Rohan Paul@rohanpaul_ai · 4月29日34

On polished tiles with minimal grip, this robot dog demonstrates impressive foot placement and balance. It appears to use real-time adjustments, keeping its body level as it moves naturally despite the challenging surface.

译在抓地力极小的抛光瓷砖上，这只机器狗展现了令人印象深刻的脚步定位与平衡能力。 它似乎运用了实时调整技术，在具有挑战性的地面上自然移动时，始终保持身体水平。

Rohan Paul@rohanpaul_ai · 4月29日62

Bloomberg: Google decided to back away from a $ 100M Pentagon drone swarm contest after first making the cut, exposing how divided Big Tech still is over military AI. The project aimed to turn voice commands like directional orders into machine instructions for groups of autonomous drones. Google’s exit appears less about raw capability than about internal limits on what kind of defense work the company is willing to own. --- bloomberg. com/news/articles/2026-04-28/google-drops-out-of-pentagon-drone-swarm-contest-after-advancing

译彭博社报道，谷歌在入围后决定退出美国国防部一项价值1亿美元的无人机集群竞赛。该项目旨在将语音指令转化为对自主无人机群的机器指令。谷歌的退出并非由于技术能力不足，而更多源于公司内部对愿意承担的国防工作类型设定了限制。这一事件凸显了大型科技公司在军事人工智能应用上仍然存在深刻分歧。

宝玉@dotey · 4月29日32

帮转招聘信息

译YouMind正在招聘增长工程实习生，面向大三至研二在读、具备技术背景和前端技能的学生，要求有独立全栈项目及社媒账号基础。实习生将参与从需求挖掘到数据复盘的全链路增长项目，并搭建增长工具提升运营效率。团队由玉伯发起，氛围自由，以异步交流为主，成员多来自大厂但风格独特。福利包括宠物陪伴、免费咖啡、YouMind MAX会员以及无限制使用Claude Code/ChatGPT且不考核Token消耗。应聘者需通过𝕏联系jaredliu_bravo并附上简历。

Google DeepMind@GoogleDeepMind · 4月29日34

As AI evolves, education must keep pace. 🍎 Since 2023, the Experience AI program – developed in collaboration with @RaspberryPi_org – has offered free resources to help students and teachers understand AI and how it works. See the impact ⬇️

译随着人工智能的发展，教育必须跟上步伐。🍎 自2023年以来，与@RaspberryPi_org合作开发的Experience AI项目——提供了免费资源，帮助学生和教师理解人工智能及其工作原理。 查看影响⬇️

OpenRouter@OpenRouter · 4月29日33

Welcome @KenTheRogers to the team! Expect to see some technical guides, cookbooks, and more coming from him. He's the author of our first X article 👇

译欢迎 @KenTheRogers 加入团队！期待看到他带来一些技术指南、cookbooks 等。他是我们第一篇 X 文章的作者👇

SemiAnalysis@SemiAnalysis_ · 4月29日66

There is a single thin-film material required by every AI chip on earth. GPUs, TPUs, custom ASICs. All of them. 98% of global supply controlled by one Japanese chemical company. Zero production-ready alternatives. One producer fully booked through 2027. Raising prices. Lead times past 6 months. NVIDIA is so scared they're paying half the capex to expand supplier fabs themselves. The keyword is “umami”. Nobody's talking about this. They will be in about

译地球上每一块AI芯片都需要一种单一的薄膜材料。GPU、TPU、定制ASIC。无一例外。全球98%的供应由一家日本化学公司控制。目前没有可投入生产的替代品。唯一的生产商产能已排期至2027年。正在提价。交货期超过6个月。NVIDIA如此担忧，以至于他们自行承担一半资本支出以扩建供应商的晶圆厂。关键词是“umami”。目前无人讨论此事。但很快人们就会开始关注。

MiniMax (official)@MiniMax_AI · 4月29日34

Excited to power @mira_t_me's AI agent on Telegram — multimodal, at scale, for everyday users. Welcome to the ecosystem 🤝

译很高兴能为 @mira_t_me 在 Telegram 上的 AI 智能体提供支持——多模态、大规模，服务于日常用户。 欢迎加入这个生态系统 🤝

Rohan Paul@rohanpaul_ai · 4月29日47

Japan tests humanoid robots at Tokyo Haneda Airport to tackle labour shortage crisis These are Unitree G1 and taller UBTECH Walker E robots for tough jobs like loading cargo and moving luggage. Japan's aviation sector struggles with staff shortages due to more tourists visiting and fewer young workers available. JAL has about 4000 people doing ground handling jobs. The airline wants the robots to take on tough physical tasks to help employees. They may later use them to clean plane cabins and run other support machines.

译为应对因游客激增和年轻劳动力不足引发的航空业人力危机，日本在东京羽田机场测试人形机器人。测试机型包括Unitree G1和UBTECH Walker E，旨在承担装载货物、搬运行李等繁重体力劳动。日本航空希望借此减轻约4000名地勤人员的工作负担，并计划未来将机器人用于清洁客舱和操作其他辅助设备。

Chubby♨️@kimmonismus · 4月28日80

http://x.com/i/article/2049134075466670081 # Musk vs. OpenAI: the tl;dr. A (neutral) summary The most consequential AI lawsuit in history opened in an Oakland courtroom this week. Elon Musk is suing Sam Altman, the CEO of the company they once built together, over what Musk calls the ultimate Silicon Valley betrayal. At stake: $134 billion in damages, the leadership of OpenAI, and a legal precedent that could fundamentally alter how mission-driven tech companies are structured and funded. Opening arguments begin today, Tuesday, April 28, before Judge Yvonne Gonzalez Rogers at the U.S. District Court for the Northern District of California. Both Musk and Altman are expected to take the stand. The trial could last several weeks. ## How it started OpenAI was founded in 2015 by Musk, Altman, Greg Brockman, and a small group of researchers as a nonprofit. The mission was explicit: build artificial general intelligence that benefits all of humanity, free from shareholder pressure and profit motives. Musk contributed roughly $44 million in the early years. The idea was radical. A group of elite researchers, backed by philanthropic capital, would develop the most powerful technology in human history and give it away. However, that is not what happened - obviously. By 2017, the founders realized that competing with deep-pocketed corporations like Google required far more capital than donations could provide. Discussions about restructuring began. Musk and Altman clashed over leadership and strategic direction. Musk pushed for a merger with Tesla. The others resisted. In 2018, Musk left the board. OpenAI cited potential conflicts of interest with Tesla. A year later, OpenAI created a for-profit subsidiary. The nonprofit technically remained in control, but the commercial arm quickly eclipsed it. Then came ChatGPT, and everything accelerated. Today, OpenAI is valued at roughly $850 billion and is preparing for what could be one of the largest IPOs in tech history. It claims nearly a billion weekly active users. The nonprofit that once defined the organization now exists largely as an oversight layer. ## What Musk claims Musk's central allegation is straightforward: he was deceived. He says Altman and Brockman convinced him to fund a nonprofit with a humanitarian mission, then quietly converted it into a commercial juggernaut that enriched them personally. In court filings, Musk's lawyers described the situation as betrayal of historic proportions and accused Altman of running a calculated deception from the start (Musk, Court Filing, 2024). Of the 26 claims Musk originally filed in November 2024, only two remain after pre-trial narrowing: unjust enrichment and breach of charitable trust. His lawyers dropped the fraud claims to streamline the case. But the remedies Musk is seeking are sweeping. He wants the court to reverse OpenAI's 2019 for-profit conversion, restore the company to full nonprofit status, remove Altman and Brockman from their leadership positions, and award up to $134 billion in damages. Musk has told the court that any financial award should go to OpenAI's nonprofit foundation, not to him personally. Microsoft, one of OpenAI's largest backers, is named as a co-defendant and accused of aiding and abetting the breach of charitable trust. Microsoft CEO Satya Nadella is on the witness list. Musk has also calls the case as a matter of public principle. He has argued that allowing this kind of nonprofit-to-profit conversion without consequence would set a dangerous legal precedent, eroding trust in charitable organizations across the United States. ## What OpenAI Claims OpenAI sees things very differently. The company has characterized the entire lawsuit as a competitive attack. On X, the company wrote: "This lawsuit has always been a baseless and jealous bid to derail a competitor. We can't wait to make our case in court where both the truth and the law are on our side." (OpenAI, April 2026) On its website, OpenAI published a detailed rebuttal, stating that Musk's actions are driven by jealousy, regret over leaving the company, and a desire to advantage his own AI venture, xAI (OpenAI, 2026). The company argues that Musk was fully aware of and participated in early discussions about restructuring OpenAI into a for-profit entity. Internal communications suggest that by 2017, all founders, including Musk, understood that the nonprofit model alone could not support the compute infrastructure required to build frontier AI. OpenAI also points out that Musk leveraged the organization for his own benefit before he left. According to internal records published by OpenAI, Musk asked the OpenAI team to help fix Tesla's Autopilot program in early 2017, leading to the recruitment of key researcher Andrej Karpathy to Tesla (OpenAI, 2026). The company contends that Musk left because he could not assume total control, not because of any principled disagreement over structure. OpenAI emphasizes that it remains governed by a nonprofit dedicated to the same mission of creating AGI that benefits humanity, and that the OpenAI Foundation is now one of the best-resourced nonprofits in history, with a valuation exceeding $180 billion (OpenAI, 2026). Altman himself has not stayed quiet. In February, he posted on X: "Really excited to get Elon under oath in a few months, Christmas in April!" (Altman, February 2026). Musk, in turn, has repeatedly called Altman "Scam Altman" on social media and stated that Altman "lies as easily as he breathes" (Musk, August 2025). ## What now? This trial is not just about two billionaires settling a grudge. It is a test case for a question that now haunts the entire tech industry: when a company is founded with a public-good mission and later pivots toward profit, what are the legal limits of that transformation? Jill Fisch, a professor of business law at the University of Pennsylvania, put it clearly: when someone invests in an organization that promises to operate in a socially responsible way, and the leadership decides to pivot, there have to be limits on how far that pivot can go (Fisch, 2026, https://www.wvia.org/news/2026-04-27/musk-vs-altman-tech-ceos-head-to-court-over-the-fate-of-openai). The outcome will ripple far beyond OpenAI. Anthropic, Musk's own xAI, and a generation of mission-driven AI startups all use different corporate structures designed to balance purpose and capital. A ruling against OpenAI could force a rethink of how these hybrid models work. A ruling in OpenAI's favor could validate the idea that nonprofits can evolve into commercial powerhouses without legal consequence. The timing amplifies the stakes. OpenAI is targeting a fourth-quarter IPO. Musk is preparing to take SpaceX public in what could be a record offering. Together, the two companies are valued at over $2 trillion on private markets. A loss for OpenAI could derail its public listing entirely. As Casey Newton, founder of the Platformer newsletter, put it: the thrust of Musk's case is to stop OpenAI in its tracks, prevent them from developing future models, and essentially knock a major player out of the AI race (Newton, 2026). The jury has been seated. Opening arguments start today. Both men will testify. The AI industry is watching.

译埃隆·马斯克在美国加州法院起诉OpenAI及其CEO萨姆·奥特曼，指控其将最初的非营利组织转变为营利性商业实体，构成背叛。马斯克要求法院撤销2019年的营利性转型，恢复非营利地位，罢免奥特曼等管理层，并索赔高达1340亿美元。他认为此案关乎公共原则，若此类转型无后果将危及全美慈善机构信任。OpenAI则反驳称该诉讼是“无根据的、出于嫉妒的竞争攻击”。此案结果可能为使命驱动型科技公司的结构与融资设立重要法律先例。

Runway@runwayml · 4月28日33

He's a goat. That's the show. Enter your #RunwayBigPitchContest submission today. Up to $100K in cash prizes. Contest closes May 4th. Learn more and submit at the link below.

译他是山羊。这就是节目。 今天提交你的 #RunwayBigPitchContest 参赛作品。现金奖高达10万美元。比赛截止日期为5月4日。 通过下方链接了解更多信息并提交作品。

Rohan Paul@rohanpaul_ai · 4月28日39

See the handover in this Chinese robot-run convenience store. The G1 walks to the refrigerated section post-payment, uses its grippers to grasp, then returns to the counter. Early commercial pilots replacing cashier roles with robots in 24/7 retail.

译看看这家中国机器人运营便利店中的交接流程。G1在付款后走向冷藏区，用它的抓取器拿取商品，然后返回柜台。 早期商业试点正在用机器人取代24/7零售中的收银员角色。

Alibaba Cloud@alibaba_cloud · 4月28日34

@AkoolInc x Alibaba Cloud: 60% less flickering in AI video! ⚡️ ● Integrates Wan, Qwen-Image & Qwen-VL ● 40-60% cost reduction via flexible deployment ● 180+ languages translated in minutes Commercial AI video at scale. Read the success story: https://int.alibabacloud.com/m/1000412139/

译@AkoolInc x 阿里云：AI视频闪烁减少60%！⚡️ ● 集成Wan、Qwen-Image & Qwen-VL ● 通过灵活部署实现40-60%成本降低 ● 支持180+种语言分钟级翻译 规模化商用AI视频。 阅读成功案例：https://int.alibabacloud.com/m/1000412139/

Chubby♨️@kimmonismus · 4月28日70

Google just signed a deal letting the Pentagon use its AI models for classified work and "any lawful government purpose." This comes despite over 600 employees urging CEO Sundar Pichai to reject the agreement, and marks a dramatic reversal from 2018 when Google pulled out of Project Maven after employee backlash. Google now joins xAI and OpenAI in having classified Pentagon AI deals, with terms that appear even more permissive than OpenAI's. The contract includes language saying Google's AI "is not intended for" mass surveillance or autonomous weapons without human oversight, but legal experts say this wording is not legally binding. Notably, the deal also requires Google to adjust its AI safety filters at the government's request. This all follows Anthropic's public refusal to drop its red lines on those exact use cases, which led to the Pentagon declaring Anthropic a supply chain risk, a designation Anthropic is currently fighting in court.

译谷歌已与五角大楼签署协议，允许其AI模型用于机密工作及“任何合法的政府目的”，此举无视了超600名员工的反对，并逆转了其2018年因员工抗议退出Project Maven的立场。协议条款看似比OpenAI的同类合约更为宽松，虽声明AI“不拟用于”大规模监控或无人监督的自主武器，但法律专家指出该措辞缺乏约束力。协议还要求谷歌应政府要求调整AI安全过滤器。这与Anthropic因拒绝在类似用途上妥协而被五角大楼列为供应链风险形成对比。

Alibaba Cloud@alibaba_cloud · 4月28日28

Alibaba Cloud is proud to support the TNG Digital FINHACK 2026, where 300+ Malaysian innovators tackled real fintech challenges from fraud prevention to financial inclusion, powered by AI. As a strategic partner, Alibaba Cloud provided cloud infrastructure, AI tools including Qwen, our #1 open-source model, Qoder, our next generation AI powered coding platform and on ground Solutions Architect support to help the future talents build scalable and real world solutions. Congratulations to all participants and winners! Together, we are accelerating Malaysia's fintech future: inclusive, secure and intelligent. #AlibabaCloud #TNGDigital #Fintech #AI #MalaysiaTech #AInnovation #Qwen #Qoder

译阿里云作为战略合作伙伴，支持了TNG Digital FINHACK 2026活动。该活动汇聚了300多名马来西亚创新者，利用人工智能应对从欺诈预防到金融普惠等现实金融科技挑战。阿里云提供了云基础设施、AI工具（包括排名第一的开源模型Qwen和下一代AI驱动编码平台Qoder）以及现场解决方案架构师支持，旨在帮助未来人才构建可扩展的、面向真实世界的解决方案。活动致力于共同加速马来西亚金融科技的未来，使其更具包容性、安全性和智能化。

SiliconFlow@SiliconFlowAI · 4月28日43

Builders are voting with their tokens 🔥 SiliconFlow is now the #1 third-party model provider by daily token usage On @OpenRouter , • ~280B tokens/day • ~1.9T tokens/month • 33 frontier models: DeepSeek V4 series, GLM 5.1, Kimi K2.6 etc. Big thanks to every dev building with us And more is coming🚀

译开发者们正在用他们的代币投票 🔥 SiliconFlow 现已成为日使用代币量排名第一的第三方模型提供商 在 @OpenRouter 上， • 每日约 2800 亿代币 • 每月约 1.9 万亿代币 • 33 个前沿模型：DeepSeek V4 系列、GLM 5.1、Kimi K2.6 等 衷心感谢每一位与我们共同构建的开发者 更多精彩即将到来 🚀

Alibaba Cloud@alibaba_cloud · 4月28日30

With close to 200 participants joining our first AI workshop, the Alibaba Cloud Singapore SG WANders AI Video Challenge is off to an exciting start! A big thank you to everyone who has registered so far - we're thrilled by the strong response. With just 1 day left before registration closes, don't miss your chance to be part of this challenge. Register now and stand a chance to win exciting prizes: https://int.alibabacloud.com/m/1000412215/ Our second workshop will be held on 4th May, giving participants another opportunity to learn, create, and refine their AI video submissions using Alibaba Cloud's WAN model. Co-organised with FizzDragon in collaboration with the Institute of Technical Education, NYP, OSC, and Silver Media Group. #AlibabaCloudSG #AlibabaCloudPartner #AlibabaCloud #AI #DigitalTransformation #LLM #Qwen #Wan #ContentCreation #DigitalUpskilling #CloudComputing #AInnovation #OnlineWorkshop #SGWANders #AIVideoChallenge #Hackathon #Singapore

译阿里云在新加坡举办的“SG WANders AI视频挑战赛”已拉开序幕，首场AI工作坊吸引了近200名参与者，反响热烈。注册即将截止，主办方鼓励参与者把握最后机会报名，并有机会赢取奖品。第二场工作坊定于5月4日举行，旨在帮助参与者利用阿里云的WAN模型学习、创作和完善其AI视频作品。本次活动由阿里云与FizzDragon联合组织，并得到了新加坡工艺教育学院等多家机构的协作支持。

Alibaba Cloud@alibaba_cloud · 4月28日26

HappyHorse is now on GMI

译HappyHorse现已登陆GMI [引用 @gmi_cloud]：Happy Horse is on GMI @HappyHorseATH 又一个将我们推向电影级制作的视频模型👇

宝玉@dotey · 4月28日69

转译原推：OpenAI 或将重新定义智能手机；联发科、高通与立讯精密，可能成为其 AI 智能体手机的关键伙伴 • 最新产业调研显示： OpenAI 正在与联发科（MediaTek）和高通（Qualcomm）合作开发智能手机处理器，立讯精密（Luxshare）则是独家系统共同设计与制造伙伴。该产品预计将在 2028 年进入量产。 • AI 智能体将重新定义智能手机： 用户真正想要的，并不是在手机里打开一堆 App。用户想要的是通过手机完成任务、满足需求。这个变化会从根本上改变人们对智能手机的理解。 我做了一版智能手机界面的概念设计，放在这篇文章最后，用 iPhone 作为例子，和今天的手机使用模式做对比。 • OpenAI 为什么要做手机？ 只有同时完全掌控操作系统和硬件，OpenAI 才能提供完整的 AI 智能体服务。 智能手机是唯一能够持续捕捉用户完整实时状态的设备，而这些实时状态，正是 AI 智能体进行实时推理时最重要的输入。 在可预见的未来，智能手机仍将是规模最大的设备品类。 • 云端 AI 与端侧 AI 将深度结合： 手机需要持续理解用户所处的上下文。因此，功耗、内存层级管理，以及基础小模型的运行能力，都会成为处理器设计时的关键考量。 更复杂、计算量更大的任务，则会交给云端 AI 来处理。 • OpenAI 的优势，在于它的消费级品牌、长期积累的用户数据，以及领先的 AI 模型。 智能手机硬件本身已经非常成熟，所以 OpenAI 可以借助供应链来开发设备。 在商业模式上，OpenAI 可能会把订阅服务和硬件绑定在一起，并与开发者共同建立一个新的 AI 智能体生态系统。 • 联发科和高通是处理器共同开发伙伴，未来可能受益于长期换机需求： 规格和供应商预计将在 2026 年底或 2027 年第一季度敲定。 以“联发科 × Google TPU Zebrafish”为例，单颗芯片带来的收入贡献，大约相当于 30–40 颗 AI 智能体手机处理器。假如初期目标锁定全球高端智能手机市场，而这个市场每年出货量约为 3 亿到 4 亿台，那么未来的换机周期可能会成为另一个重要增长引擎。 • 无论立讯精密多努力，要在苹果供应链中超越鸿海的组装地位，都会非常困难。 也正因为如此，这个项目对立讯精密尤其重要。 如果它能在早期就卡位这一供应链，就有机会成为下一代智能手机浪潮中的核心受益者。

译据报道，OpenAI正与联发科和高通合作开发智能手机处理器，立讯精密为独家设计与制造伙伴，目标2028年量产。此举旨在通过完全掌控硬件与操作系统，提供完整的AI智能体服务。智能手机被视为能持续捕捉用户实时状态的关键设备，是AI实时推理的重要输入。未来手机将深度结合云端与端侧AI，处理器设计需重点考量功耗与上下文理解。该项目对联发科和高通意味着长期增长机会，对立讯精密则是切入下一代智能手机核心供应链的关键。

TestingCatalog News 🗞@testingcatalog · 4月28日75

OpenAI models are coming to AWS Badrock in the coming weeks. Besides that, Amazon is hosting a livestream on April 28 with OpenAI leaders. > Watch Matt Garman, Colleen Aubrey, Julia White, and OpenAI leaders in a candid conversation about what's next with agentic AI.

译OpenAI模型将在未来几周内通过AWS Bedrock向客户提供。此举旨在为开发者提供更多模型选择，以适配不同任务需求。同时，亚马逊将于4月28日举办直播活动，AWS高管Matt Garman、Colleen Aubrey、Julia White将与OpenAI领导人进行对话，探讨智能体AI（agentic AI）的未来发展方向。相关详细信息将在旧金山的AWS活动中公布。

TestingCatalog News 🗞@testingcatalog · 4月28日55

Meta partners with Overview Energy to bring up to 1 GW of space solar energy to Earth! Meta also partners with Noon Energy to deploy up to 1 GW/100 GWh of energy storage. It is Meta vs SpacexAI now 👀

译Meta与Overview Energy合作，将高达1吉瓦的空间太阳能输送至地球！ Meta还与Noon Energy合作，部署高达1吉瓦/100吉瓦时的储能系统。 现在是Meta对阵SpacexAI了 👀 [引用 @Meta_Engineers]：这些与Overview和Noon的合作延续了我们的一贯策略，即通过多元可靠的解决方案助力电网强化，并为我们的AI基础设施供电。 了解更多：https://go.meta.me/635755

Microsoft Research@MSFTResearch · 4月28日51

Coming May 14 at Microsoft Research Forum: a new release and demo from MSR AI Frontiers. Plus new work on Agentic GitHub Workflows, Real-time agent verification, Energy-based fine-tuning, and Guiding the AI transition. Register now:

译微软研究论坛将于5月14日发布：来自MSR AI Frontiers的新版本和演示。 此外，还有关于Agentic GitHub工作流程、实时智能体验证、基于能量的微调以及引导AI转型的新工作。立即注册：

Chubby♨️@kimmonismus · 4月28日65

For some reason Microsoft is the only company that doesn’t benefit from AI.

译出于某种原因，微软是唯一一家没有从AI中受益的公司。

宝玉@dotey · 4月28日79

OpenAI 和微软重新谈了合作协议，核心变化一句话概括：OpenAI 不再被绑在 Azure 上了。 根据新协议，微软仍然是 OpenAI 的主要云合作伙伴，OpenAI 的产品也会优先在 Azure 上线。但有个关键松绑：如果微软无法或选择不支持某些能力，OpenAI 可以把产品部署到任何云平台上。这等于给了 OpenAI 一张多云通行证。 另一个重要变化是 IP 授权。微软对 OpenAI 模型和产品的授权延续到 2032 年，但从独家变成了非独家。也就是说，OpenAI 可以把同样的技术授权给其他公司了。 钱的部分也理顺了：微软不再向 OpenAI 支付分成，而 OpenAI 向微软的分成持续到 2030 年，比例不变但设了总额上限。微软作为大股东继续享受 OpenAI 的增长红利。 【注：此前 OpenAI 和微软的关系一直很拧巴。微软既是投资人又是云服务商，还拿着独家 IP 授权，OpenAI 每赚一笔都要分给微软。这种深度捆绑在 OpenAI 还是非营利机构时没什么问题，但随着 OpenAI 转型为营利公司、筹备 IPO，这套架构越来越不合适。新协议本质上是在 IPO 之前把关系理清楚。】 对普通用户来说，最直接的影响是：以后用 ChatGPT 或 OpenAI API 的企业客户，不一定非得走 Azure 了。用 AWS 或 Google Cloud 的公司接入 OpenAI 的服务会更方便。对微软来说，虽然失去了独占地位，但作为股东的身份让它依然能从 OpenAI 的增长中获益，只是从"锁定"变成了"竞争"。 OpenAI 正在一步步把自己从微软的生态里松绑出来，而微软也在接受一个现实：与其死守排他协议，不如当一个赚钱的股东。 https://openai.com/index/next-phase-of-microsoft-partnership/

译OpenAI与微软重新谈判合作协议，核心变化在于解除了对Azure云的独家依赖。根据新协议，微软仍是OpenAI的主要云合作伙伴，产品优先在Azure上线，但若微软无法支持特定能力，OpenAI可将服务部署到任何云平台。IP授权延续至2032年，但由独家转为非独家，OpenAI可向其他公司授权技术。财务安排上，微软不再向OpenAI支付分成，而OpenAI向微软的分成持续至2030年并设有上限。此举为OpenAI的IPO铺平道路，企业客户未来可通过AWS或Google Cloud等平台使用其服务，微软则以股东身份继续受益。

SemiAnalysis@SemiAnalysis_ · 4月28日44

While Intel gains external customers using EMIB – Google's TPU being the big one – they're moving away from it for their own products. Diamond Rapids will likely use UCIe over substrate for a long-reach die-to-die interconnect instead. At ISSCC, Intel showed a UCIe-S D2D link on 22nm hitting 48 Gb/s/lane over standard organic substrate at a reach of up to 30 mm. Beat a 3nm design with 3× higher data rate and 2.8× higher bandwidth density. 5-2-5 substrate vs 11-2-11 on EMIB. With substrate in short supply, Intel's "best" packaging tech – for everyone but Intel.

译Intel的EMIB封装技术被外部客户如Google的TPU采用，但Intel自身产品正转向UCIe技术。Diamond Rapids预计使用UCIe over substrate实现长距离die-to-die互连。在ISSCC上，Intel展示了UCIe-S D2D链接，在22nm工艺下达到48 Gb/s/lane，距离达30mm，数据率和带宽密度优于3nm设计。substrate配置为5-2-5，而EMIB为11-2-11。由于substrate短缺，Intel的“最佳”封装技术对除Intel外的客户更具优势。

宝玉@dotey · 4月28日74

GitHub Copilot 从 6 月 1 日起改按用量计费。订阅价格没变，但"用多少付多少"的规则会让重度用户的账单变得不太确定。 过去一年，Copilot 从一个编辑器里的补全助手，变成了能跑多步骤、跨整个代码仓库的 Agent 编程平台。一个简单的聊天提问和一次跑几个小时的自动编程任务，以前消耗的“高级请求次数”可能一样多，GitHub 一直在背后默默吸收那些飙升的推理成本。现在扛不住了。 新规则的核心：取消“高级请求次数”，换成 AI 积分（AI Credits）。积分按 Token 消耗计算，包括输入、输出和缓存的 Token，费率跟各模型的 API 定价挂钩。 各档订阅价不变，每月自动到账与订阅价等额的积分：Pro 是 10 美元对应 10 美元积分，Pro+ 是 39 美元对 39 美元，Business 19 美元/人，Enterprise 39 美元/人。代码补全和“下一步编辑建议”这类基础功能不消耗积分，跟以前一样包含在订阅里。 有两个细节值得注意。第一，以前高级请求用完了还能降级到便宜模型继续干活，以后这条退路没了，积分花完就是花完，除非买更多或者管理员开了预算。第二，Copilot 的代码审查功能除了消耗 AI 积分，还会额外消耗 GitHub Actions 的运行时长。 企业用户有三个月的过渡缓冲：6 月到 8 月，Business 用户每月拿到 30 美元积分（比订阅价多 11 美元），Enterprise 拿到 70 美元（多 31 美元）。企业还能把团队成员的积分打通成资源池，用不完的不浪费。 5 月初 GitHub 会上线预览账单功能，让你在正式切换前看看自己按新规则大概要花多少钱。年付用户暂时不受影响，到期后才会转到新体系。 对轻度用户来说，这次变化几乎无感。但如果你已经习惯了让 Copilot Agent 跑长任务，6 月之后最好盯一下账单。

译GitHub Copilot 将于6月1日起改用基于AI积分的用量计费模型，以支持更多Agent和高级工作流。各档订阅价格不变，每月赠送等额积分，代码补全等基础功能不消耗积分。新规则按Token消耗计费，积分用尽后无降级选项，代码审查会额外消耗Actions时长。企业用户有三个月过渡期及积分池福利。5月初将上线账单预览功能，年付和轻度用户受影响小，但重度用户需关注成本变化。

Chubby♨️@kimmonismus · 4月28日67

OpenAI has removed the AGI clause from its agreement with Microsoft. Given the upcoming IPO, it wouldn't surprise me if AGI is officially announced soon.

译OpenAI已从其与微软的协议中移除了AGI条款。 考虑到即将进行的IPO，如果很快正式宣布AGI，我也不会感到惊讶。

DogeDesigner@cb_doge · 4月28日48

Ex-board member of OpenAI calls Sam Altman a liar. He lied to the board for years, hid ChatGPT launch, lied about owning Startup Fund, falsified safety info, and lied to oust her after her paper. Board lost all trust → fired him. Sam Altman is a liar.

译OpenAI前董事会成员称Sam Altman是骗子。 他多年来对董事会撒谎，隐瞒ChatGPT的发布，在拥有创业基金一事上说谎，伪造安全信息，并在她的论文发表后撒谎以驱逐她。 董事会失去所有信任 → 解雇了他。 Sam Altman是个骗子。

Satya Nadella@satyanadella · 4月28日51

Super to see Accenture roll out 740,000+ M365 Copilot seats—our largest deployment to date! https://news.microsoft.com/source/features/digital-transformation/accenture-is-rolling-out-copilot-to-a-workforce-the-size-of-denver

译很高兴看到埃森哲推出超过74万个M365 Copilot席位——这是我们迄今为止最大规模的部署！ https://news.microsoft.com/source/features/digital-transformation/accenture-is-rolling-out-copilot-to-a-workforce-the-size-of-denver

Rohan Paul@rohanpaul_ai · 4月28日69

OpenAI is reportedly exploring a smartphone that treats AI agents as the operating model, not apps. For this, as per news, OpenAI is exploring custom mobile silicon with MediaTek, Qualcomm, and Luxshare, in a co-design and manufacturing role, because agentic computing is not just model inference, it is power efficiency, memory management, and deciding what stays on-device versus what goes to the cloud. Aiming to replace much of the app model with a system that acts across the whole device. The deeper logic is vertical integration. OpenAI has already folded Jony Ive’s io team into the company, and Chris Lehane has said its first hardware product is on track to be announced in the second half of 2026, which suggests a long game around owning more of the stack, not just shipping another companion app. The basic idea is that apps split your tasks into separate boxes, while an agent-first phone would watch context across messages, location, voice, calendar, and settings, then complete actions directly instead of handing you 10 separate screens. That matters for capability because Apple and Google decide what normal apps can access, so a full hardware and software stack could let OpenAI connect models more deeply to calling, writing, booking, search, reminders, and device controls. The likely technical setup is a split between small on-device models for fast, private, low-latency work and cloud models for heavier reasoning, which is the practical way to make an always-available assistant feel responsive without draining the phone. A phone is the densest stream of personal context most consumers generate, so a true AI agent phone would not merely answer questions better, it would see more, remember more, and infer more than an app sitting inside Apple’s or Google’s rules can. --- techcrunch .com/2026/04/27/openai-could-be-making-a-phone-with-ai-agents-replacing-apps/

译OpenAI据传正探索开发一款以AI代理为核心操作模式的智能手机，旨在取代传统应用模型。公司计划与联发科、高通及立讯精密合作设计定制移动芯片，以优化能效与云端协同，并已吸纳Jony Ive的设计团队。该设备计划于2026年下半年发布，其核心逻辑是通过软硬件垂直整合，让AI代理跨消息、位置、日历等多维度上下文直接执行任务，突破苹果与谷歌对应用权限的限制。技术架构预计采用本地轻量模型实现低延迟响应，结合云端模型处理复杂推理，以打造能深度感知与推断的个人智能助手。

Rohan Paul@rohanpaul_ai · 4月28日79

OpenAI is moving away from its exclusive Microsoft arrangement, making room for possible partnerships with Amazon and Google. At the same time, Microsoft wants to cut its dependence on OpenAI by creating its own AI models. Ending the exclusivity could also help lower antitrust scrutiny across the U.S., U.K., and Europe. Barclays analysts said the shift may free Microsoft to invest more in Copilot and expand cloud capacity. Last month, Microsoft was considering legal action against Amazon and OpenAI over a $ 50B cloud deal that may have violated its exclusive cloud partnership. The revised arrangement now lets OpenAI run its services on Amazon’s cloud without the technical changes required under the earlier Microsoft agreement. Microsoft still keeps a non-exclusive license to OpenAI model and product IP through 2032, still gets OpenAI revenue share through 2030, and still sits inside the upside as a major shareholder. Microsoft stock, falls -5% after announcing that its OpenAI license will now be nonexclusive and it will no longer pay revenue share to OpenAI. Investors can punish the loss of exclusivity on the headline, even if the deeper effect is to simplify the alliance and reduce regulatory and capital strain. But the ​software giant will no longer share revenue for the OpenAI products it sells on its cloud. Revenue OpenAI must share with Microsoft through 2030 will ​now have a cap for the total number and no longer tied to the startup's technology milestones. My read is that OpenAI gained more bargaining power than Microsoft gained certainty, which explains why investors initially treated the change like a moat reduction rather than a breakup. For Microsoft, the selloff reflects a real loss of scarcity premium, but the deeper gain is strategic - freeing capital for Copilot and other cloud capacity, while also easing antitrust pressure and reducing dependence on a partner that recently accounted for about 45% of Microsoft’s remaining performance obligation. Amazon looks like the clearest tactical winner. It already has a strategic OpenAI partnership, Bedrock distribution for OpenAI Frontier, and a large Trainium capacity commitment, so this deal turns AWS from backup infrastructure into a front-line route to OpenAI. Google Cloud gains something subtler but important: the right to compete for OpenAI workloads even while competing in models.

译OpenAI修订与微软的合作协议，结束独家云合作关系，未来可将其服务部署于亚马逊云等平台。微软保留至2032年的非独家模型许可以及至2030年的有上限收入分成。此举削弱了微软的稀缺性优势，引发其股价下跌，但有助于微软释放资本投入Copilot并缓解反垄断压力。亚马逊成为明确赢家，其AWS从备用基础设施转为OpenAI的前线平台；谷歌云也获得了竞争OpenAI工作负载的机会。OpenAI创始人确认微软仍是其主要云合作伙伴，但产品将扩展至所有云平台。

MiniMax (official)@MiniMax_AI · 4月28日27

We'll be at @stripe Sessions this week! 🔖 Booth 604 📅 Apr 29–30 📍 Moscone West, SF Come say hi 👋

译本周我们将在@stripe Sessions现场！ 🔖 604号展位 📅 4月29日至30日 📍 旧金山莫斯康西馆 来打个招呼吧 👋

Runway@runwayml · 4月28日45

Our first online Runway API Hackathon begins next week. One weekend. Free API credits. $25,000 in prizes. Build real-time video agents, agentic media pipelines or anything you can imagine. No experience required. Learn more and register at the link below.

译我们的首次线上Runway API黑客马拉松将于下周开始。一个周末。免费API额度。25,000美元奖金池。 构建实时视频智能体、智能媒体流水线或任何你能想象的项目。无需经验。 通过下方链接了解更多信息并注册。

Chubby♨️@kimmonismus · 4月27日56

OpenAI is working on a phone. They are expected to compete with apples iPhone. First product however is expected to be a HomePod style device co-worked with ex apple legend Jony Ive

译OpenAI 正在研发一款手机。 预计他们将与苹果的 iPhone 展开竞争。 不过首款产品预计将是与苹果前传奇设计师 Jony Ive 合作开发的 HomePod 风格设备。

Emad@EMostaque · 4月27日65

Looking forward to the work coming out of @IneffableLabs 🇬🇧 Largest EU/UK raise ever 😮 Beat the previous UK largest seed raise ($101m Stability AI!) by $999m 🚀 We have seen amazing things on our self-learning models so bullish on this approach 🤓

译Ineffable Labs完成了英国及欧盟有史以来规模最大的融资，其种子轮融资额远超此前由Stability AI保持的1.01亿美元英国纪录。该公司由David Silver领导，致力于汇聚全球顶尖工程师与研究人员，以开发自学习模型，目标是实现与超级智能的“首次接触”，并在此过程中攻克AI领域最艰巨的难题。

小互@xiaohu · 4月27日29

阿里的 HappyHorse 上线了 字节的回应🤣

译阿里的 HappyHorse 上线了 字节的回应🤣

Jeff Dean@JeffDean · 4月27日56

The video of my conversation with Amin Vahdat, @gilbert, and @djrosent at Cloud Next last week is now up. https://youtu.be/BpnJYJmbXcM?si=vUY3hI_aDX8K6gco Thanks for a great conversation!

译视频记录了在Cloud Next大会上与Amin Vahdat及AcquiredFM主持人的对话，核心围绕谷歌最新发布的TPU v8t和v8i芯片展开讨论。对话内容基于官方博客公布的芯片细节，探讨了其在“智能体时代”的基础设施意义与技术亮点。

TestingCatalog News 🗞@testingcatalog · 4月27日60

Meta 🚫 Manus AI China has blocked Meta’s $2.000.000.000 acquisition of Manus AI platform. The time for a new update 👀 > Manus > from China Yet, it is quite disappointing to observe. I think for Meta it will be much harder to catch up without them. Whom are they gonna buy next?

译Meta 🚫 Manus AI 中国已阻止Meta以20亿美元收购Manus AI平台。 新更新的时机来了👀 > Manus > 来自中国 然而，这观察结果相当令人失望。我认为对Meta来说，没有它们将更难追赶。 他们接下来会收购谁？

Google DeepMind@GoogleDeepMind · 4月27日37

A decade ago in Korea, AlphaGo showed AI’s potential. Together with the Korean government, we’re now looking at how this technology can help accelerate scientific discovery and create new opportunities for economic growth across the region. 🇰🇷 Find out more → https://goo.gle/4wbqUbJ

译十年前在韩国，AlphaGo展现了AI的潜力。 如今我们与韩国政府共同探索这项技术如何加速科学发现，并为整个区域创造经济增长的新机遇。🇰🇷 了解更多 → https://goo.gle/4wbqUbJ

Rohan Paul@rohanpaul_ai · 4月27日64

🇨🇳NEWS: China blocked Meta from buying the AI company Manus for 2B dollars. Regulators in China reviewed the purchase and decided it violated their foreign investment rules. They ordered everyone involved to cancel the transaction immediately. Manus began in China but moved its headquarters and key team to Singapore last year after raising money from a US investor. Meta had already started integrating parts of Manus technology into its own AI systems. Chinese officials described the purchase as a conspiratorial effort to drain the country's AI talent and resources. They even stopped two co-founders from leaving the country back in March during the review process. --- ft .com/content/1e4c269a-5258-406c-a308-e55c3d5d640f?syn-25a6b1a6=1

译中国监管机构以违反外商投资规则为由，正式阻止了Meta以20亿美元收购人工智能公司Manus的交易，并下令立即取消。监管方将此次收购定性为企图掏空中国AI人才与资源的“合谋”。Manus虽起源于中国，但去年在获得美国投资后已将总部和核心团队迁至新加坡。Meta已开始将Manus的部分技术整合进自身AI系统。在审查期间，中方甚至在三月阻止了该公司的两名联合创始人离境。

Chubby♨️@kimmonismus · 4月27日54

China has blocked Meta’s $2 billion acquisition of agentic AI startup Manus, citing concerns over foreign investment and the transfer of strategic AI technology to the US. It makes Beijing's sharper stance on AI sovereignty, turning what was once seen as a model for globally ambitious startups into a flashpoint in the US-China tech rivalry. Global race just heated up. Per Bloomberg

译中国以担忧外商投资及战略AI技术流向美国为由，阻止了Meta对智能体AI初创公司Manus的20亿美元收购案。 这标志着北京在AI主权问题上立场更趋强硬，将一度被视为全球雄心初创企业典范的案例，转变为中美科技竞争的新焦点。 全球竞赛再度升温。 据彭博社报道