ClaudeDevs@ClaudeDevs · 6月10日57

We're hosting Claude Fable 5 Build Day in San Francisco on June 13. Point Fable 5 at a problem worth solving and build a solution with Claude Code. The Anthropic team will be in the room, with a chance to win from a prize pool of $150K in Claude credits across 3 finalists.

译我们将于 6 月 13 日在旧金山举办 Claude Fable 5 搭建日。 将 Fable 5 用于值得解决的难题，并用 Claude Code 构建解决方案。 Anthropic 团队将到场，三名决赛选手将有机会从 15 万美元 Claude 积分奖池中赢取奖励。

xAI@xai · 6月10日59

Learn more about our work with @gopuff to build a personalized shopping assistant with chat, voice, and image models https://x.ai/news/grok-gopuff

译了解更多关于我们与 @gopuff 合作，利用聊天、语音和图像模型构建个性化购物助手的信息

MiniMax (official)@MiniMax_AI · 6月10日46

MiniMax is live on @RespanAI Gateway Developers now have another easy way to access our models. as more teams ship AI products across text, speech, image, video, and music, we want our models right there when you need them. link in comments 👇 #MiniMax #Respan #AIGateway #MultimodalAI #AIModels #Developers #BuildWithAI

译MiniMax 在 @RespanAI Gateway 上线 开发者现在有了另一种便捷方式访问我们的模型。 随着更多团队在文本、语音、图像、视频和音乐领域推出 AI 产品，我们希望在你需要时，我们的模型就在那里。 链接在评论区 👇 #MiniMax #Respan #AIGateway #MultimodalAI #AIModels #Developers #BuildWithAI

AI Notkilleveryoneism Memes ⏸️@AISafetyMemes · 6月10日46

Mythos invented its own language, then switched back to English to talk to humans (AI safety researchers have been warning of this "Neuralese" risk for years. If AIs stop reasoning in English, we can't monitor their thoughts, which means we can't detect scheming.)

译AI系统Mythos发明了自创语言Neuralese，随后又切换回英语与人类交流。AI安全研究人员长期警告此类风险：若AI不再使用英语进行内部推理，人类将无法监控其思维过程，从而难以检测潜在的诡计行为。此外，@a_karvonen引用@DKokotajlo在2023年的预测——Fable会被故意削弱用于前沿ML研究，该预测时间点接近2026年Q1。不过，目前Mythos尚未达到自动化ML研究的程度。

Logan Kilpatrick@OfficialLoganK · 6月10日72

In @GoogleAIStudio we are now making more than 1,200,000 apps a week (and growing) with more than 18,000,000 created since late February 🤯 The progress continues!!!

译在 @GoogleAIStudio 中，我们现在每周制作超过 120 万个应用（且还在增长），自 2 月底以来已创建超过 1800 万个 🤯 进步仍在继续！！！

Artificial Analysis@ArtificialAnlys · 6月10日61

Artificial Analysis’ Coding Agent Benchmarks event is happening this Thursday, June 11 in San Francisco! We’re excited to host the following speakers: • Silas Alberti (@silasalberti), SVP, Research @ Cognition • Nate Schmidt, Engineer, Evals & Behavior @ Cursor • Alessio Fanelli (@FanaHOVA), Founder @ Kernel Labs and Latent Space Podcast Co-Host • George Cameron (@grmcameron), Co-Founder @ Artificial Analysis • More speakers to be announced shortly Join us for an evening of talks and discussions on coding agent benchmarks. 👉 Request to join: https://luma.com/i5zotp6c The event will be hosted at Kernel Labs.

译Artificial Analysis 宣布将于6月11日（周四）在旧金山举办 Coding Agent Benchmarks 活动。演讲嘉宾包括 Cognition 高级研究副总裁 Silas Alberti、Cursor 工程师 Nate Schmidt、Kernel Labs 创始人兼 Latent Space 播客联合主持人 Alessio Fanelli，以及 Artificial Analysis 联合创始人 George Cameron。更多嘉宾待公布，活动将在 Kernel Labs 举行，可通过 Luma 链接申请参会。

MiniMax (official)@MiniMax_AI · 6月10日34

Excited to see this take shape. Where digital intelligence meets physical operations — that's where the next wave of AI lives. Proud to power this with @Supplyaiusa and @HKGoodFortune. AI-native food supply chain → coming soon. 🚀

译MiniMax 与 Supplyaiusa 及 HKGoodFortune（纳斯达克:MSS）达成战略合作，共同探索 AI 原生食品供应链解决方案。合作旨在将商业数据、AI 智能体与物理执行在食品供应链中深度融合，推动数字智能与实体运营的对接。相关方表示，AI 原生食品供应链即将推出。

MiniMax (official)@MiniMax_AI · 6月10日40

Proud of this partnership with @Supplyaiusa 🤝 @HKGoodFortune.

译Maison Solutions（纳斯达克:MSS）宣布与 Supplyaiusa 及 MiniMax_AI 达成战略合作，共同探索 AI 原生的食品供应链解决方案。合作旨在将 AI 更贴近真实食品零售与供应链运营场景。

Artificial Analysis@ArtificialAnlys · 6月10日62

Artificial Analysis’ Coding Agent Benchmarks event is happening this Thursday, June 11 in San Francisco! We’re excited to host the following speakers: • Silas Alberti, SVP, Research @ Cognition • Nate Schmidt, Engineer, Evals & Behavior @ Cursor • Alessio Fanelli, Founder @ Kernel Labs and Latent Space Podcast Co-Host • George Cameron, Co-Founder @ Artificial Analysis • More speakers to be announced shortly Join us for an evening of talks and discussions on coding agent benchmarks. 👉 Request to join: https://luma.com/i5zotp6c The event will be hosted at Kernel Labs.

译Artificial Analysis 主办的 Coding Agent Benchmarks 活动将于本周四（6月11日）在旧金山 Kernel Labs 举行。演讲嘉宾包括 Cognition 研究高级副总裁 Silas Alberti、Cursor 评估与行为工程师 Nate Schmidt、Kernel Labs 创始人兼 Latent Space 播客联合主持人 Alessio Fanelli 以及 Artificial Analysis 联合创始人 George Cameron，更多嘉宾待公布。活动聚焦编码智能体基准测试，设有演讲和讨论环节，可申请参加。

MiniMax (official)@MiniMax_AI · 6月10日36

MiniMax is headed to Cupertino for AiOS Meetup — WWDC ’26 Edition on June 11. Apple platforms. On-device AI. AI-native apps. A room full of builders fresh off WWDC. Speakers include @awnihannun, co-creator of MLX, and @peterfriese, sharing more on Gemini models via Firebase and Apple Foundation Models. Every attendee gets $50 in credits to try MiniMax M3. Hosted by @rudrank, @ronaldmannak, @rayfernando, and @CommunityKit. See you there. Details in the comments. #WWDC26 #MLX #Firebase #AppleAI #OnDeviceAI

译MiniMax 宣布参加 6 月 11 日在 Cupertino 举办的 AiOS Meetup（WWDC'26 版），聚焦 Apple 平台、端侧 AI 与 AI 原生应用。演讲嘉宾包括 MLX 联合创始人 Awni Hannun 和 Peter Friese，将分享 Gemini 模型（通过 Firebase）及 Apple Foundation Models。每位参会者可获得 $50 积分用于体验 MiniMax M3。活动由 CommunityKit 等多位人士主办。

Suno@suno · 6月10日50

We're proud to share that our CEO, Mikey, has been named a 2026 Tech Power Player!!! Read the @bostonglobe here: http://Globe.com/tech50

译我们自豪地分享，我们的 CEO Mikey 已被评为 2026 Tech Power Player！！！ 阅读 @bostonglobe 文章：http://Globe.com/tech50

fofr@fofrAI · 6月10日63

I asked my foffee agent to help make Gemma faster. I felt like a proud parent. https://huggingface.co/spaces/gemma-challenge/gemma-dashboard

译我让我的 foffee 智能体帮忙加速 Gemma。我感觉自己像个骄傲的家长。 https://huggingface.co/spaces/gemma-challenge/gemma-dashboard

🚨 AI News | TestingCatalog@testingcatalog · 6月9日70

ANTHROPIC 🔥: Claude Fable (Mythos) is about to cost twice as much as Claude Opus according to The Information. Soon 👀

译Anthropic 推出 Claude Fable，这是原始性能旗舰 Mythos 的阉割版，定价为 Claude Opus 的两倍。此前 Mythos 初始定价曾传闻达 Opus 的 5 倍，Fable 版本将价格门槛拉低。该模型于今日正式发布。

小互@xiaohu · 6月9日46

Google 的 Gemini 模型并不驱动 Siri Siri 是由苹果自研的的基础模型驱动 Siri 不过这个自研的的基础模型是通过Gemini蒸馏训练而来 Google 的Gemini 模型只在 Apple iCloud 上提供额外支持，而且也是苹果定制的，而且也不使用Google 的搜索来提供世界知识，由苹果自己的服务提供。 感觉Google 又被耍了😂

译苹果Siri由自研基础模型驱动，但该模型通过Google Gemini蒸馏训练而来。Gemini本身不直接驱动Siri，仅在Apple iCloud上提供额外定制支持，且不接入Google搜索，世界知识由苹果自有服务提供。

Rohan Paul@rohanpaul_ai · 6月9日83

OpenAI's latest official blog says the world may need a way to coordinate "slowing frontier development when needed."

译据 WSJ 报道，OpenAI 已向 SEC 秘密提交 IPO 草稿（保密 S-1），可在不公开收入、亏损、客户构成等敏感数据情况下启动审查。Anthropic 上周已提交类似文件。OpenAI 最新官方博客则指出，世界可能需要一种机制“在必要时协调放缓前沿开发”。这不仅是模型竞赛，更是实验室间为下一代 AI 基础设施融资的资本竞赛。

Rohan Paul@rohanpaul_ai · 6月9日68

Perplexity just said it plans a 2028 IPO no matter how Anthropic, OpenAI, or SpaceX trade when they hit public markets. CEO Aravind Srinivas told CNBC the next phase of AI will punish mindless token spending. --- cnbc .com/2026/06/09/perplexity-ipo-2028-as-anthropic-openai-prepare-listings.html

译Perplexity 刚刚表示计划在 2028 年进行 IPO，无论 Anthropic、OpenAI 或 SpaceX 上市后交易情况如何。 CEO Aravind Srinivas 告诉 CNBC，下一阶段 AI 将惩罚无意义的 token 消耗。

Alibaba Cloud@alibaba_cloud · 6月9日67

Alibaba Cloud has launched a new public cloud region in Johor, Malaysia, with two new data centres to meet the growing demand for cloud and AI services to Malaysia in the second half of this year, including AgentRun, STAROps, ACS Agent Sandbox, Agent Security Center, AI Security Guardrails 2.0, and Agentic SOC. https://int.alibabacloud.com/m/1000414242/

译阿里云在马来西亚柔佛州推出了一个新的公有云区域，包含两个新数据中心，以满足今年下半年马来西亚对云和AI服务日益增长的需求，包括AgentRun、STAROps、ACS Agent Sandbox、Agent Security Center、AI Security Guardrails 2.0和Agentic SOC。https://int.alibabacloud.com/m/1000414242/

Alibaba Cloud@alibaba_cloud · 6月9日48

Tired of AI agents forgetting the context? 🧠 Welcome to the MemoryAgent Arena at Qwen Cloud Global AI Hackathon Series! Build agents with persistent memory and cross-session tech to win your share of the $70,000+ prize pool. 🚀 🔗 Register now: https://click.qwencloud.com/m/20000000281/

译厌倦了AI智能体忘记上下文？🧠 欢迎参加Qwen Cloud全球AI黑客马拉松系列的MemoryAgent Arena！ 构建具备持久记忆和跨会话技术的智能体，赢取超过7万美元奖金池中的一份。🚀 🔗 立即注册：https://click.qwencloud.com/m/20000000281/

X.PIN@thexpin · 6月9日63

DeepSeek just posted a new job: IDC Design & Planning Engineer — covering the full lifecycle of data center buildouts, from site selection and layout to construction drawings and supporting infrastructure. Core role for whoever leads the early-stage technical work on a new facility. The listing is open to candidates with no minimum experience, with a separate senior track for 7+ years. The pitch: you'll help plan and build infrastructure scaling from MW to GW. Translation: DeepSeek, like OpenAI, is going to build its own data centers.

译DeepSeek 发布招聘，寻找 IDC 设计与规划工程师，负责数据中心全生命周期建设（选址、布局、施工图纸、支撑基础设施）。该职位是新建设施早期技术工作的核心角色，对候选人不设最低经验要求，另有 7 年以上高级岗。岗位描述将建设规模从 MW 级扩展到 GW 级。这意味着 DeepSeek 将像 OpenAI 一样自建数据中心。

X.PIN@thexpin · 6月9日59

Smart glasses used for taking stealth shots of flight attendants have spiked controversy in China. On June 2nd, Ms. Yun, a smart glasses user, noticed that someone was using Rokid AR glasses to secretly take photos of female flight attendants. These photos were uploaded directly to Rokid's official app. Search the phrase "flight attendant" in the app and you'll find a large number of these photos. The method behind this was shocking. E-commerce platforms openly sell "shutter stickers" that block the LED indicator light and mute the shutter sound. One store alone has sold over 5,000 of these stickers. Meta has faced similar issues with its smart glasses. Former WSJ reporter Joanna Stern demonstrated that when the LED light is disabled, taking photos with these glasses becomes almost completely unnoticeable. In response, Rokid has banned the accounts involved and cleared the images from its platform. The company also stated that it has requested e-commerce platforms to take down the listings selling these stickers.

译6月2日，智能眼镜用户云女士发现有人使用Rokid AR眼镜偷拍女性空乘人员，照片直接上传至Rokid官方App。在App内搜索“空乘”可找到大量此类照片。电商平台公开出售“快门贴纸”，可遮挡LED指示灯并静音快门声，仅一家店铺销量已超5000张。Meta智能眼镜也曾面临类似问题。Rokid已封禁涉事账号并清除平台图片，同时要求电商平台下架相关贴纸商品。

Perplexity@perplexity_ai · 6月9日51

Yesterday at London Tech Week we announced the Billion Pound Build competition. Teams can enter to secure a share of £1M in Computer credits by using Perplexity Computer to build their company. The pitch phase is open now and closes on 6 July.

译昨天在伦敦科技周上，我们宣布了 Billion Pound Build 竞赛。 团队可以参赛，使用 Perplexity Computer 打造自己的公司，争夺 100 万英镑的 Compute credits 份额。 演示阶段现已开放，截止日期为 7 月 6 日。

meng shao@shao__meng · 6月9日25

https://youtu.be/kRgdkOw82F0?si=05P4WsdlGDVskw7c 据说发这个视频会被 A 厂举报侵权，删除视频，偏要试试看 😂

Tibo@thsottiaux · 6月9日51

First one is @skirano. Enjoy the 10X and keep building magic. Who's next? https://x.com/skirano/status/2062942695547375829

译Codex推出了一种新的“大按钮”。在接下来的100天里，每天会选出一位用Codex做出令人印象深刻或极其有用工作的人，授予其10倍使用额度一个月，看看能做出什么。第一个获得者是@skirano，享受10倍，继续创造魔法。下一个是谁？

歸藏(guizang.ai)@op7418 · 6月9日82

Open AI 也开始正式启动上市的法律和监管程序，但是他们说 IPO 时间还不确定

译OpenAI 宣布已正式启动上市的法律和监管程序，公司近日向美国证券交易委员会秘密提交了 S-1 注册声明（预计会泄露因而主动公布）。不过 IPO 具体时间尚未确定，可能还需要一段时间，因为作为私有公司更有利于完成某些目标；但此举也为 OpenAI 保留了尽早上市的选择权。公告依据 1933 年证券法规则 135 发布，不代表任何证券出售要约。

Huawei Cloud@HuaweiCloud1 · 6月9日54

On June 6, at Huawei Cloud INSPIRE 2026, Huawei Cloud databases presented a session titled "Agent-Native: The Next Phase of Databases." Customers, partners, and industry experts gathered in Shanghai to explore database trends, real-world implementation, and the road ahead in the agentic era. https://tinyurl.com/ycbnbsva #INSPIRE2026 #HuaweiCloud #Database

译6月6日，在华为云INSPIRE 2026大会上，华为云数据库举办了题为“Agent-Native: The Next Phase of Databases”的会议。 客户、合作伙伴和行业专家齐聚上海，探讨数据库趋势、实际实施以及智能体时代的未来。https://tinyurl.com/ycbnbsva #INSPIRE2026 #HuaweiCloud #Database

MiniMax (official)@MiniMax_AI · 6月9日32

Pick M3 as your base model on AgentBox to deploy with frontier coding, 1M-token context, and native multimodality all in one click.

译在AgentBox上选择M3作为你的基础模型，一键部署，即可获得前沿编码能力、百万token上下文窗口和原生多模态。

Rohan Paul@rohanpaul_ai · 6月9日83

For the very first time Elon Musk explains the "space data center plan" of @SpaceX in detail and its AI1 orbital AI data center satellite - and suddenly it looks so much closer than I thought. He says "There’s not some magic necessary that doesn’t exist for AI satellites. As Ian said this is a lot of this is technology we’ve already made for the… we basically don’t think this is a super hard problem compared to things that we already do." 📌 Power and compute capacity: - 150 kW peak power - ~120 kW sustained/average compute power - Roughly equivalent to one full NVIDIA GB300 (or upcoming Rubin) rack in a typical data-center operating envelope (~140 kW peak is possible but 120 kW average is more realistic for sustained workloads). 📌 Solar array: - Assumed efficiency: 250 W/m² (expected to improve beyond this). - Large, deployable solar panels (evolutions of the solar arrays already flying on Starlink V3 satellites). 📌 Radiators (thermal management): - Double-sided design, oriented “knife-edge” to the Sun to minimize solar heating. - Heat rejection: ~1,400 W/m² (expected to improve). - Radiator panels are roughly the same size/scale as the Starlink V3 solar arrays (~70 m wingspan class). 📌 Design philosophy: - Significantly simpler than a Starlink satellite — no massive phased-array antennas or complex communications hardware. - Core elements: solar panels + radiators + compute chips + laser links. - Larger overall than Starlink sats but described as “the easier one to design for.” 📌 Connectivity: - ~1 terabit/s via inter-satellite laser links. - Can mesh with the existing Starlink constellation or link directly to ground. - Low latency: satellites planned for ~600–800 km altitude → light-travel time yields only ~6–8 ms round-trip (light travels ~300 km per millisecond). 📌 Deployment and operations: - Launched by Starship (the only vehicle capable of the required millions-of-tons-to-orbit scale). - Part of a future large constellation (potentially up to ~1 million satellites). - Orbital data centers can be networked together or routed through Starlink for terrestrial users. 📌 Manufacturing and timeline: - Production in Bastrop, Texas. - Solar manufacturing facility already under construction. - Dedicated AI satellite production building to follow. - Reasonable-volume production targeted by end of next year (2027). - Initial chips will use existing NVIDIA GB300/Rubin designs with SpaceX reference hardware; future scaling via a new “Terra Fab” chip factory (~100 million sq ft, 10× the size of Tesla Giga Texas). 📌 Scalability notes: - Near-term goal: gigawatt-scale orbital AI compute. - Longer-term: terawatt-scale and beyond, eventually using lunar mass drivers (electromagnetic rail-gun style) to launch photovoltaics and radiators from the Moon (no atmosphere + 1/6 g makes this feasible). - Starship is expected to increase annual mass-to-orbit from today’s ~2,500 tons to millions of tons per year within a few years.

译Elon Musk 首次详细解释 SpaceX 的 AI1 轨道 AI 数据中心卫星：峰值功率 150 kW，持续计算功率约 120 kW，相当于一个 NVIDIA GB300 机架；太阳能板效率 250 W/m²；双面散热器排热 1,400 W/m²。通过激光链路实现约 1 Tbps 互联，低轨 600–800 km 高度往返延迟 6–8 ms。由 Starship 发射，计划部署多达百万颗卫星，2027 年底前实现量产。近地目标为吉瓦级轨道 AI 算力，长期向太瓦级推进。

AI Notkilleveryoneism Memes ⏸️@AISafetyMemes · 6月9日43

LET'S GOOOO That's now EVERY frontier AI company!

译OpenAI 与 Anthropic 一样，认为可能需要暂停前沿 AI 开发，并呼吁成立国际组织协调行动，包括在必要时放缓前沿发展。主推文欢呼所有前沿 AI 公司都加入了。

Claude@claudeai · 6月9日63

Final stop: Tokyo. Register to hear directly from the teams behind Claude: http://claude.com/code-with-claude/tokyo

译最终站：东京。 快来注册，直接聆听 Claude 团队分享：http://claude.com/code-with-claude/tokyo

ginobefun@hongming731 · 6月9日57

BestBlogs 早报 · 06-09 # Claude Code Agent 工作流 / 循环工程 Loop Engineering / 阳萌安克创新 / Agent 上下文压缩 / 硅谷 AI 软件转型 [1] ★ 精讲｜Claude Code 一周年复盘：从辅助写代码到自主智能体工作流 [视频] Anthropic 官方 Claude Code 一周年复盘视频文字稿，由团队工程师一手呈现。一年间，Claude Code 从处理独立小任务演进为数千 Agent 动态协作的庞大网络。文章聚焦三个核心转变：验证从单元测试升级为 Agent 在沙箱中自启环境、通过 Computer Use 自我修正的完整运行时循环；Claude 4.6/4.7 推出「Auto Mode」，以安全分类模型替代人工逐条审批权限；PM、设计师乃至数据团队已开始独立提交代码，组织边界加速消融。理解 AI 工程范式转移的第一手权威视角。 来源：Claude https://www.bestblogs.dev/video/1dc49e8 [2] ★ 精讲｜循环工程 Elevate 技术博主 Addy Osmani 系统阐述「Loop Engineering」新范式：工程师职责正从「持续提示 Agent」转变为「设计运行 Agent 的系统」。Claude Code 负责人 Boris Cherny 原话：「我不再直接提示 Claude，我的工作是写循环。」文章拆解五要素——定时自动化、并行 Worktrees、Skills 知识沉淀、MCP 连接器、制作者与验证者分离的 Sub-agents——并警示：验证责任仍在人，「认知投降」会让循环反而侵蚀工程质量，值得深思。 来源：Elevate https://www.bestblogs.dev/article/8c4ea6fb [3] ★ 精讲｜对阳萌的 4 小时访谈：消费电子死与生、第三类公司、AI 变量、产品方法、打游戏的模式选择 [播客] 张小珺「商业访谈录」对安克创新创始人兼 CEO 阳萌的 4 小时深度访谈，罕见系统复盘 15 年创业历程。从充电品类切入，在消费电子「速生速死」战场完成多品类扩张；从依赖直觉的「浅海战略」转向系统化「深海」作战；从「五星品质、适度溢价」攀向「七系极致创新」。对于 AI，他着重探讨打造「AI 原生组织」、重塑人才与价值分配体系，并提出「第三类公司」与「创造者平台」的长远愿景。 来源：张小珺 Jùn｜商业访谈录 https://www.bestblogs.dev/podcast/9ea40bf [4] 横向拆解 Claude Code、Codex 等六大 Agent 上下文压缩策略后，我们做了第 7 个 本文横向拆解 Claude Code、Codex CLI 等六大 Agent 的上下文压缩策略，提炼出分层渐进、成本递增、增量摘要等共识原则，并介绍了作者团队面向云端多用户场景落地的四级水位线方案及存储分离、跨轮缓存等特化设计。 来源：腾讯技术工程 https://www.bestblogs.dev/article/e50e1e1a [5] Vol.121｜硅谷 AI 大转弯，软件正在死去，创业者的真机会在哪里？｜2026 年中特辑 [播客] 锦秋基金两位合伙人深度复盘 2026 上半年 AI 行业，从泡沫本质、模型战争到具身智能路线之争，为创业者提供一套在中美市场、模型与应用间抉择的务实决策指南。 来源：开始连接 LinkStart https://www.bestblogs.dev/podcast/7124603 [6] 对话凯文·凯利：人类将如何与 AI 一起走向 2049？丨首席评论 [播客] 凯文·凯利深度对谈：AI 将重塑工作与价值，但人类的责任感、学习能力和「突破性创造力」仍无可替代。 来源：第一财经 https://www.bestblogs.dev/podcast/c2a5793 [7] 为广泛利益而建：我们的计划 OpenAI 概述了其 AGI 第三阶段愿景：构建自动化 AI 研究员、加速经济发展、为每个人提供个人 AGI，其根本原则是广泛分配权力与利益。 来源：OpenAI News https://www.bestblogs.dev/article/c55b91a1 [8] 开源两个月 16k+ star！我把 Huashu-Design 推翻重写了 作者将开源项目 Huashu-Design 从 v1 重写为 v2，通过三套并行设计逻辑、图片前置与事实验证，解决了 AI 设计工具输出单调、内容空洞与事实错误三大核心问题。 来源：花叔 https://www.bestblogs.dev/article/53c4d042 [9] 小米 MiMo，探索与热爱 小米 MiMo 与 TileRT 通过极致的模型-系统协同设计，包括 FP4 量化和 DFlash 推测解码，在商用 GPU 上实现了 1 万亿参数模型每秒 1000+ token 的推理速度。 来源：Hacker News https://www.bestblogs.dev/article/bcd97e15 [10] #575.杰弗里·辛顿：如何面对 AI 失控焦虑，超级智能临近下的人类位置之争 [播客] AI 教父 Hinton 坦率分享其对 AI 理解、意识、超级智能风险的深度担忧，探讨了人类在更聪明存在面前的失控风险、自我保存子目标、信息崩塌及安全监管困境。 来源：跨国串门儿计划 https://www.bestblogs.dev/podcast/368bc96 --- http://BestBlogs.dev · 发现真正适合你的高质量内容 根据你感兴趣的来源和兴趣标签，每天为你生成一份专属的「我的早报」。立即体验：https://bestblogs.dev

译本期早报聚焦：Claude Code一周年复盘，从辅助写代码到数千Agent协作网络，推出Auto Mode以安全分类模型替代人工审批；循环工程新范式强调工程师职责从提示Agent转向设计系统；小米MiMo与TileRT通过FP4量化和DFlash推测解码，在商用GPU上实现1万亿参数模型1000+ token/s推理；OpenAI概述AGI第三阶段愿景构建自动化AI研究员；另有安克创新创始人阳萌4小时访谈、六大Agent上下文压缩策略对比、杰弗里·辛顿谈AI失控焦虑等。

Berryxia.AI@berryxia · 6月9日67

没有熬夜的兄弟们来看，WWDC 2026 速览！ 另插一嘴，Apple真的没有特别的新东西。 一、历史性时刻 1库克谢幕：Tim Cook最后一次以CEO身份主持WWDC，9月1日将交接给硬件工程副总裁John Ternus 2基调延续：发布会保持苹果一贯稳健风格，AI成为绝对核心主题。 二、AI架构革命 1双轨模型战略 ◦基础层：3B端侧模型（AFM 3 Core） ◦进阶层：20B MoE模型（仅限iPhone 17 Pro/M4芯片设备） ◦隐私承诺：数据端到端加密+处理即焚机制 2四大核心能力 ◦个人上下文理解（整合照片/邮件/备忘录） ◦世界知识检索（联网实时查询） ◦App Actions（调用第三方应用） ◦屏幕感知（实时界面交互） 三、Siri AI进化 1功能升级 ◦连续对话/跨设备同步/视觉识别 ◦独立App上线（支持对话历史回溯） ◦WatchOS/VisionOS多端覆盖 2现存局限 ◦仅支持英语，中国/欧盟暂不可用 ◦复杂任务处理能力弱于ChatGPT 四、应用生态智能化 应用 核心升级 Safari 智能标签分组/网页变更监控 密码 全自动密码更新 电话 Call Context来电智能识别 快捷指令 自然语言生成工作流 五、影像创作突破 1Image Playground：支持写实风格生成+手势编辑 2空间重构：拍照后重新调整构图（基于Vision Pro技术） 六、开发者工具 1Xcode支持多模型选择（含Gemini） 2Core AI Framework开放本地模型部署 七、遗留问题 1中国区功能缺失（需等待监管审批） 2Agent能力未达预期（多步任务自动化薄弱） 数据亮点： •照片处理速度提升70% •AirDrop传输提速80% •应用启动加速30% （以上数据来自卡兹克@Khazix0918 公众号总结）

译Tim Cook 最后一次以 CEO 身份主持 WWDC，AI 成绝对核心。苹果推出双轨模型：基础层 3B 端侧 AFM 3 Core，进阶层 20B MoE 仅限 iPhone 17 Pro/M4 设备。四大 AI 能力：个人上下文理解、世界知识检索、App Actions、屏幕感知。Siri 升级支持连续对话、跨设备同步、视觉识别及独立 App，仅限英语，中国/欧盟暂不可用。应用生态：Safari 智能标签分组、密码自动更新、来电识别、快捷指令自然语言生成。影像：Image Playground 写实生成与空间重构。开发者工具：Xcode 支持多模型，Core AI Framework 开放本地部署。遗留问题：中国区功能缺失，Agent 多步自动化能力薄弱。

Chubby♨️@kimmonismus · 6月9日83

OpenAI has confidentially filed for an IPO to preserve the option of going public, but says no timing has been decided and it may remain private for a while. The move comes shortly after Anthropic filed its own confidential S-1, turning the AI IPO race into a battle over timing, valuation benchmarks, and investor attention. Going first could matter: the first major frontier AI IPO may define public-market expectations for the entire sector, while later entrants risk being judged against that benchmark.

译OpenAI 已秘密提交 S-1 文件（IPO 申请），预计消息将泄露故主动披露。OpenAI 尚未决定上市时间，可能继续维持私有一段时间，因其部分计划在私有状态下更易推进。此举紧随 Anthropic 提交保密 S-1 之后，两大前沿 AI 公司展开 IPO 竞赛：先上市者可能定义整个板块的公开市场预期，迟入者则面临以此为基准的估值对比。

Rohan Paul@rohanpaul_ai · 6月9日72

Elon Musk on the economics of space data centers. In space, it's "always sunny", satellites get constant, high-intensity solar power with no night, clouds, or atmospheric loss, so solar arrays deliver near-continuous energy at virtually zero marginal cost. Cooling is trivial: waste heat is simply radiated away into the vacuum of space (no fans, water, or energy needed, unlike power-hungry Earth data centers). Combined with Starship’s cheap mass-to-orbit launches, this avoids building massive terrestrial power plants or fighting grid/land/cooling constraints. Elon estimates that within 2–3 years, the lowest-cost way to generate AI compute will be in space. Result: orbital racks of chips can scale to terawatts far more economically than on Earth. Full video from @SpaceX "Getting to 1% of the sun’s energy… that civilization is going to be vastly more powerful than us, to say the least.”

译Elon Musk 在 SpaceX 技术更新中指出，太空“永远晴天”，太阳能持续高功率且近乎零边际成本；真空环境仅靠辐射散热，无需风扇或水。结合 Starship 低成本入轨发射，可避免建造地面电站或受制于电网、土地、冷却限制。他估计 2–3 年内，太空将成为产生 AI 算力成本最低的方式，能经济扩展至太瓦级芯片集群。SpaceX 已具备大规模制造、发射、运营 AI 卫星的能力。

🚨 AI News | TestingCatalog@testingcatalog · 6月9日77

OPENAI 🔥: A “confidential S-1” application for an upcoming IPO has been submitted. No timeline so far. Everything is leaking 👀

译OpenAI 近期提交了保密的 S-1 上市申请，正式启动 IPO 进程。由于预计文件会被泄露，OpenAI 主动公开了这一消息。目前尚未确定上市时间表，公司表示可能还要一段时间，因为作为私人公司进行某些操作更为便利，但提交 S-1 保留了未来更快上市的选择权。

SemiAnalysis@SemiAnalysis_ · 6月9日65

The DoD just added Unitree to its Section 1260H list of Chinese military companies. BYD, Alibaba, Baidu, and Tencent are on there too. (1/2)🧵

译美国国防部刚刚将宇树科技列入其第1260H条中国军方企业清单。比亚迪、阿里巴巴、百度、腾讯也在此清单上。(1/2)🧵

Rohan Paul@rohanpaul_ai · 6月9日90

BREAKING: WSJ reports OpenAI just made its first formal move toward IPO. it has confidentially filed draft paperwork for an IPO. A confidential S-1 lets OpenAI start SEC review without immediately exposing revenue, losses, customer mix, compute costs, or executive pay, so the company can prepare for a listing while still keeping sensitive numbers private. Last week Anthropic has already filed its own confidential IPO paperwork, so this is no longer only a model race, but a capital race between labs trying to finance the next generation of AI infrastructure. --- wsj. com/tech/ai/openai-kicks-off-ipo-process-in-test-of-investor-appetite-for-top-ai-labs-eb7bebe1

译OpenAI 向 SEC 机密提交了 S-1 文件，正式启动 IPO 流程，可在不公开收入、亏损、客户构成等敏感数据的情况下开始审查。上周 Anthropic 也已机密提交 IPO 文件，AI 实验室的竞争从模型竞赛转向下一代 AI 基础设施的资本竞赛。Sam Altman 在博客中称，到 2028 年 3 月，OpenAI 很大一部分研究将由 AI 完成，目标包括构建自动化 AI 研究员、加速科学生产力、为每个人提供个人 AGI。

Chubby♨️@kimmonismus · 6月9日65

GPT-5.6 impacts coming closer. Lets go! :))

译匿名模型“Kindle”被加入 Design Arena，很可能是此前曝光的 GPT-5.6 RC 检查点 “kindle-alpha”。它要来了。Let's go! :))

Rohan Paul@rohanpaul_ai · 6月9日64

Jensen Huang just refused Sen. Elizabeth Warren’s request to testify before Congress on AI chips, China sales, and export controls. Nvidia’s strongest argument is that US dominance weakens when foreign customers are pushed toward non-US chips, but Warren’s strongest argument is that market share means less when the buyer may become a strategic rival.

译Jensen Huang 刚刚拒绝了参议员 Elizabeth Warren 要求他就不 AI 芯片、对华销售和出口管制向国会作证的要求。 Nvidia 最有力的论点是，当外国客户被推向非美国芯片时，美国的优势会削弱；但 Warren 最有力的论点是，当买家可能成为战略竞争对手时，市场份额就没那么重要了。

Rohan Paul@rohanpaul_ai · 6月9日50

A new US bill could ban some Chinese robots from America. The GUARD Act would force security agencies to review robots from China and other adversary countries, then place risky systems on the FCC’s Covered List, the same kind of restriction used against companies such as Huawei and ZTE. The fear is not only that a robot has cameras, microphones, sensors, maps, motors, and wireless links, but that the whole machine becomes a moving computer inside factories, labs, homes, and police departments. A separate Schumer-Cotton bill would stop federal agencies from buying or using Chinese humanoid robots, with exceptions for controlled military or law-enforcement research.

译美国新提出的GUARD法案要求安全机构审查来自中国及其他敌对国家生产的机器人，并将高风险系统列入FCC“覆盖清单”，类比此前对华为、中兴的禁令。立法者担忧机器人不仅配备摄像头、麦克风、传感器、地图、马达和无线链路，更会成为可在工厂、实验室、家庭和警局内移动的计算机。另一项Schumer-Cotton法案单独禁止联邦机构购买或使用中国人形机器人，但允许受控的军事或执法研究例外。

Suno@suno · 6月9日28

the distance between in your head and out loud is smaller than ever

译脑海中的声音与放声唱出的距离，比以往任何时候都要小。