Tibo@thsottiaux · 5月1日59

Lots of improvements across the board both for coders and general work. Update your codex app to enjoy all the improvements. We will be back next week

译为编码者和普通工作带来了全面的改进。更新您的Codex应用以享受所有改进。 我们下周回来

Tibo@thsottiaux · 5月1日62

This thing does more than what you think it does. Codex now available for non-coders

译这东西的功能比你想象的要多。Codex 现在可供非编码人员使用。

Greg Brockman@gdb · 5月1日43

Secure your ChatGPT account with Advanced Account Security:

译通过高级账户安全功能保护你的 ChatGPT 账户： [引用 @OpenAI]：现已为 ChatGPT 账户推出：高级账户安全，这是一项新的可选设置，适用于面临较高数字攻击风险的用户，提供更强大的保护，包括防钓鱼登录和更安全的账户恢复。 https://openai.com/index/advanced-account-security/

Greg Brockman@gdb · 5月1日36

Codex is for everyone, for any task done with a computer

译Codex 适用于每个人，适用于任何用计算机完成的任务

OpenAI@OpenAI · 5月1日70

It's never been easier to do everyday work with Codex. Choose your role, connect the apps you use every day, and try suggested prompts. Codex helps with everything from research and planning to docs, slides, spreadsheets, and more.

译使用Codex处理日常工作从未如此简单。 选择你的角色，连接你每天使用的应用，并尝试建议提示。 Codex能在研究规划、文档、幻灯片、电子表格等方方面面提供帮助。

Rohan Paul@rohanpaul_ai · 5月1日50

OpenAI just rolled out Advanced Account Security, an opt-in mode that turns ChatGPT and Codex accounts into phishing-resistant accounts by replacing passwords and weak recovery paths with stronger identity checks. The problem is that an AI account now stores chats, work context, connected tools, and sometimes sensitive research or political material, so a stolen inbox, phone number, or browser session can become a direct path to that data. The fix is - password login is shut off, email/SMS recovery is shut off, and access depends on passkeys, physical security keys, backup keys, and shorter-lived sessions with login alerts. OpenAI also ties this mode to privacy by making training exclusion automatic for enrolled accounts, and it says members of Trusted Access for Cyber must enable it by 06-26 unless their organization proves phishing-resistant SSO.

译OpenAI推出可选的“高级账户安全”模式，旨在将ChatGPT和Codex账户转变为防钓鱼账户。该模式彻底禁用密码登录和电子邮件/SMS恢复功能，强制用户依赖Passkey、物理安全密钥等强认证方式，并辅以更短会话周期和登录警报。其背景在于，AI账户存储了对话记录、工作上下文乃至敏感材料，传统邮箱或手机号被盗易导致数据泄露。此外，启用该模式的账户将自动排除在模型训练数据之外。OpenAI要求其“网络可信访问”计划成员在6月26日前启用此模式，除非组织已部署其他防钓鱼单点登录方案。

阿绎 AYi@AYi_AInotes · 5月1日60

Damn，传统代码扫描器90%的报警都是误报，这个困扰了安全行业十年的问题，今天被Claude彻底解决了🤨🤨🤨🤯🤯🤯 Claude今天正式发布了Claude Security，进入公开测试阶段， 仅限企业客户使用， 它不是又一个传统的SAST工具， 会直接关联你的GitHub仓库，自动扫描整个代码库， 用大模型的语义理解能力追踪完整的数据流动， 发现那些传统工具根本找不到的复杂攻击链， 最牛的是，它会主动验证每一个发现的漏洞是不是真的， 自己分裂成两个角色，一个证明漏洞存在，一个拼命反驳， 只有经过这种地狱级验证的结果，才会推送到你面前， 实测误报率只有传统工具的不到10%， 找到问题之后，它会自动生成完整的修复补丁， 带diff，带commit信息，带影响范围说明， 你只需要点一下审查，确认没问题就能合并， 整个流程从扫描到修复，全部在Claude界面里完成， 不需要自建任何工具，不需要额外的API集成， 有意思的是，这个发布正好在OpenAI公布Codex Cyber的第二天， OpenAI给你的是一个可以用来做安全的模型， Claude直接给你一个开箱即用的完整产品， 不需要你自己搭agent，不需要你写prompt， 点一下Security按钮就完事了， 这对Snyk，Semgrep，GitHub Advanced Security这些传统工具来说， 是真正的降维打击， 毕竟以前安全团队要花80%的时间过滤误报， 现在他们只需要花20%的时间审查AI生成的修复， 这意味着DevSecOps的整个工作流，从今天开始被彻底重构了， 不过它现在还只对企业客户开放，个人用户暂时用不到， 但我觉得这已经足够说明，AI正在一个一个吃掉所有的专业工具。

译Claude Security进入公开测试阶段，仅限企业客户使用。它通过关联GitHub仓库自动扫描代码库，利用大模型语义理解追踪数据流动，发现传统工具难以检测的复杂攻击链。工具主动验证每个漏洞，将误报率降至传统工具的不到10%，并自动生成带diff和commit信息的修复补丁，用户只需审查确认即可合并。与OpenAI的Codex Cyber提供模型不同，Claude Security是开箱即用的完整产品，无需自建工具或API集成。这对Snyk、Semgrep等传统工具构成降维打击，重构DevSecOps工作流，大幅减少安全团队处理误报的时间。

TestingCatalog News 🗞@testingcatalog · 5月1日49

Anthropic launched Claude Security in public beta for Enterprise customers. I also hope that at least Team users will get it eventually, too. Some old screenshots 👀

译Anthropic 面向企业客户公开测试 Claude Security。我也希望至少团队用户最终也能用上。 一些旧截图 👀

OpenAI@OpenAI · 5月1日44

Now available for ChatGPT accounts: Advanced Account Security, a new opt-in setting for people at higher risk of digital attacks, with stronger protections including phishing-resistant sign-in and more secure account recovery. https://openai.com/index/advanced-account-security/

译现已为ChatGPT账户推出高级账户安全功能，这是一项面向数字攻击高风险人群的新可选设置，提供更强大的保护措施，包括防钓鱼登录和更安全的账户恢复流程。 https://openai.com/index/advanced-account-security/

Chubby♨️@kimmonismus · 5月1日52

Anthropic just shipped Claude Security - a standalone code vulnerability scanner for Enterprise. Scans your repo, validates findings, suggests patches. Powered by Opus 4.7. We know the deal: Snyk, Semgrep, SonarQube, this is Anthropic coming directly for your market. Stocks goes down.

译Anthropic发布了Claude Security，这是一个面向企业客户的独立代码漏洞扫描工具，由Opus 4.7驱动。它能扫描代码库、验证发现以减少误报，并建议可审查的补丁。Claude Security现已对Claude Enterprise客户开放公开测试。此举直接进入代码安全市场，与Snyk、Semgrep、SonarQube等现有工具竞争，可能对相关市场格局和公司股价产生影响。

NotebookLM@NotebookLM · 5月1日64

Notebooks in @GeminiApp are officially available to Free and Paid users on mobile! 🥳 (Coming soon to additional countries in Europe!)

译Gemini App中的Notebooks功能现已正式在移动端向免费和付费用户开放，并计划很快扩展至更多欧洲国家。此次更新标志着NotebookLM与Gemini的深度集成，用户现可直接在Gemini App内访问所有个人未共享的笔记本，并能将Gemini的聊天记录用作新笔记本或现有笔记本的资料来源。功能推出将分阶段进行：首先面向网页版的Google AI Ultra、Pro和Plus订阅用户，未来几周将逐步覆盖移动端用户、更多欧洲地区以及免费用户。

Claude@claudeai · 5月1日56

Claude Security is now in public beta for Claude Enterprise customers. Claude scans your codebase for vulnerabilities, validates each finding to cut false positives, and suggests patches you can review and approve.

译Claude Security 现已面向 Claude Enterprise 客户开启公开测试版。 Claude 会扫描您的代码库以查找漏洞，验证每个发现以减少误报，并建议您可以审查和批准的补丁。

Replit ⠕@Replit · 5月1日64

Introducing Replit Slides! Made an entire pitch deck in under a minute without touching a single slide. Just describe what you want, chat to iterate, and export to PowerPoint, Google Slides, or PDF (or publish a live URL). Slide Decks are available to Core and Pro users. 🎯

译介绍Replit Slides！ 不到一分钟就制作了整个演示文稿，无需手动操作任何一张幻灯片。只需描述你的需求，通过聊天进行迭代修改，然后导出为PowerPoint、Google Slides或PDF格式（或发布实时链接）。幻灯片功能面向Core和Pro用户开放。🎯

Berryxia.AI@berryxia · 5月1日62

Stripe Sessions 直接把 Agent 经济推上新高度了！💳 @patrickc 亲自总结本次大更新： ✅ 整个经济正在“replatforming” ✅ Agents 将在不久后负责大多数交易 ✅ AI 让“开发者优先”变得前所未有重要（Agents 对好 DX 的饥渴程度远超人类开发者） ✅ 推出 Link AI 钱包：直接把 Agent 指向 https://link.stripe.com → 让它用一次性安全 token 帮你购物 ✅ 新增 Pix、UPI、稳定币支持，Machine Payments 协议新增微支付和循环支付 ✅ Checkout Studio、Adaptive Pricing（订阅版）、Terminal 新硬件、Treasury 多币种扩展…… 从支付基础设施到 Agent 时代经济层，Stripe 正在全面布局！ 完整公告 + 所有新功能戳这里👉 https://stripe.com/sessions

译Stripe在年度大会上宣布一系列战略更新，以迎接AI Agent主导交易的新经济时代。CEO指出，经济正经历“平台重构”，未来多数交易将由Agent完成，这使得“开发者优先”战略至关重要。核心发布包括Link AI钱包，允许Agent使用安全令牌代用户购物，并新增Pix、UPI及稳定币支持。同时，Machine Payments协议增加了微支付和循环支付功能。此外，Checkout Studio、Adaptive Pricing订阅版、新款终端硬件T600以及Treasury的多币种扩展等产品，共同标志着Stripe正从支付基础设施向Agent时代的经济层全面演进。

NotebookLM@NotebookLM · 5月1日64

Notebooks in @GeminiApps are officially available to Free and Paid users on mobile! 🥳 (Coming soon to additional countries in Europe!)

译Google的Gemini应用正式向移动端免费和付费用户推出Notebooks功能，并即将扩展至更多欧洲国家。该功能允许用户在Gemini应用内直接访问所有个人未共享的笔记本，并能将与该AI的对话内容作为新笔记本或现有笔记本的资料来源。此次发布首先面向网页端的Google AI Ultra、Pro和Plus订阅用户，后续将逐步推广至移动端、更多欧洲地区及免费用户。

TestingCatalog News 🗞@testingcatalog · 4月30日61

MANUS 🚨: Manus AI announced Cloud Computer, an always-on machine that works for you 24/7! Always-on agent as an infrastructure 🔥 > Users can host their own 24/7 Bots for Discord and Telegram via Cloud Computer > Build a Persistent Knowledge Base > Deploy and host any open-source tool > Run Scheduled Scrapers > Use CLI Developer Tools Cloud Computer on Manus is now available on the web and mobile for paid personal plans. This needs testing time 👀

译Manus AI宣布推出Cloud Computer，这是一种作为基础设施的、7x24小时持续运行的云端计算机。用户可通过它托管Discord和Telegram的持久运行机器人，构建持久知识库，部署并托管任何开源工具，运行定时网络爬虫，以及使用命令行开发者工具。该服务现已面向付费个人用户开放网页版和移动版访问，团队计划即将推出。

OpenRouter@OpenRouter · 4月30日52

alphaXiv x OpenRouter is live! When an AI paper mentions a model, alphaXiv now turns it into a preview: provider, description, use-case rankings, and a direct link to the OpenRouter model page. Go from research to model in one click with @askalphaxiv

译alphaXiv x OpenRouter 现已上线！ 当 AI 论文提及某个模型时，alphaXiv 现在会将其转换为预览：提供商、描述、用例排名，以及直达 OpenRouter 模型页面的链接。 通过 @askalphaxiv 一键从研究直达模型。

Berryxia.AI@berryxia · 4月30日59

🚀 Qwen 重磅开源 Qwen-Scope！ 稀疏自编码器完整套件正式发布，把 SAE 特性变成真正能落地的实用工具，模型可解释性直接起飞！ 1. Inference：直接操纵内部特征实现输出控制，完全无需 prompt engineering 2. Data：用极少种子样本就能分类和合成目标数据，解决长尾能力问题 3. Training：精准追溯 code-switching 和重复生成根源，从源头修复 4. Evaluation：通过特征激活模式分析智能挑选 benchmark，减少冗余 Qwen 模型家族的深度可解释性神器，社区快来挖掘新机制和应用！ 项目地址： https://huggingface.co/collections/Qwen/qwen-scope

译Qwen开源了Qwen-Scope，这是一个为Qwen模型家族设计的稀疏自编码器完整套件，旨在将SAE特征转化为实用工具。该套件提供四大核心功能：在推理方面，可直接操纵模型内部特征以控制输出，无需依赖提示工程；在数据方面，能用极少样本对目标数据进行分类和合成，增强模型的长尾能力；在训练方面，能精准追溯代码切换和重复生成等问题的根源并进行修复；在评估方面，可通过分析特征激活模式来智能筛选基准测试，减少冗余。Qwen希望社区能利用此工具深入探索模型内部机制并开发更多应用。

Berryxia.AI@berryxia · 4月30日52

🦾 想象一下： 工厂的流水线灯火通明，一个银灰色的身影悄然走来。它只有123cm高，却拥有15-31个自由度，像科幻电影里走出来的“未来伙伴”。 它叫 Unitree 双臂人形机器人。 今天，Unitree正式宣布：售价$4290！ 它不是冰冷的机器。 它有双目立体视觉，能“看”清深度； 有阵列麦克风和扬声器，能听懂你的自然语言；双臂灵活如人类，可秒换两指爪、三指爪甚至灵巧人手。 拖拽示教、底层全开放SDK，想怎么玩就怎么玩。 它轻量、紧凑，却能上流水线装配零件、实验室精准操作、甚至未来帮你做家务…… 从科研到工业，从教育到创业，它把“人形机器人”这四个字，从科幻拉进了普通人的现实。 这不是遥不可及的梦想，这是你现在就能买来、改装、一起创造未来的起点。 机器人革命，从这一刻，真正开始“平民化”。 ？ 你准备好迎接你的第一个AI搭档了吗？

译Unitree正式推出双臂人形机器人，售价为4290美元，标志着人形机器人进入“平民化”新阶段。该机器人高123厘米，具备15-31个自由度，集成了双目立体视觉、阵列麦克风和扬声器，支持自然语言交互。其双臂灵活，末端执行器可快速更换，并支持拖拽示教和底层全开放SDK，便于开发。设计轻量紧凑，目标应用场景涵盖工业装配、实验室操作乃至未来家务，旨在为科研、教育及创业等领域提供可及的人形机器人平台。

1X@1x_tech · 4月30日24

Building Your NEO

译构建你的NEO

Runway@runwayml · 4月30日28

You don't need to shut down the streets to bring a big idea to life. Just great storytelling. From idea to execution, this short concept was made by a single creative in just one day with Runway. Try it yourself at the link below. #MadeWithRunway

译实现宏大创意无需封锁街道。只需精彩的叙事。从构想到执行，这个短片概念由一位创作者仅用一天时间通过Runway完成。点击下方链接亲自尝试。#MadeWithRunway

meng shao@shao__meng · 4月30日57

Glean 发布 Waldo：自研"智能搜索专用模型"，作为前置环节运行在前沿大模型之前，专门负责检索规划，把"找资料"和"做推理"这两件事拆开 @glean https://www.glean.com/blog/waldo-launch Waldo 基于 NVIDIA Nemotron 3 Nano（30B/3B MoE），采用 instruct 模式而非 reasoning 模式，以压低延迟。使用 Thinking Machines Tinker API 做 LoRA 微调 为什么做 Waldo · 观察：企业 AI 任务无论多复杂，几乎都从搜索开始（多轮迭代检索 → 综合作答） · 痛点：让前沿模型同时做"检索规划"和"深度推理"，是用最贵的算力干最机械的活——慢且贵 · 思路：高频、定义清晰的子任务（如搜索）应该交给专门训练的小模型；前沿模型只做综合与生成 架构关键决策 · Waldo 作为 第一步 运行，而非作为前沿模型的 sub-agent 被调用。 · 子智能体方案需要 3 次串行推理；Waldo 前置最优情况下只需 1 次前沿模型调用 · Waldo 调用工具集（Glean Search、员工搜索、Web Search），完成后不生成自然语言，而是把检索好的上下文"原状"交给前沿模型，让前沿模型像自己搜过一样直接作答 · 不替代 Glean 现有的语义搜索与企业知识图谱，而是在其之上做"规划层" 衍生能力：自适应推理路由 Waldo 跑完后，其自身执行轨迹（调用次数、命中文档数、是否稀疏、是否需越界工具）天然成为路由信号，用来决定下游前沿模型应启用多深的推理档位。模型"读自己的活"来评估任务复杂度。 实际效果 · 单次 LLM 调用：Waldo 比默认推理模型（GPT-5.4 medium）快 10×+（~250ms vs ~3s） · 端到端集成后：延迟降低 ~50%，Token 消耗降低 ~25%，质量无回退 · 约一半查询走"快路径"，根本不需要前沿模型的完整能力

译Glean推出自研的智能搜索专用模型Waldo，旨在将企业AI任务中的“检索规划”与“深度推理”分离。Waldo基于NVIDIA Nemotron 3 Nano构建，作为前置环节运行，专门负责调用Glean Search、员工搜索和Web搜索等工具进行检索规划，并将检索到的上下文直接交给前沿大模型进行综合作答。这种架构避免了让昂贵的前沿模型处理机械的搜索任务，使单次调用延迟降低至约250毫秒，比默认推理模型快10倍以上。集成后，系统端到端延迟降低约50%，Token消耗减少约25%，且约一半查询可走“快路径”无需调用完整前沿模型。该设计印证了专用小模型在处理重复性任务上的效率优势，是智能体AI架构的重要演进。

Baidu Inc.@Baidu_Inc · 4月30日47

http://x.com/i/article/2049739903970529280 # Anyone Can Build Now, with MeDo At the start of this month, a lighthearted trend unexpectedly went viral: SBTI, a meme-style personality quiz that spread rapidly across Chinese social media, driven by how easy it was for users to create, remix, and share their results. It was entertaining, but it also pointed to something deeper. SBTI took off because it lowered the barrier to participation. Anyone could create something personal, recognizable, and instantly shareable. That same principle is now driving a new wave of AI applications, especially vibe coding tools. What once required technical expertise can now start with a simple prompt. This is the shift we've been watching closely. And it's exactly what we built Miaoda and MeDo for. We first launched Miaoda in 2024 as a no-code, conversational application development platform powered by generative AI, able to deliver multi-agent collaboration and multi-tool invocation. As the platform evolved, we introduced Miaoda 2.0 in 2025, significantly expanding its capabilities and launching its international version, MeDo, bringing the same building experience to developers worldwide. At its core, Miaoda and MeDo allow anyone to describe an application in natural language and receive a fully functional, deployable product in minutes, without writing a single line of code. Behind the scenes, the platform orchestrates a team of 10+ specialized AI agents that work together across the full development process, from requirement analysis and code generation to testing and deployment. Users are not just generating prototypes. They are building full-stack, production-ready applications that can be continuously refined through simple, iterative prompts. ## 🌍 Expanding Who Gets to Build There are nearly 8 billion people in the world, but only around 30 million are professional programmers. That means just 0.4% of the global population has traditionally had the ability to turn ideas into working software. For everyone else, the barriers have been clear: complex syntax, steep learning curves, and high development costs. Countless ideas never make it past the concept stage. By shifting creation from writing code to describing intent, we're opening the door for a much larger group of people to build digital tools, whether for business, education, or everyday problem-solving. As of the end of 2025, Miaoda has powered the creation of 500K+ commercial apps across 200+ real-world scenarios. Notably, 81% of creators are non-programmers, and the apps they're building are already serving 10M+ users with nearly 100K daily users. But who are the people that are actually building with it? Take Sean, a UK-based developer with no technical background. He tried to teach himself game development through online tutorials but gave up when the complexity became too much. Last November, he discovered AI coding tools and found MeDo through a Google search. After his first session, he was hooked. Since then, he's built multiple games with high production quality and sophisticated mechanics — the kind of projects that would have previously required a professional development team. Then there's Bubu, a medical escort service provider based in Guangzhou, China. Working on the front lines of that market, Bubu understood exactly what clients cared about most: a trustworthy, professional service with standardized processes and real-time updates on how appointments were progressing — especially for adult children living away from home who needed to stay informed about their parents' care. Using Miaoda, Bubu was able to create and launch a fully functional WeChat mini-program built around those exact needs — clean interface, clear workflows, and a seamless delivery experience. After launch, it attracted 8 new clients through organic traffic alone, generating over RMB 7,000 in additional revenue and meaningfully boosting overall business income. These examples point to a broader shift. When the barrier to building drops, more people experiment, and more ideas turn into real products. ## 🤝 AI for Good: Silent Guardian Miaoda and MeDo are also enabling projects with a deeper purpose. One example is Silent Guardian, an anti-fraud application designed for the hearing-impaired community. Hearing-impaired individuals are often more vulnerable to scams due to communication barriers and limited access to real-time verification tools. Using the platform, a single creator was able to build a comprehensive solution that includes fraud education, evidence collection, and real-time alerts. The app uses AI to convert speech into text and visual sign language cues, helping users better understand potential risks and respond in real time. For communities like this one, the value of a tool isn't just what it builds — it's who finally gets to use it. ## 🚀 Building Momentum: From Individuals to Teams As more users build with Miaoda and MeDo, we're seeing these tools move beyond individual use cases and into real-world environments. That shift is already visible in our developer ecosystem. Recent Miaoda hackathons in Beijing have brought together creators experimenting with new ideas and applications, showing how quickly simple prompts can turn into functional products. At the same time, this activity is expanding globally. The Build with MeDo Hackathon is now underway, inviting creators from around the world to push the boundaries of what can be built with no-code AI. Submissions are open through May 20, with top projects showcasing how quickly ideas can turn into fully functional applications. As the community grows, we're also bringing these tools into more structured environments. This month, we introduced the MeDo Enterprise Version, designed to bring the same no-code, natural language development experience to teams. With a structured workspace for enterprises, teams, and individual members, it enables collaborative building while keeping data private and resources easy to manage. Taken together, these developments point to a broader transition, from individual experimentation to collaborative, real-world deployment across teams and organizations. ## 🏁 What's Next: Baidu Create 2026 As more people gain the ability to build, the definition of a "developer" is starting to change. Looking ahead, this shift will come into focus at Baidu Create 2026, our flagship developer conference in Beijing on May 13–14. For the first time, it will be held alongside the Yunzhi Summit (the company's annual AI Cloud event), bringing together developers, businesses, and industry leaders to explore the latest in AI and agents. This year's theme, "Agents at Scale," reflects a clear direction for the industry, moving from early experimentation to real-world deployment, and from building tools to building with autonomous, agent-driven systems. Here's a quick look at what else we've been working on this month across AI models and developer tools: > GenFlow 4.0 Launched at Baidu AI Day - GenFlow 4.0 is a major upgrade to our general AI agent, with a fully revamped Office Agent at its core. With a single instruction, users can now invoke PPT, Excel, and Word agents in parallel, handling the full range of workplace tasks in one place. - It is now deeply integrated with OpenClaw, and users can deploy it directly from the Baidu Drive PC or mobile app. > PaddleOCR Becomes the Most-Starred OCR Project on GitHub - PaddleOCR has officially become the #1 most-starred OCR repository globally on GitHub, surpassing long-standing benchmarks including Google's Tesseract. - Built on ERNIE foundation models, it offers a robust end-to-end pipeline for high-precision text recognition and structured document parsing across 110+ languages. > ERNIE-Image Open-Sourced for Developers - We released ERNIE-Image, an open 8B parameter text-to-image model that delivers strong performance across instruction following, multilingual text rendering, and structured visual generation, while remaining lightweight enough to run on consumer hardware. - ERNIE-Image comes in two versions: the SFT model optimized for stronger general quality in 50 inference steps, and ERNIE-Image-Turbo, optimized for speed and aesthetics in just 8 steps. > Famou-Agent 2.0 Sets a New SOTA on MLE-Bench - Famou-Agent 2.0 has once again ranked #1 on MLE-Bench, setting a new SOTA for the second time since it first topped the benchmark last October. Famou-Agent is a general-purpose multi-agent framework trusted by thousands of enterprises across manufacturing, finance, transportation, and beyond to tackle complex real-world challenges. - The 2.0 upgrade brings key improvements across evolution strategies, long-horizon memory, and infrastructure. A full reveal is coming at Baidu Create next month — stay tuned. Have a question about building with MeDo, or something you'd love us to cover next? Leave a comment or DM us! Until our next roundup, keep up with our latest AI developments and innovations by following us on LinkedIn and X.

译SBTI迷因测试的流行，揭示了低门槛参与的趋势正驱动AI应用开发工具的演进。Miaoda及其国际版MeDo是一个生成式AI驱动的无代码对话式应用开发平台，用户仅需通过自然语言描述，即可在几分钟内获得功能完整、可部署的应用，无需编写代码。平台背后由10多个专用AI代理协作，覆盖从需求分析到部署的全流程。传统上全球仅0.4%的人口是专业程序员，而该平台已助力创建超50万个商业应用，其中81%的创建者是非程序员，服务超1000万用户。这标志着开发方式从编写代码转向描述意图的根本性变革。

向阳乔木@vista8 · 4月30日51

如果你用Dia 浏览器，可以新开一个Tab，对话框中输入：coffeeonjosh，体验这个新功能。 会自动连接Gmail，也可以连接Notion等工具，制作一份晨间早报。

Chubby♨️@kimmonismus · 4月30日36

$4290 for a dual-arm robot. Holy.

译4290美元买一个双臂机器人。 天啊。

Alibaba Cloud@alibaba_cloud · 4月30日44

HiClaw v1.1.0 is here! 🚀 🔹 K8s Native: Production-ready Helm chart with HA & RBAC. 🔹 Hermes Support: Autonomous coding agent as a first-class worker. 🔹 Lightweight: Image size reduced by ~1.7GB. 🔹 Enterprise Ready: Multi-tenant isolation & pluggable storage. Upgrade now for scalable AI orchestration! Learn more: https://int.alibabacloud.com/m/1000412675/ #HiClaw #Kubernetes #AI

译HiClaw v1.1.0 现已发布！🚀 🔹 K8s 原生：具备高可用和 RBAC 的生产就绪 Helm chart。 🔹 Hermes 支持：自主编码代理作为一等工作者。 🔹 轻量化：镜像大小减少约 1.7GB。 🔹 企业就绪：多租户隔离与可插拔存储。 立即升级以获得可扩展的 AI 编排能力！ 了解更多：https://int.alibabacloud.com/m/1000412675/ #HiClaw #Kubernetes #AI

Orange AI@oran_ge · 4月30日39

大家发在 X 上的提到了 ColaOS 的帖子，都会被官网自动收录了 欢迎大家多多分享使用案例 😄 吐槽...也行吧 https://colaos.ai/community

Alibaba Cloud@alibaba_cloud · 4月30日68

Introducing HappyHorse, the latest breakthrough from Alibaba Cloud. Key Features Demonstrated: - Cinematic 1080p Quality: Crystal clear visuals that breathe life into your ideas. - Native Audio-Visual Sync: Perfect lip-sync and sound alignment generated instantly. - Multi-Shot Consistency: Maintain character identity across complex scenes and camera movements. - Instant Generation: Go from prompt to production in seconds. Try HappyHorse for FREE today: https://int.alibabacloud.com/m/1000412663/

译阿里云推出最新AI视频生成模型HappyHorse。该模型具备多项突破性功能：可生成影院级1080p高清画质；原生实现精准的音画同步，确保口型与声音对齐；在复杂场景和镜头切换中保持多镜头角色一致性；支持从文本提示到视频成片的秒级即时生成。用户现可免费试用。

宝玉@dotey · 4月30日35

Hermes Agent 加了一个新功能：Curator（策展人）。 Hermes Agent 有个“自我进化”机制，用着用着会自己创建各种技能文件来优化工作流。但技能攒多了就成了数字垃圾抽屉，重复的、过时的、太细碎的全堆在一起。Curator 干的就是定期收拾这个抽屉。 它默认每周自动跑一次，做几件事：统计每个技能的使用频率和更新时间，把功能重叠的技能合并，把太久没用的清理掉，把过于具体的小技能降级成模板或脚本、塞进更通用的技能里。外部安装的技能、内置技能、以及你手动"钉住"的技能它不会碰，只动智能体自己生成的和用户手写的。 不想让它自动跑也行，在 config.yaml 里关掉就好，或者用 hermes curator run 手动触发。 也算是对于技能越来越多的一个痛点解决方案，至于效果如何，可以观察看看。

译Hermes Agent新增Curator功能，旨在解决其“自我进化”机制持续生成技能文件导致的技能库臃肿问题。Curator会每周自动运行，追踪技能使用频率与更新时间，自动合并功能重叠的技能、清理闲置技能，并将过于具体的技能转化为模板或脚本整合到更通用的技能中。该功能仅处理智能体自动生成和用户手写的技能，不会影响外部安装、内置或被用户“钉住”的技能。用户也可在配置文件中关闭自动运行，或通过命令行手动触发。

歸藏(guizang.ai)@op7418 · 4月30日37

我操，Codex 太牛逼了！ 自己给我做了一个类似于《杀戮尖塔》的爬塔游戏，从代码到素材全是自己搞。 我就跟他说了一个要做类似《杀戮尖塔》的游戏，要中国风格的 这是能玩的！

小互@xiaohu · 4月30日59

Stripe 发布 Link agent wallet 让 AI agent 可以代表你发起付款，但不会拿到你的真实银行卡信息。 AI agent 可以为你创建一次性支付凭证，用完就失效。 你会收到提醒，并在 Link 里批准或拒绝每一笔购买。 真实卡号不会交给 agent，也不会暴露给它正在操作的工具。 Link CLI 已经开源，开发者可以把它接进 Claude Code、MCP 客户端或自己的 agent 工作流。

译Stripe发布Link agent wallet，使AI agent能代表用户发起支付，同时保护银行卡信息安全。其核心机制是让AI agent生成一次性支付凭证，每笔交易需用户在Link中手动批准或拒绝。真实卡号既不会提供给AI agent，也不会泄露给其调用的外部工具。此外，Stripe已开源Link CLI，开发者可将其集成至Claude Code、MCP客户端或自定义的agent工作流中。

Sam Altman@sama · 4月30日21

goblinblog dropped

译goblinblog 发布了

Kling AI@Kling_ai · 4月30日39

🔍 Zoom in. Every tiny detail you've been missing. Kling 4K doesn't blink. And neither will you.

译🔍 放大。每一个你曾错过的微小细节。 Kling 4K 不眨眼。你也不会。

Alibaba Cloud@alibaba_cloud · 4月30日60

Did you know that Wan2.7‑Image also has its own “color palette” feature? 😉 By just inputting reference images, exact color codes, or even inserting your own palette, the model ensures color code matching for brand consistency and overcomes inconsistent color reproduction that has long plagued AI imagery 🎨 Watch the video and control your colors perfectly now! #AlibabaAI #Wan #Innovation

译你知道吗？Wan2.7‑Image 也有自己的“调色板”功能？😉 只需输入参考图像、精确色码，甚至插入自己的调色板，该模型就能确保色码匹配以保持品牌一致性，并克服长期困扰 AI 图像的色彩还原不一致问题 🎨 观看视频，立即完美掌控你的色彩！ #AlibabaAI #Wan #Innovation

Alibaba Cloud@alibaba_cloud · 4月30日53

The agent economy needs open payment rails — just as the internet needed HTTP. Today, @okx launches the Agent Payments Protocol (APP): the first open standard for how AI agents pay, get paid, and do business with each other — across every chain. @AlibabaCloud and OKX have been close partners in cloud infrastructure and AI. Now we're taking it further — helping build the financial layer of the agent economy together. As a founding partner of APP, we couldn't be more excited. The AI era is just getting started, and we look forward to going deeper with OKX. The payment rails are open. Let's see what agents build next. #AlibabaCloud #OKX #APP #AgentEconomy

译AI代理经济需要开放的支付通道，正如互联网依赖HTTP协议。OKX推出首个跨链开放标准Agent Payments Protocol（APP），为AI代理之间的支付与商业交互提供基础框架。OKX与AlibabaCloud的合作从云基础设施与AI领域进一步延伸，共同构建代理经济的金融层。作为APP创始合作伙伴，双方认为AI时代刚刚起步，开放支付通道将激发代理生态的创新潜力。

TestingCatalog News 🗞@testingcatalog · 4月30日62

XAI 🚨: A new Imagine Agent is rolling out on Grok web in Beta! Imagine Agent can generate images and videos on an open canvas view. It can perform quite complex tasks, such as "generate a 1-minute film", "generate UGC product stories", "generate a complete manga set", and more. This is quite a huge release! Did you get it too?

译XAI 🚨：全新的 Imagine Agent 正在 Grok 网页版 Beta 测试中推出！ Imagine Agent 可以在开放画布视图中生成图像和视频。它能执行相当复杂的任务，例如“生成一部1分钟的电影”、“生成 UGC 产品故事”、“生成完整的漫画集”等等。 这是一个相当重大的发布！你也用上了吗？

Berryxia.AI@berryxia · 4月30日55

越来越方便了，Agent来帮你完成Cloud Flare 中的创建账号，启动付费订阅，注册域名… 任务。

Rohan Paul@rohanpaul_ai · 4月30日8

Japan just showed off a drone built completely from cardboard. It flies at 120km/h, comes together in 5 minutes, and is made for use in large swarms. the most interesting point bit is that it can be mass produced by any ordinary cardboard factory.

译日本刚刚展示了一款完全由纸板制成的无人机。它以120公里/小时的速度飞行，5分钟内即可组装完成，专为大规模集群使用而设计。 最有趣的一点是，它可以在任何普通纸板工厂进行大规模生产。

阿绎 AYi@AYi_AInotes · 4月30日70

Google的AI大模型终于有点像样的的更新更新，这波直接让不少靠AI生成文档的初创公司感受到巨大压力🤯👀 Gemini现在可以在聊天里直接生成Docs、Sheets、Slides、PDF、Word、Excel等主流办公文件， 一句话prompt就能下载， 完全不用复制粘贴、不用手动排版、不用调格式。 支持的范围非常实用：包括带LaTeX公式的学术文档、表格、图表等。 
最震撼的是官方演示——
上传两张手写的潦草笔记，说一句“转成带图表的统计学学习指南PDF”， 几秒钟后，一个标题、目录、公式、表格、视觉元素全都专业排版的完整文档就生成了，直接就能下载。 而且这个功能免费给全球所有Gemini App用户开放，没有Pro门槛，没有地区限制，任何人现在打开手机就能用。 以前的AI，本质上还是聊天机器人：它输出的是文字，你还得自己花时间加工成能交付的东西。 
从这波更新开始，AI输出的不再只是文字，而是可以直接交付的生产力资产——它从“陪你聊天的助手”，升级成了“直接替你干活的员工”。 Google最狠的地方，其实不是模型技术本身，而是把过去二十年积累的Workspace生态，直接变成了Gemini的原生输出底座。 
别人还在卷“谁生成的文字更通顺”，Google直接卷“谁的输出能直接发给老板或客户”。 
那些靠AI生成PPT、PDF、表格吃饭的工具，今天起就从“必备”变成了“可选中间层”，商业模式不得不重新思考。 这才是真正的降维打击。 我理解AI的下一个战场，已经从“谁更会聊天”，彻底转向“谁能直接产出可交付的成果”。 
而Google，借着Workspace的深度集成，先手拉开了一大截领先优势。 对普通用户来说：以后写报告、做PPT、整理笔记，再也不用半夜改格式了，生产力直接起飞。 
对行业来说：这标志着AI从“对话时代”正式进入“交付时代”。

译Google Gemini迎来重磅更新，用户现可在聊天中通过一句话指令，直接生成并下载Docs、Sheets、Slides、PDF等主流办公文件，无需手动复制排版。该功能支持含LaTeX公式的学术文档、表格和图表，且免费向全球Gemini App用户开放。这标志着AI从输出文字的“对话时代”，迈向了直接产出可交付生产力资产的“交付时代”。Google凭借与Workspace生态的深度集成，实现了降维打击，对依赖AI生成文档的初创公司构成巨大压力，并推动行业竞争焦点转向直接产出可用成果。

OpenClaw🦞@openclaw · 4月30日58

OpenClaw 2026.4.27 🦞 🧠 DeepInfra provider 📎 better file attachments 🛡️ operator-managed proxy routing 🧭 stricter model selection + local model fixes 🔧 gateway, channel, and session reliability Ships more than it brags. https://github.com/openclaw/openclaw/releases/tag/v2026.4.27

译OpenClaw 2026.4.27 🦞 🧠 DeepInfra 提供商 📎 更好的文件附件功能 🛡️ 运营商管理的代理路由 🧭 更严格的模型选择 + 本地模型修复 🔧 网关、通道和会话可靠性 实干多于自夸。 https://github.com/openclaw/openclaw/releases/tag/v2026.4.27