Shopify CEO Tobi Lütke 宣布开源内部 Agent 自我改进闭环基础设施组件 Tangle。同时,Elon Musk 透露 X 平台 @Bot 正在引入自动识别待解决问题、完成代码修复并异步推进流程的能力,Anthropic 团队亦展示了 Claude Tag 跨 Slack 与表格自主生成报告闭环。
采用 Ben Thompson 的 Aggregation Theory(聚合理论)视角看,当底层基座大模型的推理能力趋于平民化,商业价值正在快速向上层的控制环路(Harness)转移。过去大家以为 Agent 的核心是 Prompt 技巧,但 Shopify 与 xAI 的动作表明,真正决定商业可用性的是系统级自我纠错与长流程异步执行闭环。段永平常强调「本分就是做对的事,把事做对」,在软件工程中,把状态持久化、搭建可自主校验的容错机制,才是做对的事。我认为,模型调用本身已不再是壁垒,谁能搭建起承载长任务、具备自愈能力且不依赖人工介入的 Harness,谁就掌握了 AI 应用层的核心杠杆。
前抖音负责人卷卷在《十字路口》播客中分享 AI 3D 创业经验,指出基础大模型无法完全吞噬物理制造与供应链交付,数美万物选择从 3D 大模型 Hi3D 延伸至线下物理工厂与拆件摆盘,试图打造面向实体制造业的操作系统。
运用芒格的多学科思维模型(物理世界摩擦力)与段永平的「能力圈/本分」原则来审视:纯软件生成代码毫无边际成本,但一旦延伸至物理实体交付,原子世界的材料、供应链与履约将产生巨大的交易成本。马斯克在《马斯克传》中一再强调「制造工厂比原型设计难上万倍」。我认为,生成式 AI 的核心杠杆在于「信息场的重组」,而非「原子场的履约」。基础大模型厂商可以通过算力与数据吞噬纯数字内容生成,但绝对无法单靠 API 替代扎根在供应链深处的物理执行管线。
🎯 商业机会雷达
缺口物理实体与数字 AI 之间存在巨大的适配断层,但线下供应链智能化需要极高的资金积累与重交付团队。
观点头部厂商正以高密度推动轻量级与垂直领域模型迭代。对 CEO 而言,这预示着推理成本与安全护栏能力的进一步提升,应及时评估产品架构的模型选型。
Introducing Gemini 3.8 Flash, another upgrade in under a month! And the new 3.8 Flash Cyber pushes the frontier of cyber defense. Relentless progress ...
Had early access to Gemini 3.8 Flash, it is a very good Flash model, but not equivalent to a frontier model, though. Here is Gemini 3.8 Flash's versio...
观点头部 SaaS 厂商正优先获得下一代模型早测权,新模型在复杂企业场景的可用性持续增强。作为 CEO 应建立自有的企业场景 Eval 评估集,以便新模型发布时快速验证与调优升级。
At Box, we’ve been testing Fable 5.1 in early release against our complex enterprise work eval. Fable 5.1 delivers a huge 7 percentage point jump ove...
鲍里斯·切尔尼@bcherny ✓· 2026-09-03 06:22
7.4
鲍里斯·切尔尼
@bcherny ✓ · Boris Cherny · 424K followers
Anthropic Claude Code 产品负责人,主导这款 AI 编程工具的产品决策;推文常分享 Claude Code 的使用技巧、产品思路和更新预告,是了解第一手产品方向的信源
Fable 5.1 让 Claude Tag 变得更加实用。在这里,它利用指标电子表格以及 Slack 等平台的数据,快速生成了一份供领导层使用的汇报 PPT。
事实Anthropic 团队展示了 Claude Tag 结合 Fable 跨应用(Slack、表格)获取 Context 的能力。
观点自动化办公 Agent 正在从单一终端交互走向跨系统数据整合与最终交付物生成。
对你意味着应关注企业级工作流中多源数据连接器的沉淀。
Fable 5.1 makes Claude Tag even more useful. Here it builds a last-minute leadership deck from a metrics spreadsheet and other data across Slack, spot...
重大提升。AI/ML API:Muse Spark 1.3 对比 1.2:雕刻维京手办。@AIatMeta 今天发布了 Muse Spark 1.3 —— 我们将其与 Muse Spark 1.2 进行了对比测试。两个模型收到了相同的需求:在一个自包含的 HTML 中生成三个可收集的 3D 手办(维京头盔、镶钻战斧、带船员的长船)。
事实Meta 发布 Muse Spark 1.3,Scale AI 团队对其自包含 3D HTML 渲染能力进行了基准对比。
观点前端代码与 3D 交互一体化生成能力正在加速成熟,CEO 应关注此类多模态模型对复杂 UI 与 3D 原型开发效率的提效潜力。
big improvementsAI/ML API: Muse Spark 1.3 vs 1.2: sculpt viking figurines @AIatMeta dropped Muse Spark 1.3 today — we ran it against Muse Spark 1.2both models got the same brief: three collectible 3D figurines — a viking helmet, a diamond-studded axe, a longship with a crew — one self-contained HTML
Had early access to Claude Fable 5.1. Its a real advance in long-run work that requires judgement and taste, but less of an advance in the Claudish. H...
观点研发工具链对最新顶尖模型的敏捷接入正成为效率分水岭,CEO 需关注内部 AI 编程工具链的升级以保持研发效能优势。
Claude Fable 5.1 is now available in Cursor! It's the most capable model we’ve run on CursorBench 3.2, scoring 73.4% at max effort. We found it espec...
Moderna CEO Stéphane Bancel 就个性化 mRNA 癌症疫苗史上首次获得阳性三期临床结果发表看法:“在为个性化癌症疫苗努力工作了 10 年之后……”
事实Moderna 历时 10 年研发的个性化 mRNA 癌症疫苗首次取得三期临床阳性结果。
观点计算生物与精准医疗正在迎来商业化拐点。对于 AI 创业者而言,AI 与硬科技领域的落地虽然周期长,但一旦跨越临床与工程验证,将构建极深的数据与技术护城河。
Moderna CEO Stéphane Bancel on the first-ever positive Phase 3 results for an individualized mRNA cancer vaccine: "After working 10 years on an indiv...
很高兴能与 React、开源和 Web 的推动者 @tannerlinsley 及其团队达成合作并提供支持。这对我们的客户意味着:无论你...
事实Vercel 正在通过战略合作深度绑定 TanStack 等头部开源项目团队。
观点前端基础设施服务商正加速向底层状态与数据流框架延伸。
对你意味着在选型 AI Web 应用的技术栈时,主流框架与云厂商的绑定将更为紧密,需注意架构解耦与锁定的平衡。
Excited to partner with and support @tannerlinsley and team, champion of React, open source, and the web What this means for our customers: whether yo...
最高可自动节省 65% 的成本。这就是 Auto Mode 的魅力所在。Replit 工程主管 Rachael Fuller 讲解了智能模型路由(Intelligent Model Routing)的运行机制...
言外之意,
事实Replit 在 Auto Mode 中引入智能模型路由以降低成本。
观点混合模型调度与动态路由正成为 AI 应用层的标配架构。CEO 在搭建产品时应重视路由层建设,通过将不同复杂度的任务分发至对应性价比的模型,控制算力成本并提升单位经济效益。
Up to 65% cost savings, automatically. That’s the beauty of Auto Mode. Replit Engineering Lead Rachael Fuller explains how Intelligent Model Routing ...
观点这表明 AI 原生应用正在加速打入企业核心业务层。私网直连消除了企业对数据过公网的合规顾虑,为构建高安全性企业级 AI 系统降低了门槛。
Vercel 🤝 AWSVercel Developers: AWS PrivateLink is now available on Vercel.Connect Vercel Functions and builds to AWS-hosted services without sending traffic over the public internet. Available for Pro and Enterprise teams with Advanced Networking.https://vercel.com/changelog/aws-privatelink-is-now-available-on-pro-and-enterprise
上个月 Moderna 报道了首个通过 III 期临床试验的个性化癌症疫苗。Moderna CEO Stéphane Bancel 与 a16z 的 Jorge Conde 讨论了这意味着什么...
事实Moderna 个性化癌症疫苗通过 III 期临床验证了计算生物学与 mRNA 结合的落地可行性。
观点顶级 VC 正在集中关注 AI 与生命科学交叉领域的端到端落地。对 AI CEO 而言,前沿科技在物理与生物世界的垂直闭环正在加速成熟,数据壁垒与高门槛领域落地方案将重新定义新一代伟大公司的估值结构。
Last month Moderna reported the first individualized cancer vaccine to pass a Phase 3. Moderna CEO Stéphane Bancel and a16z's Jorge Conde on what tha...
Pieter Levels 转发 David Pantera 推文:我们两人用几部手机拍摄了这段视频,并使用 Atlas 将其制作成一部短片。这只是 World Labs 技术博客中的一个剪辑片段。Atlas 能够生成任何摄像头都未实际拍摄到的场景部分,还可以冻结时间,让你重新构图视角。
事实World Labs 展示了 Atlas 模型通过手机视频补全未拍场景与重构视角的 3D 生成能力。
观点空间智能正加速降低三维内容生产门槛。对 CEO 意味着 3D 与视频资产制作成本将大幅下降,需提早规划 AI 原生内容工作流。
InsaneDavid Pantera: Two of us filmed this on a couple phones, and we used Atlas to turn it into a short film. This is just a clip from the @theworldlabs tech blogAtlas is able to generate the parts of the scene that none of the cameras actually filmed.It can freeze time, allowing you to reframe
ExtractBench is now live on @Kaggle. It tests schema-guided document extraction on the documents most likely to break downstream agents and workflows,...
Omarchy Quattro 在短短 18 天内实现了 20 万次 ISO 下载!自 Quattro 发布以来,平均每 8 秒就有人拉取一个全新的 Omarchy。
事实Omarchy Quattro 在 18 天内实现 20 万次下载,展示出开发者对私有化与自托管部署的极高热情。
观点
对你意味着在 AI 时代除了拥抱公有云 API,也需关注基础设施的自主可控性与成本效率。
Two hundred thousand ISO downloads of Omarchy Quattro in just 18 days! On average, someone has pulled a fresh Omarchy every 8 seconds since Quattro sh...
对你意味着在 AI 重构软硬件范式期,针对特定人群的原生操作系统与终端入口正在被重新定义,布局下一代数字原住民的入口卡位具有长期战略价值。
I love this so much. We're going to make Omarchy Kids the best possible OS for the first computer (and beyond). As a father of three, this is just meg...
对你意味着,在 AI 创业公司面临高额云端算力与运维成本的背景下,软硬件生态正加速完善自建部署与私有化降本方案。
Love to see that @dell and @intel are supporting Rails World this year! The timing couldn't be better now that there finally is an OS option for Rails...
SpaceX 与 4iG 签署了欧洲首个此类合作伙伴关系,在四个国家扩展 Starlink Mobile 网络。该协议由 4iG 首席执行官 Gellért Jászai 与 SpaceX 总裁兼首席运营官 Gwynne Shotwell 在德克萨斯州巴斯特罗普签署。Starlink Mobile 将首先在...
事实SpaceX 正在通过跨国合作推进 Starlink Mobile 在欧洲落地。
观点全球通信基础设施的无缝化覆盖正在消除物理网络盲区,这为户外具身智能、边缘 AI 终端以及全球分布式 Agent 服务的商业化落地提供了重要的底层网络支撑。
StarlinkDogeDesigner: BREAKING: SpaceX and 4iG sign a first-of-its-kind European partnership to expand @Starlink Mobile across four countries.The agreements were signed in Bastrop, Texas, by 4iG CEO Gellért Jászai and SpaceX President and COO Gwynne Shotwell.• Starlink Mobile will first launch in
观点独立开源生态正在获得强劲资金支持。这对 CEO 意味着在选型基础设施时,资金充足的开源基金会项目能提供更好的长期稳定性和技术自主度。
Omacom Foundation funding has topped $13m with four new Distinguished Patrons: @ryanrhughes, @dxhuang, @mschoening, and Adrien Treccani. The Omarchy c...
Pieter Levels 为其产品 Hoodmaps 新增自动翻译功能,灵感来自于 X 平台。Hoodmaps 上许多城市的标签是用当地语言撰写的,新增翻译功能方便用户阅读与交互。
对你意味着
事实Pieter Levels 在 Hoodmaps 中接入自动翻译以解决多语言标签阅读痛点。
观点跨语言内容屏障是全球化产品的通用痛点,利用翻译 API 或 AI 模型实现无缝本地化已成为全球化应用提升留存与增长的低成本标配方案。
✨ Added [ Auto translate ] inspired by @X itself Many cities on Hoodmaps have lots of tags written in the local language, and it'd be nice to especia...
153,000,000 American driver licenses leaked in KYC hack This is 63% of all American driver licenses In US driver licenses function as valid government...
Meet the team behind the Replit MCP this Friday at 9AM PT. Join the Replit community team live with Amadeo Pellicce (@amapel) and Jean-Luc Thumm (@jea...
to make the “ai software factory” work, we need to be qualitymaxxing. every software product should have this running in the background in six month...
确实如此,我认为现在可以看到 AI 公司正在吸走整个行业(许多是独立开发者活跃的领域)的真实经济效应。我也认为 SaaS……
对你意味着轻量级单点 SaaS 的防御壁垒正被快速抹平。
事实Pieter Levels 观察到 AI 公司正在吸收独立开发者传统领地的市场收益。
观点AI CEO 需避免做简单的 API 套壳,应聚焦于端到端工作流闭环与深层数据壁垒。
True I think you have the real economic effects of AI companies sucking up entire industries now (a lot where indie hackers operate) I do think the Sa...
The Bull Case for Linear Being a $100BN Company: "Linear is the clear winner. They have built an agentic product first that allows us to build 100x mo...
观点打造 AI 驱动的消费级应用时,CEO 不应被公共舆论茧房误导,传统社交巨头的网络效应与用户黏性依旧稳固,切入消费级 Agent 或社交入口的竞争依然极其残酷。
Hating on social media is in vogue. So is declaring its "enshittification." Contrary to the narrative, user satisfaction is up on every major social p...
✨ Added a channel switcher now It auto tags videos with @xai and then shows you them The moment you write a prompt yourself it will switch back to �...
观点高频即时反馈形成的沉浸式体验是 AI 产品提升留存的核心。CEO 应关注产品能否建立无摩擦的连续人机交互闭环。
I still need to record my Fable 5.1 video. The problem is, in order to do that, I have to stop prompting Fable 5.1 for long enough to go upstairs and ...
观点提醒 CEO 在构建 AI 内容生成与分发平台时,需提前规划内容审核策略,防止高误杀率损害用户体验。
No idea why everything is flagged as adult content Here's some random broadcast I recorded now It feels a bit like Adult Swim, and is all extremely ab...
观点对应用层 AI 创业者而言,依赖单纯的模型能力差异将不可持续,必须迅速抢占垂直工作流与私有数据,建立自身独立的业务护城河。
The two ways that large language models will win “If you are a model provider, you are basically looking to dominate the platform era and get really ...
World Labs CEO 李飞飞谈及对世界模型的需求:“大自然中并没有语言。你走到大自然中,天空上并没有写着文字……”
言外之意,纯语言模型与上下文工程无法完全解决物理世界的感知与交互难题。
事实李飞飞创立 World Labs 研发 3D 空间智能与世界模型。
观点对 AI CEO 而言,若业务涉及具身智能、空间计算或物理交互,需提早布局物理世界模型路线。
World Labs CEO @drfeifei on the need for world models: "There's no language out there. You don't go out in nature and there's words written in the sky...
Have you ever felt scared by how powerful your model is? This really makes me wonder: are we living in a simulation? The whole world may just be a sta...
观点AI 创业公司极易陷入“持续优化模型/添加新功能”的研发陷阱。 CEO 需保持警惕,快速发布以获取真实市场反馈,避免在假设中浪费迭代周期。
When I did office hours with this startup in the middle of August, I told them their product was already good enough and that they should stop writing...
观点对 AI CEO 而言,打造全球化公司时,团队与目标市场需具备极高的行动力与创新容忍度,美国依然是验证 AI 原生产品商业价值的首选高地。
You stop wondering why the US is so insanely rich when you travel the rest of the world In most of the world there is: - no real willingness to solve ...
Omacom Foundation 的金库账户刚刚开立,但如果我们保持这种增长速率,我们应该能够买下微软和苹果……
事实DHH 借金库初建即按增长率计算收购巨头的段子,讽刺了行业中普遍存在的增长外推幻觉。
观点
对你意味着在评估自身 AI 产品 PMF 或向投资人汇报数据时,切勿把早期基数极低时的增速当作长期复利,需警惕虚荣指标造成的战略误判。
Omacom Foundation treasure account was just opened, but if we keep this rate of increase, we should be able to afford to buy both Microsoft and Apple ...
I want to prevent a race into unmonitorability kicked off by confused reporting. The depth of the computation graph for our present frontier models, i...
网络安全领域的 AI 应用即将呈现垂直爆发态势。大模型在发现和利用漏洞方面的能力正变得极其强悍。前沿模型正...
事实前沿模型在自动寻找与利用漏洞能力上显著提升。
观点网络安全正从辅助工具转向由 AI 主导的自动化攻防。
对你意味着须立即提升企业 AI 应用的安全防护,同时关注该领域由 AI 驱动的垂直创业机会。
AI for cyber is about to go vertical. The models increasingly becoming insanely good at finding and exploiting vulnerabilities. Frontier models are ah...
Dwarkesh and I had a great conversation. We cover the swarm's many ambitious cheating R&D projects, discuss how much more serious it could have been i...
Everything is codeDamian Kofman: I never imagined spending 80% of my time in front of a CLI to build an outbound sales process. But as @rauchg says, everything is code now… even sales.
这未必是大家希望因 AI 发展而看到激增的岗位类型,但它很好地例证了随着 AI 能力提升,市场对工作需求发生急速转移的现象。
对你意味着
事实Ethan Mollick 观察到 AI 能力提升正在引发岗位需求的快速结构性转移。
观点AI 的落地不仅是替代原有岗位,还会衍生新的过渡与协同需求。CEO 在组织设计与招聘时,需突破传统岗位定义,构建随 AI 能力迭代而快速重组的敏捷团队。
Not exactly the type of jobs you want to see boom as a result of AI, but a good example of how what work is in demand may shift rapidly as AI capabili...
观点言外之意,CEO 无法依赖渐进式的组织升级,必须立刻推动全员 AI 原生工作流重构,提升组织认知迭代的速率。
"The Industrial Revolution gave a farmhand's son forty years to become a factory hand. This revolution may offer five." @Konstantine on the cognitive ...
随着 Agent 的发展,我们在 AI 的实际能力与公众认知之间再次面临巨大的差距。指数级增长意味着这种差距正随着时间推移不断扩大。突然之间,...
事实大众与传统客户对 Agent 的认知仍停留在对话助手阶段,而一线技术能力已快速迭代。
观点这种认知迟滞造成了巨大的“落差红利”。
对你意味着,在竞争对手低估 Agent 交付能力时,率先基于最新能力重塑工作流,可形成显著壁垒。
With agents, we are at another large gap between AI abilities & public perception. Exponential gains mean that the gap is growing over time. Suddenly,...
What changes when you're building software for agents not for humans? “The experience is going to be defined by the slowest point in that chain. The ...
What is required to get the most out of models? “To take the most advantage out of models, you need to do something a little different from just rout...
多伦多大学数学家 Daniel Litt 讨论了 AI 如何改变学术数学领域的奖励激励机制,以及他是如何轻松证明这一点的:“目标……”
事实AI 正在改变学术研究的评价与验证门槛。
观点
对你意味着高智力岗位的考核逻辑面临重构。构建 AI 原生企业时,需警惕传统绩效指标被自动化工具操纵,应重新设计结合 AI 验证机制的人才评估体系。
University of Toronto mathematician Daniel Litt on how AI has changed academic mathematics' reward incentives, and how easily he proved it: "The goal ...
University of Toronto mathematician Daniel Litt on the risk that AI narrows mathematics when everyone runs the same generative slot machine: "We've se...
Prinz has been running legal benchmarks against AI systems and the latest models are very good... but cheaper/low reasoning models are not very good. ...
本周 AI 领域的发布极其密集,如今 Muse 也推出了重大的新模型更新。在能力进阶方面,我们似乎正达到逃逸速度……
事实本周 Muse 等多个模型厂商密集发布更新。
观点基础模型演进正呈现加速态势,底层能力的频繁突破持续挤压应用层包装空间。
对你意味着产品架构必须保持极高的解耦灵活性,避免在当前模型暂时的缺陷上做过于沉重的定制开发。
This week is absolutely nuts for AI releases, and now a major new model update from Muse. We appear to be reaching escape velocity in capability progr...
World Labs co-founder Justin Johnson on how five photos from your phone could teach a robot to work in a room it has never seen: "I think there's a co...
Cache writes are the vast majority of the "usage" of Fable 5.1 in Claude Code, at over 65% of the cost in my real world usage. The cache read price cu...
多伦多大学数学家 Daniel Litt 表示,AI 在数学方面的能力瓶颈在于验证:“我的感觉是,[AI 模型] 的原因……”
事实数学家指出 AI 数学能力受限于验证机制。
观点对 CEO 而言,核心壁垒正从单纯的“大模型生成”转移至“领域自动验证器(Verifier)”。在构建垂直 Agent 和复杂推理产品时,打造高精度的自动验证反馈闭环比单纯依赖生成能力更具战略价值。
University of Toronto mathematician Daniel Litt says AI's math capabilities are bottlenecked on verification: "My sense is the reason [AI models are] ...
University of Toronto mathematician Daniel Litt and a16z's Lisha Li on AI's impact on mathematics: The models are good at a narrower slice of math tha...
We're both making installing Omarchy on M and trying it out first way better, way easier, and way faster. Unbelievable pace from the Omarchy M team �...