Greg Brockman@gdb · 5月29日61

How @CGRTeams is working with @OpenAI to improve motorsports performance:

译CGRTeams如何与OpenAI合作提升赛车运动表现：

Berryxia.AI@berryxia · 5月29日25

我们不八卦了，来点干的。 Claude Opus 4.8 VS ChatGPT 5.6 发售在即。 受挫期待住了，看看能不能有惊喜啊！ 周末又要熬了~~~

译我们不八卦了，来点干的。 Claude Opus 4.8 VS ChatGPT 5.6 发售在即。 受挫期待住了，看看能不能有惊喜啊！ 周末又要熬了~~~

Xiaomi MiMo@XiaomiMiMo · 5月28日59

Today we're excited to announce the first 13 ecosystem partners joining the Xiaomi MiMo Orbit Program. A sincere thank you to every partner for your trust and collaboration. @gitlawb and 12 other ecosystem partners. This is just the beginning. MiMo Orbit Program is still open, and we'd love to collaborate with you! → business-mimo@xiaomi.com

译今天我们很高兴地宣布，首批13家生态伙伴加入小米MiMo Orbit计划。 衷心感谢每一位伙伴的信任与合作。@gitlawb 及其他12家生态伙伴。 这仅仅是开始。MiMo Orbit计划仍在开放中，我们期待与您合作！ → business-mimo@xiaomi.com

Rohan Paul@rohanpaul_ai · 5月28日66

Morgan Stanley now says European banks could remove up to 20% of jobs as AI takes over repeatable banking work. The bank expects 30% productivity gains, meaning fewer people may be needed for tasks like compliance checks, risk monitoring, KYC reviews, AML screening, reporting, and other back-office work. The forecast has doubled from 10% to 20%, or roughly 200K to 400K possible roles by 2030, because banks are moving from AI pilots to real restructuring plans. Most cuts would likely come through retirements, attrition, and managed exits, because European labour rules make sudden mass layoffs harder than in the US. The workforce will not just shrink, it will change, with fewer traditional processors and more data engineers, AI operators, and model-risk specialists. --- thenextweb. com/news/european-banks-20-percent-jobs-ai-morgan-stanley

译摩根士丹利最新预测，AI技术可能促使欧洲银行裁减高达20%的工作岗位，较此前预期的10%翻倍。该行预计AI能带来30%的生产力提升，主要影响合规审查、风险监控、KYC审核等后台重复性工作。预测显示，到2030年，受影响的岗位可能从20万增至约40万，原因是从试点转向实质性重组。变革将主要通过自然减员实现，并催生对数据工程师、AI操作员等新角色的需求。

Rohan Paul@rohanpaul_ai · 5月28日59

China has started giving every humanoid robots 29-character ID codes, with 28,000+ assigned so far. The system works like a robot passport: each unit gets a code that records its country link, maker, model type, and individual serial identity. The goal is not only paperwork, because humanoids are moving from demos into public roads, factories, homes, and service jobs where failures need clear responsibility. A robot that falls, damages property, leaks data, or gets modified after sale becomes easier to trace when regulators can link the machine to its maker, seller, user, service history, and recycler. --- scmp .com/tech/policy/article/3354747/china-give-every-humanoid-robot-digital-id-push-boost-industry-standards

译中国开始为人形机器人分配29位的身份代码，已发放超过2.8万个。该系统类似机器人护照，记录制造国、厂商、型号及唯一序列号。主要目的是明确责任归属，因为人形机器人正从演示进入工厂、家庭等真实场景，需要清晰追责。当机器人发生故障、损坏财产、泄露数据或被改装时，监管方可借此追溯至制造商、销售商、用户、服务历史及回收商。

Rohan Paul@rohanpaul_ai · 5月28日65

Elon Musk just told investors that SpaceX’s Anthropic AI compute deal is not a locked multi-year rental, but a 180-day lease for Colossus with a 90-day cancellation path. The older reading made the deal look like $1.25B/month through May-29, but Musk says SpaceX wanted the short term because AI compute may become too scarce to rent away for years. SpaceX wants flexibility because Colossus is not just a side asset, since the same compute infra trains xAI models, support internal AI systems, or become a paid cloud-style business. --- reuters .com/technology/musk-says-spacex-did-not-commit-long-term-colossus-lease-with-anthropic-2026-05-28/

译Elon Musk向投资者澄清，SpaceX为Anthropic提供AI算力的Colossus并非长期锁定租赁，而是一份为期180天的租约，并附带90天的取消路径。此前外界认为该交易价值约每月12.5亿美元并持续至2029年5月，但Musk解释称，SpaceX选择短期条款是因为考虑到AI算力未来可能变得稀缺，不宜长期外租。他强调，Colossus并非闲置资产，同一套计算基础设施将用于训练xAI模型、支持内部AI系统，或可能在未来发展为付费云服务，因此SpaceX需要保持运营灵活性。

Rohan Paul@rohanpaul_ai · 5月28日57

China’s humanoid robot race has moved from lab demos to real shipment. Global humanoid shipments grew nearly 800% in 2025, China now has 140 humanoid robot makers, & 330 new models launched in just 12 months (per IDC Global Humanoid Robot Market Analysis) AGIBOT ranked #1 globally by humanoid robot shipments in 2025, with about 5,200 units and roughly 39% global share.

译中国人形机器人竞赛已从实验室演示转向实际交付。 全球人形机器人出货量在2025年增长近800%，中国目前有140家人形机器人制造商，仅12个月内就推出了330款新型号（据IDC全球人形机器人市场分析）。 AGIBOT在2025年按人形机器人出货量排名全球第一，出货约5,200台，全球市场份额约39%。

Rohan Paul@rohanpaul_ai · 5月28日59

NVIDIA published a report on Vera CPU benchmarks, done by Phoronix. Compares Vera directly against a current 128-core x86 CPU and claims a 1.5x overall performance advantage Compared with the prior-generation NVIDIA Grace CPU, Vera delivered a 1.6x geometric mean increase in Phoronix’s testing. Vera delivered over 4x the memory bandwidth per core compared with traditional x86 CPUs. Vera delivers 1.2TB/s bandwidth using LPDDR5X, while keeping memory power under 30W, compared with more than 100W for many DDR5 server setups. ---- To note, Vera uses Armv9.2, not x86, so NVIDIA is basically saying that their Arm-based CPU can beat the usual Intel and AMD server CPUs. For agentic AI, this CPU-side work becomes much heavier because the AI is not only generating text, but also calling tools, reading files, writing code, using browsers, running sandboxes, and managing workflows.

译NVIDIA发布Vera CPU基准测试报告。Vera采用Armv9.2架构，在Phoronix测试中，其整体性能比128核x86 CPU高1.5倍，比前代Grace CPU提升1.6倍（几何平均）。其每核心内存带宽是传统x86 CPU的4倍以上，使用LPDDR5X实现1.2TB/s带宽，内存功耗低于30W。该报告旨在表明NVIDIA的Arm架构CPU性能已超越Intel和AMD的x86服务器CPU，并强调在智能体AI场景下，因涉及工具调用、文件读写、代码生成等复杂任务，CPU侧工作负载变得更重。

Rohan Paul@rohanpaul_ai · 5月28日68

Jensen Huang just moved Nvidia closer to China’s elite business and policy network by accepting the coveted Tsinghua University advisory role while Washington is blocking Nvidia’s top AI chips from China. The role is on Tsinghua SEM’s 65-member advisory board, chaired by Apple’s Tim Cook and packed with global CEOs from tech, finance, and industry. Gives Huang access to a rare forum where Chinese policymakers, academics, and corporate leaders still meet Western executives. Nvidia may not be able to ship its best silicon to China, but it does not want to lose the relationships that shape future demand. --- firstpost .com/world/nvidia-chief-jensen-huang-joins-advisory-board-of-chinas-tsinghua-university-14016101.html

译英伟达CEO黄仁勋接受邀请，加入清华大学经管学院顾问委员会。该委员会由苹果CEO库克担任主席，拥有65名成员，汇集了科技、金融和工业界的全球CEO。此举发生在美国限制英伟达向中国出口顶尖AI芯片的背景下，旨在帮助英伟达维持与中国政策制定者、学术界和企业领袖的联系，以参与塑造未来的市场需求。

X.PIN@thexpin · 5月28日77

Exclusive: DeepSeek plans to file for a STAR Market (A-share) IPO immediately after closing its current ~$50B (¥350B) funding round. Source: a large fund manager involved in this round.

译独家：DeepSeek计划在完成当前约500亿美元（3500亿人民币）融资轮后，立即申请科创板（A股）IPO。 来源：参与本轮融资的一位大型基金经理。

Alibaba Cloud@alibaba_cloud · 5月28日73

Qwen3.7-Max hits #1 on the @OpenRouter Trending LLM chart with 77.3B tokens in usage. And we are just getting started. 👇 https://int.alibabacloud.com/m/1000413314/

译Qwen3.7-Max 以 77.3B tokens 的使用量登顶 @OpenRouter 热门大语言模型榜单。 而我们才刚刚开始。 👇 https://int.alibabacloud.com/m/1000413314/

Alibaba Cloud@alibaba_cloud · 5月28日62

📢Qwen3.7-Max just hit #3 on ITbench-AA — a fresh benchmark testing how well models handle real-world enterprise IT tasks, agentic-style. 🔧Agentic era, go with Qwen.🏃🏃 API: https://int.alibabacloud.com/m/1000413314/

译通义千问（Qwen）团队宣布，其Qwen3.7-Max模型在新兴的ITBench-AA基准测试中位列第三。该测试由Artificial Analysis与IBM Research合作推出，旨在评估模型解决真实企业IT任务的能力，当前聚焦于站点可靠性工程（SRE）领域。测试包含59个Kubernetes故障诊断任务。结果显示，Claude Opus 4.7以47%的得分排名第一，GPT-5.5（xhigh）以46%紧随其后，Qwen3.7-Max以42%排名第三。所有前沿模型得分均低于50%，表明该测试具有较高挑战性。

Kling AI@Kling_ai · 5月28日62

Kling AI is taking the stage at AI on the Lot's Community Day - the world’s largest AI film & media conference to proudly present 20 original short films from the Prompt Club's filmmakers, each pushing the boundaries of what AI cinema looks like at native 4K resolution. Community Day is where the industry comes together to not only talk about AI filmmaking, but to actually watch it on the big screen. Kling AI's showcase is in incredible company with the first NA theatrical releases of Gossip Goblin’s The Patch Wright and Higgsfield’s Hell Grind as well as Emmy Awards Winner Bernie Su’s HOLYWATER: Liminality. Join us on May 29 at 1PM at The Culver Theater, 9500 Culver Blvd, Culver City, California.

译可灵AI将在全球最大AI影视会议AI on the Lot的社区日上，展示由Prompt Club的电影制作人创作的20部原创AI短片。所有短片均为原生4K分辨率，旨在探索AI电影的边界。该展示将于5月29日在加州卡尔弗市的卡尔弗剧院举行。

Alibaba Cloud@alibaba_cloud · 5月28日59

📢Qwen3.7-Max just hit #3 on ITbench-AA — a fresh benchmark testing how well models handle real-world enterprise IT tasks, agentic-style. 🔧Agentic era, go with Qwen.🏃🏃

译由 Artificial Analysis 和 IBM Research 合作推出的首个评估模型处理真实企业IT任务能力的基准测试 ITBench-AA，聚焦于站点可靠性工程（SRE）任务。测试结果显示，通义千问（Qwen3.7-Max）以 42% 的分数排名第三。该测试中，所有前沿模型得分均低于 50%，其中 Claude Opus 4.7 以 47% 领先，GPT-5.5（xhigh）以 46% 紧随其后。在开源模型中，GLM-5.1（Reasoning）以 40% 领衔。该基准未来将扩展到财务运营（FinOps）等任务。

Deedy@deedydas · 5月28日53

This is painfully embarrassing. The national board of education in India just generated an image on ChatGPT to “prove” that they’re secure after a 19yo showed you can edit marks of 2M test takers on their platform. That is after trying to deny they got hacked using a domain they didn’t own (to prove it, they bought by the domain and redirected it to the blog post about the hack). Then they claimed (falsely) that no production data was on those servers that was immediately proven false. These are levels of incompetence we have never seen. The fate of 2M+ students a year rests in these hands. Pathetic.

译印度中央教育委员会（CBSE）在曝出重大数据安全事件后，其应对措施备受批评。一名19岁学生发现其平台存在漏洞，可篡改200万名考生的成绩。官方后续试图用一个不属于他们的域名来否认被黑客攻击（后为此购买该域名），并虚假声称服务器上没有生产数据，但被立即证伪。最终，他们用ChatGPT生成了一张图片，试图以此“证明”系统安全。该委员会每年负责超过200万名学生的命运。

Fei-Fei Li@drfeifei · 5月28日58

It’s a real honor to receive an honorary doctorate of science from @BrownUniversity . 😍

译很荣幸获得布朗大学颁发的荣誉理学博士学位。😍

Alibaba Cloud@alibaba_cloud · 5月28日31

Delighted to invite you to the AI Film Festival Monaco 2026 with Alibaba Cloud: ⛱ June 9, Pre-Event: AI Hackathon and Beach Networking. 🎥 June 10, Official Event: Keynote, Panel, Jury Talks and Awards. From strategic dialogues to AI filmmaking in practice with WAN, HappyHorse, and HappyOyster, we are pushing the boundaries of what’s possible in digital storytelling. The future of production starts in Monaco. 🌍🚀 🎟 Register now to secure your spot: https://int.alibabacloud.com/m/1000412352/ (Or scan the QR code in the poster below) #AlibabaCloud #AIFilmFestival #WAN #HappyHorse #HappyOyster #Monaco2026 #AIArt #FilmTech #CreativeAI #AInnovation

译阿里巴巴云宣布与AI Film Festival Monaco 2026合作。活动将于6月9日举行AI Hackathon和沙滩社交预热，6月10日举办正式活动，包含主题演讲、专家小组讨论及颁奖典礼。活动将围绕WAN、HappyHorse、HappyOyster等AI工具在电影制作中的实践应用展开，旨在探索数字叙事的更多可能。

Alibaba Cloud@alibaba_cloud · 5月28日21

⏰ 48 HOURS LEFT. #DiveIntoYourWorld closes May 30. Your last chance to: ✦ Create something extraordinary with AI ✦ Win tickets to the World Swimming Championships ✦ Get invited to MuleRun's Global User Conference One idea. MuleRun does the rest. → https://int.alibabacloud.com/m/1000412352/ @WorldAquatics @alibaba_cloud @mulerun_ai

译⏰ 还剩48小时。 #DiveIntoYourWorld 活动将于5月30日截止。 你最后的机会： ✦ 用AI创作非凡作品 ✦ 赢取世界游泳锦标赛门票 ✦ 获邀参加MuleRun全球用户大会 一个创意，MuleRun完成其余。 → https://int.alibabacloud.com/m/1000412352/ @WorldAquatics @alibaba_cloud @mulerun_ai

Alibaba Cloud@alibaba_cloud · 5月28日24

Join the AInnovation Summit at VivaTech! Catch our opening speech & 3 fireside chats to unpack global strategic perspectives, open innovations, and full-stack agentic AI engines. 📅 June 17, 15:00 - 17:00 CEST 📍 Paris Expo Porte de Versailles | Hall 7.3 Workshop A

译参加 VivaTech 的 AInnovation Summit！聆听我们的开幕演讲和 3 场炉边对话，深入探讨全球战略视角、开放创新和全栈智能体 AI 引擎。 📅 6月17日，15:00 - 17:00 CEST 📍 巴黎凡尔赛门展览馆 | 7.3号馆 Workshop A

Greg Brockman@gdb · 5月28日33

please report any ChatGPT bugs in the thread below — team (and codex) working super hard to resolve them:

译请在下方线程中报告任何 ChatGPT 漏洞——团队（和 Codex）正在全力解决：

Rohan Paul@rohanpaul_ai · 5月28日45

Brett Adcock, CEO of Figure AI: "we're working until midnight every night... we are here every weekend. By end of 2026, we'll be able to put a robot into home and be able to do fairly long horizon work."

译Figure AI CEO Brett Adcock："我们每晚工作到午夜……每个周末都在这里。 到2026年底，我们将能够把机器人送入家庭，并能完成相当长时间跨度的工作。"

Rohan Paul@rohanpaul_ai · 5月28日77

"There's about 30-35mn software engineers in the world today. We wanna make all of them 10 times more efficient, and then we think there is a lot more than 10 times more software to build." ~ Cognition CEO Scott Wu Talking about their $1 Bn fund raise today. Their revenue climbed from $37M to ~$500M in 1 Year. ---- From "Bloomberg Technology" YouTube channel, (link in comment)

译Cognition AI完成超10亿美元融资，投前估值达260亿美元。其年化收入一年内从3700万美元增长至约4.92亿美元。其核心产品Devin定位为可自主工作的初级工程师，超越传统代码助手，能通过多步骤工作流进行规划、测试和部署。Cognition采用模型无关策略，结合自身模型与OpenAI、Anthropic的大语言模型。CEO Scott Wu表示，目标是提升全球约3000-3500万软件工程师的效率10倍。

Google AI Developers@googleaidevs · 5月28日24

We love a good dogfooding session! 🐶 Before bringing these updates to you, we asked Googlers to try building @AndroidDev apps in @GoogleAIStudio. Here are some of our faves…

译我们喜欢内部测试！🐶 在向大家推出这些更新之前，我们让Googlers在@GoogleAIStudio中尝试构建@AndroidDev应用。 以下是我们的一些最爱…

Rohan Paul@rohanpaul_ai · 5月28日63

Today’s edition of my newsletter just went out. 🔗 https://www.rohan-paul.com/p/chinas-huawei-reveals-a-new-chip 🗞️ China’s Huawei reveals a new chip design breakthrough which can close its gap with TSMC and Intel 🗞️ New Alibaba + Nanjing Univ paper shows standard LLMs can handle very long context faster by making attention selectively sparse. 🗞️ Deep Dive on DeepSeek: Their real story is not cheaper chatbots, but architecture that turns hardware scarcity into strategy. 🗞️ New Meta + Stanford + Illinois survey paper argues that AI agents work better when code becomes their main working layer. 🗞️ Anthropic billionaire cofounder backs Pope Leo, warning that AI job losses will create a historic moral crisis 🗞️ xAI just Dropped ‘Grok Build’: The Terminal-Native Agentic AI for all all SuperGrok and X Premium+ users.

译华为披露新的芯片设计突破，旨在缩小与台积电及英特尔的差距。阿里巴巴与南京大学的论文提出，标准大语言模型可通过选择性稀疏注意力机制更高效地处理长上下文。对DeepSeek的深度分析指出，其核心价值并非提供更廉价的聊天机器人，而是将硬件稀缺性转化为战略优势的架构设计。Meta、斯坦福及伊利诺伊大学的调查论文主张，当代码成为AI智能体的主要工作层时，其效能会更高。Anthropic联合创始人警示AI导致的失业将引发历史性道德危机。xAI为SuperGrok与X Premium+用户推出了终端原生的智能体AI产品“Grok Build”。

Replit ⠕@Replit · 5月28日30

Co-founder and Head of Design at Replit, @HayaOdeh shapes the brand, user experience, and product design of one of the world’s most accessible vibe coding platforms. See her take the stage on day two of Vibecon. NYC, June 17–18. Get your tickets at http://vibecon.ai

译Replit联合创始人兼设计负责人@HayaOdeh塑造了全球最易用的vibe编码平台之一的品牌、用户体验和产品设计。 请于6月17-18日在纽约观看她在Vibecon第二天的演讲。 购票请访问 http://vibecon.ai

swyx@swyx · 5月28日75

cognition is now the largest independent agent lab in the world. take the 200% utilization that everyone is hitting from this chart and run out the sales growth from this, i encourage you to go thru the exercise if you are new to investing a lot of you have read my cog initation post, but when i talk to people they do not sufficiently appreciate that this is the perfect storm of all the trends: - first* koding agent - first* cloud dev infra - best reviewed code review/security guy - first* llm wiki knowledge base - s-tier GTM - most IOI golds - somehow also cracked at Smash Bros and Poker agent lab gets you: - long model diversity - long reasoning/toolcalling models - long harnesses - long domain specific RLFT - long coding data/expertise/evals - long full agentic SDLC - chief partner to CIOs in combating tokenmaxxing slop from humans and from agents - SOTA foodbench from house chefs and this is why you can see this is Peter Thiel's biggest AI bet *i dont like "first" as an adjective, but "first + 2 years of serious scaling" really serves as a shorthand for "most enterprise battle tested" - see customer list in screenshot - literally 10s of thousands of devs, 10s of thousands of repos, PER CUSTOMER is the kind of weight I put on the simple word "first"

译Cognition宣布已成为全球最大的独立智能体实验室。公司完成超10亿美元融资，估值达260亿美元，由Lux Capital、General Catalyst等领投。其企业使用量自年初增长超10倍，年化收入增至4.92亿美元。Cognition于两年前推出Devin，定位为首个AI软件工程师。公司强调其拥有多项领先优势，包括首个编码智能体、顶级代码审查能力等，并得到了Peter Thiel的重大投资。

小互@xiaohu · 5月28日63

OpenRouter 完成了 1.13 亿美元 B 轮融资 在过去 6 个月里，OpenRouter 的每周交易量从 5T Tokens 增长到 25T Tokens... Token 专卖看来是个大生意... 中转的也赚的很厉害

译OpenRouter 完成了 1.13 亿美元 B 轮融资 在过去 6 个月里，OpenRouter 的每周交易量从 5T Tokens 增长到 25T Tokens... Token 专卖看来是个大生意... 中转的也赚的很厉害

Rohan Paul@rohanpaul_ai · 5月28日69

OpenAI’s controlling nonprofit just committed $250M to study and soften the job-market shock from AI systems that can replace paid tasks. The announcement details the three focus areas: - Understanding AI’s economic impacts through independent research and measurement. - Supporting workers and communities facing near-term disruption (e.g., job transition support, insurance mechanisms, retraining). - Exploring long-term approaches to sharing economic gains from AI more broadly. OpenAI Foundation is unusually powerful because it owns 26% of OpenAI’s for-profit arm, a stake that could be valued at $260B (Assuming OpenAI is valued at $1T).

译OpenAI的控股非营利机构OpenAI Foundation承诺投入2.5亿美元，用于研究和缓解能替代付费任务的AI系统对就业市场造成的冲击。资金将支持三个方向：通过独立研究衡量AI的经济影响、为面临近期冲击的工人与社区提供支持（如职业过渡、保险机制、再培训），以及探索更广泛分享AI经济收益的长期途径。该基金会因拥有OpenAI营利部门26%的股份而地位特殊，按OpenAI估值1万亿美元计算，该股份价值可达2600亿美元。

Replit ⠕@Replit · 5月28日60

Replit has been named to @Redpoint's 2026 InfraRed 100. A list of the companies building what AI runs on. Grateful to be in it, alongside alumni like Stripe, Snowflake, and HashiCorp. To every builder shipping on Replit: this one's yours. https://www.redpoint.com/infrared/report/

译Replit 被列入 @Redpoint 的 2026 InfraRed 100 榜单。 这是一份构建 AI 运行基础设施的公司名单。很荣幸能入选，与 Stripe、Snowflake 和 HashiCorp 等往届入选者并列。 致每一位在 Replit 上发布产品的构建者：这份荣誉属于你们。 https://www.redpoint.com/infrared/report/

Rohan Paul@rohanpaul_ai · 5月28日71

Another great win for agentic coding. Cognition AI just raised over $1B at a $26B pre-money valuation. Revenue reportedly climbed from $37M in annualized run-rate to $492M, while customers like Goldman Sachs and Mercedes-Benz suggest Devin is moving from demo rooms into production workflows. Cognition's progress is driven by its flagship product, Devin, which aims to function as an autonomous junior engineer, going beyond typical coding assistants. Devin can plan, test, and deploy code through multi-step workflows in secure environments. Cognition combines its own models with OpenAI and Anthropic rather than relying on one model. Cognition is basically pitching Devin as a model-agnostic agent layer: the LLM does the reasoning and code generation, while Devin supplies the engineering workspace, repo context, terminal access, file edits, tests, and model choice around it. Last July, Cognition agreed to buy the remains of coding startup Windsurf after Google struck a $2.4 billion deal for Windsurf’s top talent and licensing rights.

译Cognition AI完成超10亿美元融资，投前估值达260亿美元。其年化收入从3700万美元增长至4.92亿美元，客户包括Goldman Sachs和Mercedes-Benz，标志着其产品Devin正进入生产环境。Devin定位为自主初级工程师，能通过多步骤工作流规划、测试和部署代码。Cognition采用自有模型与OpenAI、Anthropic相结合的模型无关技术路线，而非依赖单一模型。此外，该公司于去年7月同意收购编程初创公司Windsurf的剩余资产。

Rohan Paul@rohanpaul_ai · 5月28日79

Jensen Huang presents new Taiwan campus. Nvidia plans to invest around $150 bn a year in Taiwan, This is only after a week when rival AMD said it would invest more than $10 bn in Taiwan's AI sector.

译黄仁勋展示了新的台湾园区。 英伟达计划每年在台湾投资约1500亿美元。 就在竞争对手AMD宣布将向台湾AI领域投资超过100亿美元一周后。

Claude@claudeai · 5月28日35

Michele Catasta (@pirroh) is President and Head of AI @replit, the platform where anyone can build software in natural language. At 16, he set out to make software open to everyone. Today, over 50 million people are building on Replit with Claude:

译Michele Catasta (@pirroh) 是 Replit 的总裁兼 AI 负责人，该平台让任何人都能用自然语言构建软件。 16岁时，他立志让软件向所有人开放。如今，超过5000万人正在 Replit 上使用 Claude 进行构建：

Nathan Lambert@natolambert · 5月28日51

The most likely way continual learning manifests in the coming few years is through products used directly for knowledge work. Sort of how cursor can continually train their models with real-world data and RL, Claude, Copilot, and co will see if they can for knowledge work. I was chatting with Ronak a few weeks ago when this was crystalizing for me, so it's fun to see a startup in that area.

译创业公司Trajectory宣布成立，并获得Conviction、Bessemer、Radical Ventures、Jeff Dean及李飞飞等投资的1500万美元融资。该公司旨在构建持续学习平台，利用产品使用数据中的信号，帮助企业对大规模智能体模型进行持续后训练，使其性能超越前沿模型。目前，Trajectory已与Harvey、Decagon AI、Mercor、Rogo AI等AI原生公司建立合作，部分已进入生产环境。团队成员来自DeepMind、OpenAI、Apple、Meta Superintelligence等顶级机构。其理念是AI产品将通过每一次用户交互（如修正、重试、编辑）持续变得更智能。

Sam Altman@sama · 5月28日61

AI should dramatically increase quality of life and individual freedoms for people around the world. The OpenAI Foundation is making an initial $250M commitment to measurement, transition support, and new approaches to broadly shared prosperity. http://openaifoundation.org/news/economic-futures-in-the-age-of-ai

译AI应显著提升全球人民的生活质量和个体自由。 OpenAI基金会初步承诺投入2.5亿美元，用于衡量、转型支持以及广泛共享繁荣的新方法。 http://openaifoundation.org/news/economic-futures-in-the-age-of-ai

OpenRouter@OpenRouter · 5月28日54

Honored to be included in @Redpoint’s 2026 InfraRed 100! https://www.redpoint.com/infrared/report/

译很荣幸入选 @Redpoint 的 2026 InfraRed 100 榜单！

Suno@suno · 5月28日42

See you in LA!!!

译洛杉矶见！！！

Runway@runwayml · 5月28日51

Join us in New York for a pre-show conversation with special guest, Ron Howard. Followed by our 10 finalist films and the Best of Advertising, Design, New Media, Gaming and Fashion. Tickets available at the link below.

译欢迎加入我们在纽约的活动，与特邀嘉宾罗恩·霍华德进行展前对话。随后将展映我们的10部入围影片，并颁发广告、设计、新媒体、游戏和时尚领域的最佳奖项。 门票可通过下方链接购买。

Tibo@thsottiaux · 5月27日55

Team is investigating elevated latencies across the board for API and ChatGPT. Updates will be posted on https://status.openai.com/

译团队正在调查API和ChatGPT整体延迟升高的问题。更新将发布在 https://status.openai.com/

Rohan Paul@rohanpaul_ai · 5月27日40

Huawei's new breakthrough "LogicFolding". To close the gap with TSMC and Intel without relying only on smaller transistors, by making chip signals travel less distance. Their main ideas is that chip progress should now be measured by time saved, not just space shrunk. What they call "τ scaling" - changes the question from “how small is the transistor?” to “where is time being lost?” LogicFolding is Huawei’s physical answer to that question inside a chip. LogicFolding partitions digital, analog, and memory circuits across vertically stacked active tiers, so the chip behaves less like a city spread across a plain and more like a compact building with elevators between floors. Because chip delay is often not just the transistor switching time, but the resistance and capacitance of wires between logic gates. By folding critical paths upward into another active layer, the design can shorten wires, reduce parasitic delay, tighten clock skew, and raise frequency without changing the device node. So LogicFolding is not simply “3D chips” as a packaging. Its deeper idea is that topology has become a scaling tool. When lithography stops giving easy gains, the next improvement may come from rearranging where logic, memory, power, and communication physically live.

译华为推出“LogicFolding”芯片设计技术，旨在不依赖单纯缩小晶体管制程，以缩小与竞争对手的差距。其核心思想是通过垂直堆叠活跃电路层，将关键信号路径“折叠”向上，从而缩短导线长度、减少寄生延迟并收紧时钟偏差，最终在不改变器件节点的情况下提升芯片频率。此技术并非传统3D封装，其深层创新在于将芯片拓扑结构本身变为一种性能缩放工具。华为指出，该技术能大幅压缩相邻触发器间的传播时间，收紧关键路径。预计其高端芯片将在2031年实现等效于14 Å（1.4 nm）工艺的晶体管密度。

Berryxia.AI@berryxia · 5月27日67

我今天刷到OpenRouter的公告。 过去半年，他们每周处理的token量从5T直接冲到25T。 这可不是实验阶段的小打小闹，是实打实的生产级使用在爆发。 企业把AI真正塞进工作流里，每天都在跑大规模任务。 他们刚完成了1.13亿美元的B轮，由CapitalG领投。 a16z、Menlo Ventures、NVIDIA的NVentures、ServiceNow、MongoDB、Snowflake、Databricks这些机构都跟进了。 这笔钱背后，是整个行业对“多模型时代基础设施”的押注。 你不用再为每个模型单独对接、比价、处理限流。 一个统一API就能切换500多个模型，包括50多个免费的，还带私有聊天和可探索数据。 增长数据把一个趋势讲得很清楚： AI的下一阶段不是谁的参数更大，而是谁能让生产环境真正稳定、高效、可控。 OpenRouter正好卡在这个节点上，把路由这件事做成了基础设施。

译OpenRouter宣布完成1.13亿美元B轮融资，由CapitalG领投，a16z、NVIDIA的NVentures等跟投。过去6个月，其平台每周处理的token量从5T增长至25T。该平台提供统一API，可切换包括50多个免费模型在内的500多个大语言模型，并提供私有聊天功能。此轮融资和增长数据反映了市场对其多模型时代基础设施价值的认可。