凡人小北@frxiaobei · 5月20日51

DeepSeek 招聘，帮转，有志青年可以看看。

译DeepSeek 发布招聘信息，计划在北京组建新团队开发 Code Harness 项目（可能称为 DeepSeek Code）。团队将负责从零开始构建该工具，招聘职位包括 Harness Product Manager 和 Harness R&D Engineer，分别侧重于产品管理和研发工作，强调研究与产品的结合。此次招聘面向有志于AI技术领域的青年人才，工作地点位于北京。

Runway@runwayml · 5月20日42

Tickets are still available for the Runway AI Festival in New York on June 11 and Los Angeles on June 18. Join us in celebrating the very best work being made with AI across film, design, new media, fashion, advertising and gaming. Get your tickets at the link below.

译Runway AI艺术节将于6月11日在纽约、6月18日在洛杉矶举行，门票仍在发售中。 诚邀您共同见证AI在电影、设计、新媒体、时尚、广告及游戏领域创作的杰出作品。 请通过下方链接购票。

宝玉@dotey · 5月20日65

DeepSeek 招做 Agent 的产品经理和研发

译DeepSeek正在招聘Agent方向的产品经理和研发工程师，并为此组建了新的Harness团队。该团队目标是构建名为Code Harness（或类似产品）的代码相关AI工具，招聘地点位于北京。这标志着DeepSeek正投入资源开发AI Agent及开发者工具。

Ethan Mollick@emollick · 5月20日62

One interesting side feature of recursive self-improvement, to the extent that is happening, is that it makes the Big Three labs more appealing to talent, and shortens the runway for launching a potential competitor instead at the same time.

译递归自我改进的一个有趣附带效应是，只要它还在发生，就会让三大实验室对人才更具吸引力，同时缩短潜在竞争对手的启动窗口。

Chubby♨️@kimmonismus · 5月20日63

Holy: Leaked audio from a Meta all-hands on April 30: Zuckerberg told employees the company is using them to train AI models before mass layoffs hit. His argument? Meta's engineers are smarter than any external workforce, so having them solve coding tasks internally will make Meta's models better, faster than competitors. The layoffs are expected Wednesday at 4 a.m. Train your replacement, then get walked out. That's the deal now.

译泄露的Meta内部音频显示，CEO扎克伯格在4月30日的全员会议上向员工表示，公司正在利用他们的工作成果来训练AI模型，随后即将启动大规模裁员。其核心论点是，Meta工程师的平均智力远高于外部可雇佣的群体，让这些顶尖人才在内部完成编码等任务，能更快速、更有效地提升公司AI模型的编程能力，从而超越竞争对手。裁员计划预计在周三凌晨启动，员工需先完成对替代者的培训。

Alibaba Cloud@alibaba_cloud · 5月20日43

Qwen Conference 2026: The Keynote Agenda AI-Native Cloud, Agent Native Cloud architectures, the Future of Inference, and Multimodal Visual drops. No fillers. Just engineering blueprints for global scale. Register: https://click.qwencloud.com/m/20000000190/

译Qwen Conference 2026：主题演讲议程 AI原生云、Agent原生云架构、推理的未来以及多模态视觉发布。没有空谈，只有面向全球规模的工程蓝图。 注册：https://click.qwencloud.com/m/20000000190/

meng shao@shao__meng · 5月20日63

两年前，Andrej Karpathy 离开 OpenAI 后启动了 Eureka Labs，要做 AI 时代的教育，让很多人包括我自己都满怀期待，这位 AI 领域极具声望的 AK 大神到底会到来什么样的全新 AI 教育？ 两年后，Andrej Karpathy 官宣加入 Anthropic，Eureka Labs 就此锁推 🔐，AI 教育就此告一段落。 Andrej Karpathy 到底看到了什么？我们无从得知！但我们能够知道的是，AI 领域少了一位能够带更多人入门了解 LLM 的大神，Anthropic 多了一位 Member of Technical Staff。。

译Andrej Karpathy在创立AI教育机构Eureka Labs仅两年后，宣布加入Anthropic。这导致旨在用AI变革教育的Eureka Labs项目终止。Karpathy表示，LLM的前沿在未来几年将至关重要，他为此兴奋并回归研发岗位。他同时承诺，仍对教育充满热情，计划日后重启相关工作。此举意味着AI领域少了一位重要的LLM教育布道者，而Anthropic则新增了一位技术成员。

Alibaba Cloud@alibaba_cloud · 5月20日60

Qwen Conference 2026: The Keynote Agenda AI-Native Cloud, Agent Native Cloud architectures, the Future of Inference, and Multimodal Visual drops. No fillers. Just engineering blueprints for global scale. Register: https://click.qwencloud.com/m/20000000190/

译Qwen Conference 2026：主题演讲议程 AI原生云、Agent原生云架构、推理的未来，以及多模态视觉技术发布。没有冗余内容，只有面向全球规模的工程蓝图。 注册：https://click.qwencloud.com/m/20000000190/

X.PIN@thexpin · 5月20日50

DeepSeek has formed a new internal team: Harness. This team is to build a coding agent aimed squarely at Anthropic's Claude Code. Senior researcher Chen Deli confirmed it: the goal is essentially "DeepSeek Code Harness," a direct answer to Claude Code. They're hiring a PM and an engineer for it, both based in Beijing at their Haidian office near Peking & Tsinghua.

译DeepSeek已成立一个新的内部团队：Harness。该团队旨在打造一个直接对标Anthropic的Claude Code的编程智能体。高级研究员陈德利确认了这一点：目标本质上是“DeepSeek Code Harness”，作为对Claude Code的直接回应。他们正在为此招聘一名产品经理和一名工程师，工作地点均在北京海淀办公室，靠近北京大学和清华大学。

X.PIN@thexpin · 5月20日57

Russia is looking to source Chinese chips to train its GigaChat AI. The news was announced during Putin's recent visit to Beijing. Sberbank's CEO confirmed that the company is eyeing Huawei's Ascend 950, the most advanced Chinese-made AI chip despite it trailing behind Nvidia's H200.

译俄罗斯正寻求采购中国芯片以训练其GigaChat AI。该消息于普京近期访华期间宣布。俄罗斯联邦储蓄银行CEO确认，公司正关注华为的昇腾950，这是目前最先进的中国产AI芯片，尽管其性能仍落后于英伟达的H200。

Alibaba Cloud@alibaba_cloud · 5月20日39

As AI agent deployments scale, how do you secure AI without slowing innovation? Join us for ClawTalks Episode 4: Secure AI Agents Across the Full Lifecycle! Set a reminder now! 🔽 📅 May 22, 2026 | 5:00 PM (UTC+8) https://youtu.be/mm9Fl1LcBXI We'll: ✅ Reveal real-world threats targeting AI agents ✅ Share Alibaba Cloud's 7 proven security best practices ✅ Demo the Agent Security Center—discover, map, and secure your AI assets in real time Don't miss these enterprise-grade AI security insights—set a reminder now! #ClawTalks #AISecurity #EnterpriseAI #AlibabaCloud #AgentSecurity #Cybersecurity

译阿里巴巴云将举办ClawTalks研讨会，聚焦大规模AI代理部署中的核心矛盾：如何在保障安全的同时不拖累创新。活动将揭示针对AI代理的真实威胁，分享七项经实践验证的安全最佳实践，并现场演示Agent Security Center工具，该工具可实时发现、测绘并保护企业AI资产。会议旨在提供企业级的AI安全见解。

swyx🛬 SFO@swyx · 5月20日65

oh no contextual got windsurfed

译噢不，Contextual AI被挖墙脚了

Alibaba Cloud@alibaba_cloud · 5月20日51

Qwen Conference 2026 is coming! Join us at Sands Expo Singapore on May 26. Featuring our keynotes on full-stack AI breakthroughs, 3 parallel forums, and a 1,000 m² exhibition with 7+ models and 6+ Agent tools. Register now: https://click.qwencloud.com/m/20000000142/

译Qwen Conference 2026即将举行！ 5月26日，新加坡金沙会展中心。 主讲全栈AI突破，设3个平行论坛， 1000平方米展区展示7+模型与6+ Agent工具。 立即注册：https://click.qwencloud.com/m/20000000142/

Rohan Paul@rohanpaul_ai · 5月20日69

FT: Google DeepMind founder Demis Hassabis was an early angel investor in Anthropic. While Google later put billions into the same company through cloud and AI deals. DeepMind diaspora is actually super strong: former researchers have built more than a dozen AI companies and raised at least $14B since 2021. DeepMind has really become less like one company and more like an AI school whose graduates now compete with, supply, and pressure Google at the same time. --- ft. com/content/8f2a529e-7a1b-4d8e-95be-338d0c4c98f5

译据英国《金融时报》报道，Google DeepMind创始人德米斯·哈萨比斯曾是Anthropic的早期天使投资者，而Google随后也通过云服务与AI合作向该公司投入了数十亿美元。更值得关注的是DeepMind显著的“人才外溢”效应：自2021年以来，其前研究人员已创立超过十家AI公司，累计融资至少140亿美元。这表明DeepMind正逐渐从一个研究机构演变为一所顶尖的AI人才“学校”，其“毕业生”同时在与Google竞争、为其提供技术，并对整个市场施加压力。

Rohan Paul@rohanpaul_ai · 5月20日65

FT: Nvidia just turned its AI chip lead into a $90B financing machine that funds the companies buying, renting, building, and extending its own computing stack. The company committed $47B to investments and partnerships through Jan-25, then earmarked another $43B in the next 4 months, covering 145+ companies across model labs, cloud providers, chip designers, and suppliers. That creates a loop where startups get capital, suppliers expand capacity, cloud firms buy more Nvidia GPUs, and Nvidia becomes harder to replace inside the AI infrastructure layer. --- ft .com/content/c6b362b8-ab6b-4723-af48-28082bdfcac2?syn-25a6b1a6=1

译英伟达将其AI芯片的领先优势转化为一个巨大的融资机器。公司在过去数月内承诺了总计900亿美元的投资与合作，覆盖145多家涉及模型研发、云计算、芯片设计和供应链的公司。这一策略创造了一个自我强化的闭环：初创公司获得资本，供应商得以扩张产能，云厂商采购更多Nvidia GPU，从而进一步巩固了Nvidia在AI基础设施层中难以被替代的核心地位。

X.PIN@thexpin · 5月20日53

🚨 Tesla China just posted a wave of jobs tied to autonomous-driving testing--fueling speculation that FSD (Full Self-Driving) is finally about to land for real. The roles: real-vehicle test technicians, test engineers, track-testing specialists--across 9 cities including Beijing, Shanghai, Shenzhen, Guangzhou, Suzhou, and Wuhan. A reporter posing as a buyer asked support if the ¥64K (~$9K) package could unlock via OTA. The reply: "Your vehicle doesn't support it yet. We're working through approval under national regulations and will push it as soon as it clears." The backstory: - FSD launched in China Feb 2025--but rebranded from "Full Self-Driving" to "smart assisted driving." - A limited free trial last March was paused after one week. - FSD has struggled with local traffic rules--likely starved for local data. But since @elonmusk 2024 visit, Tesla's location-based data restrictions have mostly melted away. This hiring push could finally give FSD its full data diet. If things break right: a full-strength FSD on Chinese roads in 2026. Musk came to China. Looks like the trip paid off again.

译特斯拉中国近期发布大量自动驾驶测试相关职位招聘，涵盖测试技术人员、工程师等岗位，涉及北京、上海等9个城市，引发市场对FSD（全自动驾驶）即将在中国真正落地的猜测。客服回应称，价值6.4万元的FSD包能否通过OTA推送取决于国家法规审批进展。FSD虽于2025年2月在中国推出，但更名为“智能辅助驾驶”，去年3月的免费试用仅持续一周后暂停，可能因缺乏本地数据而受限。自马斯克去年访华后，特斯拉在地理位置数据方面的限制大多已解除，此次招聘有望为FSD提供所需数据支持。若进展顺利，完全版FSD或将于2026年在中国道路上运行。

Kling AI@Kling_ai · 5月20日50

Kling AI Cannes Showcase — House of David & Trailer of The Old Stories: Moses Deep integration into Hollywood-grade pipelines. House of David, the first Hollywood production to openly discuss the use of AI video generation technology in the production process on an industrial level, has captivated over 44M viewers worldwide, ranked among the U.S.’s top 10 new series debuts, and reached #1 on Prime Video in the U.S. Kling AI works seamlessly with live-action production, delivering consistent, high-quality shots at scale and meeting strict industrial delivery standards — proving AI as a reliable, efficient new foundation for global filmmaking.

译在戛纳展示的Kling AI视频生成技术已深度融入好莱坞工业级制作流程。其首个公开应用于剧集《David之家》的项目，吸引了超过4400万全球观众，位列美国新剧前十并登顶Prime Video。该技术能够无缝对接实拍制作，以高标准交付大量高质量镜头，证明了AI作为全球电影制作可靠、高效新基础的潜力。

PixVerse@PixVerse_ · 5月20日16

Size ain’t everything! Hamster just out-lifted a hippo in the Olympics！

译体型不是一切！仓鼠在奥运会上举重超过了河马！

Kling AI@Kling_ai · 5月20日47

Kling AI Cannes Showcase — Technical Discussion on Born of the Tide AI-driven animation across the full industrial creative workflow. Set against the historical and cultural backdrop of the Tanka people, Born of the Tide explores a new way of bringing local stories to the screen through AI-powered animation. From art direction and environment design to character movement and cinematic shot planning, Kling AI helps streamline the animation process, making high-quality, epic-scale visual storytelling more accessible, efficient, and scalable. By lowering production barriers and reducing reliance on large teams and long production cycles, the project points to a more agile and creatively open future for original animation.

译Kling AI展示了其在动画《生来如潮》中全流程应用AI技术的实践。该项目以疍家文化为背景，利用Kling AI辅助完成了从美术指导、环境设计到角色动作、电影镜头规划等关键环节。通过AI驱动的动画制作，项目降低了制作门槛，减少了对大规模团队和漫长制作周期的依赖，使高质量、史诗级的视觉叙事变得更高效、更易扩展，指向了更具敏捷性和创作自由度的动画未来。

Greg Brockman@gdb · 5月20日79

openai offering to invest $2M in API credits in every @ycombinator startup in the current batch. compute for powering the next generation of startups.

译OpenAI宣布向Y Combinator当前批次的每家创业公司提供价值200万美元的API信用额度投资，以换取股权。此举旨在为初创公司提供AI算力资源，支持下一代创新。这让人联想到Sam Altman在YC合伙人时期，Yuri Milner向每家创业公司提供投资的先例。OpenAI希望通过此举释放创业者潜力，推动AI应用创新。

Alibaba Cloud@alibaba_cloud · 5月20日56

Qwen Conference 2026 is coming! Join us at Sands Expo Singapore on May 26. Full-stack AI breakthroughs, 3 parallel forums, and a 1,000 m² exhibition featuring 7+ models and 6+ Agent tools. Register now: https://click.qwencloud.com/m/20000000142/

译Qwen Conference 2026即将到来！ 5月26日，新加坡金沙会展中心，全栈AI突破、3场平行论坛、1000平方米展区，展示7+模型与6+ Agent工具。 立即注册：https://click.qwencloud.com/m/20000000142/

Sam Altman@sama · 5月20日66

i am excited to see what will happen with tokenmaxxing startups, both for how they work internally and the products they can build. openai offered to invest $2M in tokens into every startup in the current yc batch. happy building!

译OpenAI向当前Y Combinator（YC）批次的每家创业公司提供了价值200万美元的OpenAI代币，以换取股权。这一举动被称为“tokenmaxxing”。此做法让人联想到早年Yuri Milner向所有初创公司提供投资的历史。这一举措旨在激励最具动力的创始人去探索和构建由AI驱动的新产品与内部运作模式，业界正期待这能解锁新的创新可能性。

Tibo@thsottiaux · 5月20日24

I like to think that Codex is the work of a great team working together in unison. It's the most collaborative team I've gotten the chance to work with. But I don't think we could have done it without @ajambrosino's magic. He's been the driving force behind what makes the Codex app the app that everyone wants to emulate. And we are barely getting started.

译我喜欢将Codex视为一个伟大团队协同工作的成果。这是我有幸合作过的最具协作精神的团队。 但我认为，没有@ajambrosino的魔力，我们无法完成这一切。他一直是Codex应用成为众人争相模仿对象的核心驱动力。而我们才刚刚起步。

AYi@AYi_AInotes · 5月20日63

Damn! GitHub这次被黑,可以说是人类历史上第一次AI网络战了吧？ 官方只轻描淡写说在调查内部仓库未授权访问,但社区已经炸锅了。 据网络犯罪论坛爆料,黑客疑似拿到了Anthropic的Mythos安全AI,用它精准突破了GitHub的防线,偷走了约4000个核心内部仓库， 里面有Copilot的源码,CodeQL的算法,还有Actions运行时和整个计费系统。 但这还不是最可怕的， 以前黑客靠挖漏洞吃饭， 现在黑客靠模型吃饭， 以前防御方靠打补丁。 现在防御方需要比对手更好的AI。 也就是说，攻防的天平已经彻底倾斜了， 防御方需要做到完美无缺， 攻击方只要找到一个突破口,再用AI放大一万倍， 模型泄露就等于是核武器扩散， 今天是Mythos打GitHub， 明天可能就是任何一个顶级模型,打任何一个公司， GitHub说目前没有客户数据泄露， 但这句话前面加了一个currently， 我觉得稳妥起见还是立刻备份你的核心仓库，感觉这不像是演习啊😲 #GitHub #网络安全 #AI

译近日，GitHub疑似遭黑客利用Anthropic的Mythos安全AI模型突破防线，窃取约4000个核心内部仓库，包括Copilot源码和CodeQL算法。此事件被视为AI网络战的开端，彻底改变了攻防平衡：攻击方可借助AI模型将漏洞无限放大，而防御方则需依赖更强大的AI防守。这意味着顶级AI模型一旦泄露，危害堪比核扩散。GitHub官方已确认正在调查内部仓库未授权访问，目前暂未发现客户数据受影响。

ginobefun@hongming731 · 5月20日56

#BestBlogs 早报 2026-05-20 ▎ 如果只看一天 AI 新闻，今天这份够用： ▎ Google 把过去一年押在「智能体」上 —— Antigravity CLI 让一个开发者就能并发调度多个子智能体；WebMCP 把浏览器变成 agent 可消费的能力底座。 ▎ Karpathy 重返一线研发选了 Anthropic，比任何一篇技术博客都更像「下一阶段路线背书」。 ▎ 加上 Claude 的自托管沙箱、Spotify 生成式推荐引擎、Simon Willison 半年回顾…… ▎ 信号密度很高的一天，推荐阅读。

译2026年5月20日，AI领域释放密集信号。Google通过Antigravity CLI和WebMCP强化智能体生态；Andrej Karpathy加入Anthropic，被视为对其发展路径的背书。此外，Claude自托管沙箱、Spotify生成式推荐引擎等进展，表明AI正从基础模型竞争转向具体应用、工具和基础设施构建。

StepFun@StepFun_ai · 5月20日68

Welcome, another 3.5 Flash 👋

译Google推出Gemini 3.5 Flash模型，现已向所有用户开放，并集成到其产品与API中。相比Gemini 3.1 Pro，3.5 Flash在几乎所有基准测试中均有提升，编程能力进步尤为显著。该模型性能媲美顶尖模型，但生成速度极快（比其他前沿模型快4倍），在“智能与输出速度”的评估维度中位于顶尖象限，实现了效率与效果的卓越平衡。

NotebookLM@NotebookLM · 5月20日40

We're currently experiencing some system issues 🛠️ If you are on a paid plan, you might notice temporary restrictions on your usage and limits. We're working to get this resolved and bring your quotas back to normal ASAP! Thank you so much for your patience and understanding 🙏

译我们目前正经历一些系统问题 🛠️ 如果您使用的是付费计划，可能会注意到使用量和限制暂时受到约束。我们正在努力解决此问题，并尽快恢复您的配额！ 非常感谢您的耐心和理解 🙏

Replit ⠕@Replit · 5月20日56

We're proud to be named to CNBC's annual Disruptor 50 list of the most innovative private companies for the first time. A decade of work has brought us here, and we're accelerating into the next.

译我们很自豪首次入选CNBC年度最具创新力50强私企榜单。 十年耕耘让我们走到今天，我们正加速迈向下一个十年。

Rohan Paul@rohanpaul_ai · 5月20日81

Andrej Karpathy had already talked about this possibility 1 month back, that he might join one of the big AI labs and why he may do so. "They (the AI labs) are kind of like at the edge of capability of what's possible. And they are working on what's coming down the line. And I think if you're outside of that frontier lab, your judgment fundamentally will start to drift because you're not part of what's coming down the line. And so I feel like my judgment will inevitably start to drift as well. And I won't actually have an understanding of how these systems actually work under the hood. It's a big system. I won't have a good understanding of how to develop and etc. And so I do think that in that sense, I agree and something I'm nervous about. I think it's worth basically being in touch with what's actually happening and actually being in the frontier lab. And if some of the frontier labs would have me come for some amount of time and do really good work for them, then I think that's maybe a good setup" --- From 'No Priors: AI, Machine Learning, Tech, & Startups" YT channel (link in comment)

译Andrej Karpathy在一个月前的访谈中解释了考虑加入大型AI实验室的原因。他指出，身处能力前沿的实验室正主导着技术发展方向；脱离其中会导致个人判断力逐渐偏离真实进展，且无法深入理解大型系统底层运作机制。因此，他认同保持接触技术前沿、身处核心研发环境的价值。近期，Karpathy已正式宣布加入Anthropic，他认为未来几年大语言模型的前沿发展将至关重要，并期待重返研发一线。

meng shao@shao__meng · 5月20日81

啊？ Andrej Karpathy 加入 Anthropic 了？！

Orange AI@oran_ge · 5月20日56

看 Google IO 的的感受是这样的： 前半段，Google 太强了，真的什么都做了，从模型到应用，牛逼 后半段，Google 太乱了，怎么什么都做三遍，功能都一样，累了

译看 Google IO 的感受是这样的： 前半段，Google 太强了，真的什么都做了，从模型到应用，牛逼 后半段，Google 太乱了，怎么什么都做三遍，功能都一样，累了

Rohan Paul@rohanpaul_ai · 5月20日66

Looks like the AI coworker category is on fire. Accel just led a $75M Series A in Viktor. The first AI coworker native to Slack and Microsoft Teams. Viktor sits inside team channels, connects to 3,000+ tools, runs scheduled tasks, keeps memory through reusable Skills, and produces work such as reports, updates, data pulls, drafts, and operational follow-through. A major shift is happening, the interface is moving from “ask a model” to “assign a role,” where the system needs memory, tool access, scheduling, permission handling, and enough task judgment to reduce supervision.

译AI同事Viktor获得Accel领投的7500万美元A轮融资，成为首个原生集成于Slack与Microsoft Teams的AI员工。它能驻留团队频道，连接3000余种工具，执行定时任务，并通过可复用技能保持记忆，从而自主生成报告、更新等实际工作产出。此次融资标志着AI交互范式正从“向模型提问”向“指派AI角色”转变，系统需具备记忆、调度与任务判断能力。据披露，Viktor在上线10周内便达成了1500万美元年化收入，其用户企业实现了团队效率提升、成本显著削减与业务增长，展现了其作为能创造真实价值的“AI员工”的巨大潜力。

Chubby♨️@kimmonismus · 5月20日83

Andrej Karpathy joining Anthropic is the biggest flex Anthropic could have hoped for. 10m views after 6h. Best hire ever.

译Andrej Karpathy加入Anthropic，是Anthropic所能期望的最大手笔。 6小时内获得1000万次浏览。史上最重磅招聘。

Josh Woodward@joshwoodward · 5月20日20

Glad you like it!

译很高兴你喜欢！ [引用 @mweinbach]：我认为 Google 在新的 Gemini 应用上做得非常出色。 他们同时采用了来自 Android（Material You）和 iOS（Liquid Glass）的自适应原生设计系统，并且仍然保持了 Google 整体设计的独特性。 他们现在做到了我所见过的所有应用中最好的。

Google AI@GoogleAI · 5月20日33

A few weeks ago, we asked our community to use @GoogleAIStudio or Canvas in @GeminiApp to help us create the Google I/O countdown. Thanks SO much to everyone who submitted, and special shoutout to the creators whose submissions helped us set the right ~vibes~ on the stage today: @thgie_8_eight @r_mevo @EdinsonNM @gunny_5 @xabila @GuyMosheTamam @mikkelhundborg @rajatxr Shahab Uddin Charles Zuo

译几周前，我们邀请社区使用 @GoogleAIStudio 或 @GeminiApp 中的 Canvas 功能，帮助我们创作 Google I/O 倒计时动画。 衷心感谢所有提交作品的用户，特别鸣谢以下创作者，他们的作品为今天的舞台营造了恰到好处的氛围： @thgie_8_eight @r_mevo @EdinsonNM @gunny_5 @xabila @GuyMosheTamam @mikkelhundborg @rajatxr Shahab Uddin Charles Zuo

Replit ⠕@Replit · 5月20日37

Sneaky interventions, interactive installations, tools built for other artists to use. @kcimc's practice lives where machine learning, computer vision, and social technology collide. See his new interactive installation on day one of Vibecon. NYC, June 17–18. Get your tickets at http://vibecon.ai

译隐秘的干预、互动装置、为其他艺术家打造的工具。@kcimc 的实践存在于机器学习、计算机视觉与社会技术的交汇处。 在 Vibecon 首日观看他的新互动装置。 纽约，6月17-18日。 购票请访问 http://vibecon.ai

Artificial Analysis@ArtificialAnlys · 5月20日37

Congratulations Cerebras on going public last week! Artificial Analysis benchmarks were cited in Cerebras' S-1 filing regarding inference performance. We have benchmarked Cerebras’ serverless API since the day it launched in August 2024. Since then, Cerebras has consistently offered the fastest output speeds for supported models in Artificial Analysis benchmarks. Congratulations @andrewdfeldman and the whole @cerebras team.

译祝贺 Cerebras 上周成功上市！ Artificial Analysis 基准测试在 Cerebras 的 S-1 文件中被引用，涉及推理性能。自 2024 年 8 月 Cerebras 无服务器 API 上线以来，我们一直对其进行基准测试。此后，在 Artificial Analysis 基准测试中，Cerebras 在支持的模型上始终提供最快的输出速度。 祝贺 @andrewdfeldman 和整个 @cerebras 团队。

SemiAnalysis@SemiAnalysis_ · 5月20日67

Warren Buffett's Berkshire Hathaway first invested in Google in Q3 2025, coincidentally the same time that SemiAnalysis called out a huge increase in TPU purchases from Anthropic. In Q1 Berkshire added more to their position. Buffett actually said this about his own diligence into Google's AI Infrastructure supremacy: "I don't buy what I can't understand. So Greg sat me down with the TPU v5p spec. 8,960 chips wired in a 3D torus, every chip talking to six neighbors over ICI links at 4.8 Tbps a pop, wraparound rails so no chip is ever at the end of the line, and optical circuit switches throwing the junctions mid-job to carve out whatever submesh your sharded matmul needs. Add in ring all-reduce running consists in both directions along each torus axis, collective permutes shuffling shards between sidings, and bandwidth-optimal SPMD partitioning across the data, model, and pipeline dimensions. Folks, it's just BNSF. Six neighbors, scheduled consists, a yardmaster throwing switches, trains that never stop. And I've been understanding railroads since 1942." Want to understand the TPU system architecture as deeply as Buffett? Read more here: https://newsletter.semianalysis.com/p/tpuv7-google-takes-a-swing-at-the

译巴菲特旗下伯克希尔·哈撒韦于2025年第三季度首次投资谷歌，并于2026年第一季度加仓。其投资决策基于对谷歌AI基础设施的深度理解。巴菲特将复杂的TPU v5p芯片集群——包括其3D环面互联、光电路交换与分布式训练算法——比作他熟悉的BNSF铁路网络：芯片互联如轨道连接，调度算法如列车编组，环形拓扑确保系统稳定。这一类比揭示了他并非盲目跟风，而是通过对底层系统运作原理的剖析做出的审慎投资。

Chubby♨️@kimmonismus · 5月20日24

Just sat down with @rmstein , Vice President of Product at @Google Search, to talk about AI Mode, the future of Search, and how generative AI is reshaping the way people discover information. Super insightful conversation. Video coming soon.

译刚刚与@Google Search产品副总裁@rmstein会面，探讨AI模式、搜索的未来，以及生成式AI如何重塑人们发现信息的方式。 对话非常有深度。 视频即将发布。

Josh Woodward@joshwoodward · 5月20日13

Glad you like it! We have more accents coming! :)

译很高兴你喜欢！我们即将推出更多口音！:)