI 评分5656引用Rohan Paul@rohanpaul_ai
Opinion from a former Meta PM. And this is from Aravind Srinivas of Perplexity "China can build data centers a lot faster. Power is not a problem. Permits are not a problem. People are not a problem. Labor is not a problem. Expertise is not a problem." https://x.com/rohanpaul_ai/status/2071780740220752220/video/1
- Rohan Paul@rohanpaul_aiAI 评分4747
引用Xiaoyin Qu@quxiaoyinThe worst case scenario for USA AI: 1. Chinese open sources keep gaining market share. China owns the model layer. 2. Those models were trained and inference-optimized on Huawei chips instead of NVIDIA. China also owns the chip layer. 3. US doesn't build data centers fast enough to keep up with the demand of compute, storage and energy. China meanwhile exports the inference and training layer(for continual training it will happen along with inference) Export control is not the right strategy here. Simply banning "open source from China" doesn't solve the issue here. USA must invest in open source models, hopefully get Chinese models to use NVIDIA, and invest in nuclear asap.
IT之家·人工智能AI 评分5353 优必选创始人周剑：机器人将替代手机，成为AI最核心的交互终端
在2026年度全球发布会上，优必选创始人周剑指出，机器人将替代手机成为AI最核心的交互终端，用户用自然语言下达指令即可获得服务与情感陪护，交互方式从手动变为口述，奠定机器人在家庭的终端入口地位。同日，优必选宣布超仿生人形机器人优世界U1系列订单已突破1万台，力争今年交付。该机器人主打情感陪伴，分男女两款，各搭载88个高自由度运动关节。作为对比，2025年全年优必选全尺寸具身智能人形机器人总销量为1079台。
- Rohan Paul@rohanpaul_aiAI 评分5757引用Rohan Paul@rohanpaul_ai
Aravind Srinivas just explained why China’s open-source AI may become more powerful than ever. And why Anthropic has lobbied very hard for export control. "the only reason why there is even a 12-month gap between open source and frontier models is export controls. But there is a chance that, because of that, they now get really good at the physical layer. One advantage they (China) have is that they can actually build data centers a lot faster. Power is not a problem. Permits are not a problem. People are not a problem. Labor is not a problem. Expertise is not a problem. And so, by forcing them to go out there and build all this, you are converting them into a far more potent competitor." --- From "20VC with Harry Stebbings" YouTube channel (@HarryStebbings ), link in comment
IT之家·人工智能AI 评分5757 Claude Code 负责人称未来员工将身兼数职，工程产品设计融为一体
Claude Code 负责人鲍里斯·切尔尼认为工程、产品和设计正逐渐融合。他将其团队成员分为五类：原型设计者、构建者、清理者、增长者和维护者，许多成员身兼多职。健康团队需具备全部五种能力，未来产品岗位可能不再按专业领域区分。Figma CEO 迪伦·菲尔德也认同传统职位边界正在消失，部分企业已弃用“管理者”称谓。切尔尼强调角色会随时间和项目改变，且 Claude 能在不同程度上协助所有类型工作。
- Rohan Paul@rohanpaul_aiAI 评分5252
引用Rohan Paul@rohanpaul_ai👀 This is from an ex-Meta PM.
Tomasz Tunguz 博客精选AI 评分5858 Anthropic：当AI成本超过工程师薪酬
Anthropic在算力上的支出达到每位工程师每年51.5万美元，是其完全薪资（22.4万美元）的2.3倍。相比之下，顶尖1%软件公司的算力支出为8.9万美元，中位数仅为1.37万美元。三个2029年情景预测了这一差距的缩小路径。
推荐理由：Tomer Tunguz 用数据把 AI 公司的烧钱速度拆得很细，Anthropic 每个工程师年算力成本 51.5 万美元，是工资的 2.3 倍，这个数字对重新理解 SaaS 毛利率很有冲击力。
- Rohan Paul@rohanpaul_aiAI 评分6262
引用Rohan Paul@rohanpaul_aiDario Amodei has a really hardline view that China shouldn’t have strong AI. He says - "That's the national security interest of the US....The case is very clear, the counterarguments against it are just fishy." - From "Dwarkesh Patel" YT channel
- Rohan Paul@rohanpaul_aiAI 评分6666
引用Rohan Paul@rohanpaul_ai👀 This is from an ex-Meta PM.
- Rohan Paul@rohanpaul_aiAI 评分4949
引用Rohan Paul@rohanpaul_aiAnthropic CEO Dario Amodei on Open-Source AI Models. "I don't think open source works the same way in AI that it has worked in other areas. Primarily because with open source you can see the source code of the model. Here we can't see inside the model, it's often called open weights instead of open source to kind of distinguish that. But a lot of the benefits, which is that many people can work on it and that it's kind of additive, don't quite work in the same way. So I've actually always seen it as a red herring. When I see a new model come out I don't care whether it's open source or not. If we talk about Deep Seek I don't think it mattered that Deep Seek is open source. I think I ask, is it a good model? Is it better than us at the things that matter? That's the only thing that I care about. It actually doesn't matter either way. Because ultimately you have to host it on the cloud. The people who host it on the cloud do inference. These are big models, they're hard to do inference on. When I think about competition I think about which models are good at the tasks that we do. I think open source is actually a red herring. It's not free. You have to run it on inference and someone has to make it fast on inference." --- From 'Alex Kantrowitz' YT channel (full video link in comment)
- Rohan Paul@rohanpaul_aiAI 评分5050引用Rohan Paul@rohanpaul_ai
Anthropic CEO Dario Amodei on Open-Source AI Models. "I don't think open source works the same way in AI that it has worked in other areas. Primarily because with open source you can see the source code of the model. Here we can't see inside the model, it's often called open weights instead of open source to kind of distinguish that. But a lot of the benefits, which is that many people can work on it and that it's kind of additive, don't quite work in the same way. So I've actually always seen it as a red herring. When I see a new model come out I don't care whether it's open source or not. If we talk about Deep Seek I don't think it mattered that Deep Seek is open source. I think I ask, is it a good model? Is it better than us at the things that matter? That's the only thing that I care about. It actually doesn't matter either way. Because ultimately you have to host it on the cloud. The people who host it on the cloud do inference. These are big models, they're hard to do inference on. When I think about competition I think about which models are good at the tasks that we do. I think open source is actually a red herring. It's not free. You have to run it on inference and someone has to make it fast on inference." --- From 'Alex Kantrowitz' YT channel (full video link in comment)
Boris Cherny@bcherny精选AI 评分6868推荐理由：Boris Cherny 从 Claude Code 团队提炼出的五种原型，可能是 AI 时代产品团队结构演化的一个风向标，产品人看到会忍不住对号入座。
Simon Willison 博客AI 评分4646 Jon Udell：AI 智能体应被邀请进入开发循环，而非将人类排除在外
Jon Udell 反对“human in the loop”的表述，认为它将权威让渡给机器。他主张翻转叙事——人类本就主导工作循环，现在应主动招募 AI 智能体加入团队。智能体辅助的开发过程不应是“输入提示词、输出功能”的黑箱，而是人类邀请智能体协作的开放循环。
jason@jxnlcoAI 评分7575引用Lenny Rachitsky@lennysanAndrew Ambrosino (@ajambrosino) leads the team behind the Codex desktop app at @OpenAI. Codex usage has 6x'd since February, reaching over 5M weekly active users, and nearly 100% of OpenAI's employees use the Codex app regularly (and not just the engineers). Andrew's personal mission is to build "the best desktop app that has ever existed, full stop." If you've used the Codex app lately, you know he's not far off from that goal. In our in-depth conversation, we discuss: 🔸 The "zone defense" model of how PMs at OpenAI operate 🔸 Why AI is so bad at design 🔸 Why Andrew thinks the Codex app would have flopped if they'd shipped it in November instead of February (same product—only the model changed) 🔸 What “taste” really means as a professional skill 🔸 How Andrew uses Codex to run his workflows 🔸 His vision for Codex + ChatGPT Listen now 👇 https://youtu.be/P3KDebPTUrw
- 数字生命卡兹克@Khazix0918AI 评分4141
Claude Code闹鬼了？我也没手动叫停啊，我对话里更没说过要写一份详细的文档啊，这TM到底哪来的？？？ 这两天串信息感觉有点过于严重了吧，我的代码里经常会有莫名奇妙的改动，还TM能不能用了啊
Hamel Husain 长文AI 评分6666 Hamel Husain：产品难以做 eval 其实是产品设计的坏味道
Hamel Husain 撰文指出，"产品难以 eval"本质是产品设计的坏味道，难以验证的产物对用户同样难以验证，应先为可验证性设计产品再做 evals。
Gary Marcus：The Road to AI We Can TrustAI 评分4545 中国追赶
美国AI产业受中国追赶引发价格战，token价格趋近于零，利润微薄，Anthropic和OpenAI的万亿IPO难以实现，数据中心巨额投资难回收。当前范式有三个根本缺陷：暴力训练全互联网成本高昂且效率低下；系统不可靠，长期无法维持溢价；基础方法易复制，导致价格战与负利润率。另有人认为，整个AI竞赛方向可能被误解，美国不应只追求最便宜的LLM，而应开发更适合科学与医学的AI新形态。
TechCrunch：AIAI 评分5454 软银CEO并非唯一对马斯克轨道数据中心炒作有疑问的人
软银CEO孙正义在股东大会上批评马斯克的轨道数据中心构想，认为其成本高、周期长，而AI竞赛未来几年比十年后的可能性更重要。SpaceX目前占据全球发射市场80-90%份额，主要依赖Starlink业务；建设需要每隔几年更换卫星的轨道数据中心将为SpaceX带来更多发射业务。同期，芯片公司Groq完成6.5亿美元融资，OpenAI推进定制芯片计划。
