Nathan Lambert@natolambert · 5月26日63

Gemma 4 adoption numbers outpacing Qwen 3.5/3.6 for the same sized models is a big shift in the international balance of influence via open models.

译Gemma 4 在同规模模型上的采用率超过通义千问 3.5/3.6，标志着开源模型国际影响力格局的重大转变。

Google DeepMind@GoogleDeepMind · 5月26日34

Our Gemini for Science tools could help scientists unlock their next breakthrough. 🧬

译我们的 Gemini for Science 工具可以帮助科学家实现下一个突破。🧬

Chubby♨️@kimmonismus · 5月26日59

I'm not sure if Google is winning the AI race. However, I think they're winning the AI distribution race, which is a different thing. 900M Gemini users is impressive on a slide. But a huge chunk of that is Android users who got a default app swap and Search users who got AI Overviews without opting in. But that doesnt mean its a bad thing. 9.7 trillion tokens/month two years ago. 480 trillion last year. 3.2 quadrillion now. That's a 7x jump in twelve months. To keep that going, Google plans to spend $190 billion on infrastructure this year. OpenAI has been trying to reach the 1b user milestone for some time now. For Google, on the other hand, it's a simpler game. Why? With billions of Android devices, and combined with Google and its AI mode, they have the ability to introduce everyone to AI, specifically Gemini, for free. How do they do it? TPUs! Google not only laid the foundation for modern LLMs with their 2017 paper "Attention is all you need," but also made a far-sighted decision back in 2012 to invest in TPUs - their own in-house chips that are particularly well-suited for machine learning tasks. Now in its eighth iteration, they even have two chips: one particularly good for inference, and one particularly good for training. This makes them more independent. Furthermore, they have a solid foundation that generates strong revenue and good profits, allowing them to subsidize AI usage for free, and without ads, unlike OpenAI (this is not a judgment, just a statement of fact). TPUs Therefore, Google has a very good chance of winning the game thanks to this outstanding starting position and free distribution. But to be fair: the game *is* far *from over*. However, the starting position is outstanding for Google. Image: The Economist article

译文章的核心论点是 Google 凭借其分发优势，在 AI 分发竞赛中占据了有利位置。目前 Gemini 拥有 9 亿用户，这主要归功于向 Android 用户进行的默认应用替换，以及向 Google 搜索用户推送的 AI 概览。其大语言模型 token 用量在 12 个月内从 480 万亿增长至 3.2 千万亿。为支撑此规模，Google 计划今年投入 1900 亿美元用于基础设施。Google 的关键优势在于能够利用庞大的 Android 设备基础，通过其搜索和 AI 模式免费向用户推广 Gemini。这一策略的部分成本优势源于自研的 TPU 芯片，使其在推理和训练上更独立，并能基于自身盈利补贴免费 AI 服务。尽管游戏远未结束，但 Google 的开局位置非常出色。

Rohan Paul@rohanpaul_ai · 5月26日57

New Meta, Stanford, Google and many other top labs paper proposes AutoResearchClaw. Shows that automated research improves when AI can fail, recover, and ask humans at the right moments. The paper is less about an “AI scientist” than about turning research into a governed loop. Most systems still treat science like a production line: generate an idea, run code, write a paper, then stop when the chain breaks. AutoResearchClaw treats failure as evidence, using debate, repair, verification, memory, and selective human input as parts of the same machine. That is the main point: autonomy gets better when it is constrained by process, not when it is simply given more freedom. On ARC-Bench, the system beat AI Scientist v2 by 54.7%, with its sharpest gains in result analysis, where claims had to match measurements rather than merely sound plausible. The human result is more interesting: CoPilot reached an 87.5% accept rate, while full autonomy reached 25% and step-by-step oversight reached 50%, suggesting that too little judgment and too much supervision can both degrade science. The most revealing failure was a case where every cross-validation method returned identical zero-bias outputs, which passed numeric verification but failed scientific meaning. That is the boundary this paper exposes: machines can verify that numbers are real, but humans still notice when the experiment has stopped asking the right question. ---- Paper Link – arxiv. org/abs/2605.20025 Paper Title: "AutoResearchClaw: Self-Reinforcing Autonomous Research with Human-AI Collaboration"

译Meta、斯坦福等机构提出AutoResearchClaw，这是一个通过AI智能体进行自主研究的框架。其核心理念是将科研过程转化为一个受流程约束的循环，而非简单的生产线。系统整合了辩论、修复、验证、记忆和选择性的人类反馈，并将失败视为有效证据。在ARC-Bench基准测试中，该系统在结果分析等任务上性能比AI Scientist v2提升54.7%。人类协作实验显示：CoPilot模式（适时介入）接受率达87.5%，完全自主仅25%，逐步监督为50%。一个关键失败案例揭示了当所有交叉验证方法返回相同零偏差输出时，系统虽通过数值验证却失去了科学意义，凸显了人类判断的关键作用。

Google DeepMind@GoogleDeepMind · 5月26日71

SynthID has already watermarked over 100 billion pieces of content, but transparency is a team sport. That’s why we’re partnering with @OpenAI, @ElevenLabs and Kakao to add SynthID watermarking to their models – accelerating the industry-wide momentum we started with @NVIDIA.

译SynthID已为超过1000亿条内容添加水印，但透明度需要行业协作。 因此，我们正与OpenAI、ElevenLabs和Kakao合作，将SynthID水印技术集成到他们的模型中——这延续了我们与NVIDIA共同推动的全行业趋势。

ginobefun@hongming731 · 5月26日68

Turing Post 概述了构建长时间运行 AI Agent 的 5 种模式：检查点与恢复、委派审批、分层记忆上下文、后台处理以及集群编排。

译Turing Post 总结了构建长时间运行 AI 智能体的 5 种关键模式，旨在解决生产环境中的运行挑战。这 5 种模式分别是：检查点与恢复（按批次保存进度以实现故障恢复）、委派审批（智能体可暂停流程并保持上下文等待人类响应）、分层记忆上下文（分离工作与长期记忆并实施严格治理）、后台处理（通过后台智能体实时响应事件而非硬编码策略）以及集群编排（使用协调器编排专业智能体以实现独立运行与扩展）。

Rohan Paul@rohanpaul_ai · 5月26日69

New Google paper says LLMs should stop pretending certainty and instead clearly show when they are unsure. Hallucination is less about machines being wrong than about machines sounding certain when they should hesitate. That distinction changes the target-problem. The paper changes the target from making models perfectly factual to making them honest about their own uncertainty. For years, the obvious goal has been to make language models know more, so they make fewer factual mistakes. Perfect factuality may be very hard, but a model that clearly separates “I know this” from “I am guessing” can stay useful without quietly damaging trust. This paper argues that the harder missing skill is not knowledge, but self-knowledge. A model can be well calibrated in the broad sense, knowing that answers like this are correct about 60% of the time, yet still fail to identify which particular answer is the dangerous one. That is the trap: to eliminate errors, the system must refuse many answers that would have been right. The authors call this the utility tax, and it explains why products keep drifting toward confident usefulness rather than cautious truth. Here's the key point. A wrong answer wrapped in honest uncertainty is not the same social object as a wrong answer delivered as fact. It gives the user a different instruction: verify this, treat it as provisional, do not build too much on it. The proposed fix is “faithful uncertainty,” where the model’s language mirrors its internal confidence instead of smoothing doubt into authority. For agents, this becomes even more important, because uncertainty is what should decide when to search, when to trust a source, and when to stop. Tools expand what a model can access, but metacognition governs whether access is used wisely. ---- Paper Link – arxiv. org/abs/2605.01428v1 Paper Title: "Hallucinations Undermine Trust; Metacognition is a Way Forward"

译Google最新论文指出，LLM的幻觉问题核心在于模型在该犹豫时仍表现确定，而非单纯事实错误。论文将优化目标从追求完美的事实准确性，转向让模型能诚实地区分“我确知”与“我猜测”。作者提出了“忠实不确定性”概念，要求模型的表述与其内部置信度相符。文章还引入了“效用税”概念，解释了为何产品倾向自信但可能错误的回答。对于智能体而言，元认知能力至关重要，它决定了何时调用工具、何时信任信息源。

Chubby♨️@kimmonismus · 5月26日79

Apple isn’t just adding Gemini to Siri. It is reportedly using a custom 1.2T-parameter Google model as the brain behind parts of the next Siri overhaul (Reuters). That is no small size. The question, therefore, is how Apple’s Gemini will perform and how fast it runs. In particular, simple queries are expected to run locally. Gemini 3.5 Flash is estimated to have around 300 billion parameters; Apple’s model will thus be significantly larger. However, the question also remains whether size actually pays off in this context. Apple’s model must deliver answers to everyday queries quickly-and be fast enough while doing so. Anyways, next months will be exciting in so many ways: -WWDC / Apple Intelligence with Gemini - GPT-5.6 - Sonnet4.8/Opus 4.8 (probably) -Gemini 3.5 pro (confirmed) Super hyped!

译据报道，苹果为改造下一代Siri，正使用一个定制版、参数规模达1.2T的Google大模型作为其核心，这显著大于预估约300B参数的Gemini 3.5 Flash。该模型将驱动Siri的部分功能，其中简单查询预期会在本地设备运行。苹果面临的关键挑战是确保该大模型能够足够快速地响应日常问题。此外，下个月AI领域预计将有多项重要发布，包括WWDC上的Apple Intelligence与Gemini整合、GPT-5.6、可能的Sonnet 4.8/Opus 4.8，以及已确认的Gemini 3.5 Pro。

🚨 AI News | TestingCatalog@testingcatalog · 5月25日58

GOOGLE 🔥: Antigravity now has Gemini 3.5 Flash (Low) option available with a 45% fewer token consumption in comparison to Medium variant. This addition follows Gemini quota reset as well!

译Antigravity平台新增Gemini 3.5 Flash (Low)选项，其token消耗相比Gemini 3.5 Flash (Medium)版本减少约45%。据内部测试，该新选项在SWE任务上的表现通常优于上一代的Gemini 3 Flash (High)。此举旨在优化简单任务的token使用，同时所有付费计划的Gemini配额已重置。

ginobefun@hongming731 · 5月25日49

#BestBlogs 早报 05-26 今天 BestBlogs 早报精选 10 篇，3 篇深度精讲： 📌 Google 把 FDE 面试压缩到 2 天、OpenAI 斥资 40 亿成立独立部署公司，AI 商业竞争主战场已从训练转向落地； 📌 Anthropic 研究 PM 首度揭秘下一代 Claude 的「dreaming 机制」与意识研究； 📌 以及一篇反直觉洞察：超级个体不是被培训出来的，而是被完整 Closed-loop 激发的。 听播客 / 读图文版，链接在 http://BestBlogs.dev 👇

译早报精选10篇文章并精讲3篇：包括Google与OpenAI将AI商业竞争焦点转向部署落地，Anthropic研究PM揭秘下一代Claude的「dreaming机制」，以及关于“超级个体”是通过完整Closed-loop被激发而成的反直觉观点。

🚨 AI News | TestingCatalog@testingcatalog · 5月23日38

ICYMI 👀: Gemini in Chrome is now available to some users in Europe. I finally got Gemini for Chrome feature here in Germany on my Canary build, with Gemini 3.5 Flash, Skills, and Gemini Live. Does it work for you as well?

译如果你错过了 👀：Gemini在Chrome浏览器中现已向部分欧洲用户开放。 我终于在我的Canary版本中获得了德国的Chrome Gemini功能，包含Gemini 3.5 Flash、技能和Gemini Live。 你也能用了吗？

Alibaba Cloud@alibaba_cloud · 5月23日61

The velocity of the Qwen3.7-Max development is unreal. This is what relentless innovation looks like. #AlibabaCloud #Qwen

译阿里云Qwen团队新发布的Qwen3.7-Max模型在极短时间内（不到一个月）实现了多模态生成能力的惊人进步。独立测试显示，该模型已从此前表现落后，跃升至在特定测试中与Gemini 3.5 Flash持平，并超越了GPT-5.5与Claude Opus 4.7。其渲染的图像（如足球运动员与足球）在比例和真实感上表现尤为突出，展现出卓越的空间推理能力。

Google DeepMind@GoogleDeepMind · 5月23日60

We’re expanding our partnership with Singapore to help safely deploy AI at scale. 🇸🇬 Together with country experts, our new programs will focus on accelerating scientific discovery, advancing pandemic preparedness, and improving healthcare. Find out more → https://goo.gle/49jGwjv

译我们正在扩大与新加坡的合作，以帮助安全地大规模部署AI。🇸🇬 与各国专家合作，我们的新项目将重点加速科学发现、加强大流行病防范并改善医疗保健。 了解更多 → https://goo.gle/49jGwjv

Rohan Paul@rohanpaul_ai · 5月23日64

New Google paper shows that wearable data becomes far more useful when AI learns the person behind the signals. It's is not another heart-rate algorithm, but a general model trained on more than one trillion minutes of sensor data from five million people. The authors propose SensorFM, a foundation model trained on more than 1 trillion minutes of unlabeled wearable data from 5 million people, so it can learn general patterns of human physiology before seeing specific health tasks. That scale changes the problem from measuring isolated events to learning patterns of lived physiology: sleep, movement, temperature, oxygen, heart rhythms, and their ordinary daily messiness. Wearables are not weak because they lack data; they are weak because most systems compress that data into crude summaries before the meaningful structure has a chance to appear. SensorFM tries to learn that structure first, then reuse it across tasks, which is why the same representation can help with cardiovascular, metabolic, mental health, sleep, lifestyle, and demographic predictions. The evidence is strongest as a scaling story: larger models trained on more data performed better, and the learned embeddings beat engineered-feature baselines on 34 of 35 prediction tasks. ---- Paper Link – arxiv. org/abs/2511.15352v3 Paper Title: "People readily follow personal advice from AI but it does not improve their well-being"

译谷歌研究院提出基础模型SensorFM，通过学习超过500万人产生的逾1万亿分钟可穿戴设备传感器数据，掌握了人类生理活动的一般性模式。该模型超越了将数据压缩为简单指标的传统方法，能够从数据中提取出有意义的结构并将其复用于多种健康预测任务。实验显示，模型规模和数据量越大性能越强，且其学习到的数据表征在35项预测任务中的34项上，均优于基于工程特征的基线方法。

Google Gemini@GeminiApp · 5月23日87

http://x.com/i/article/2057551138384470016 # The Gemini app becomes more agentic, delivering proactive, 24/7 help ## Gemini is becoming a more helpful AI assistant, with an intuitive new UI, proactive daily briefs and Gemini Spark, an agent to help you get things done around the clock. It’s been a banner year for the Gemini app. Last year at Google I/O, Gemini was serving 400 million users. Today, more than 900 million people across 230 countries and more than 70 languages turn to Gemini for help every month. In time for Google I/O 2026, here’s what's new: - Gemini 3.5 Flash: The first in our next generation of models that combines frontier intelligence with lightning-fast action. - Neural Expressive: A vibrant, dynamic and completely reimagined design language for Gemini. - Gemini Omni: Our new model that can seamlessly transform text, images and video prompts into cinematic, high-quality video outputs. - Daily Brief: A new agent that gives you a personalized morning brief and organizes exactly what you need to know to start your day. - Gemini Spark: A 24/7 personal AI agent designed to proactively manage tasks and help you navigate your digital life, all under your direction. - MacOS app: Our desktop app will be integrating Gemini Spark so it can operate on your local machine, and it will also add powerful new voice features. ## Neural Expressive: A new design language for the AI era We’ve redesigned the entire Gemini experience from the ground up, introducing a stunning new design language we call Neural Expressive. The interface now features fluid animations, vibrant colors, new typography and haptic feedback. We’ve also integrated the Gemini Live conversational experience directly into Gemini. Now, you can seamlessly switch from typing a quick question to diving deep into a free-flowing conversation — and back again — without missing a beat. We also re-engineered the mic so you can tap and talk through a complex idea at your own pace without getting cut off mid-thought. And soon, we’ll start offering regional dialects, allowing you to choose a voice that truly resonates with you. Finally, we’re using the power of our Gemini models to make responses more engaging and easier to understand. Instead of throwing a wall of text at you, Gemini now designs tailored responses in real time — incorporating rich imagery, interactive timelines, narrated videos and dynamic graphics. Neural Expressive is rolling out globally across the web, Android and iOS for everyone. ## Gemini Omni: Turning your ideas into cinematic videos To unlock your creative potential, we're introducing Gemini Omni, a model designed to turn your imagination into reality. By seamlessly combining text, images and video inputs, Gemini Omni allows you to generate stunning, high-quality video outputs effortlessly. With Gemini Omni, video editing becomes a fluid, natural conversation. You can apply cinematic zooms or swap out backgrounds with a simple prompt. Just upload footage from your camera roll, apply built-in templates with a single tap and create polished content without expensive equipment or specialized technical jargon. You can even drop yourself directly into the action by creating a custom AI avatar that looks and sounds exactly like you. Gemini Omni is rolling out to Google AI Plus, Pro and Ultra subscribers worldwide. ## Daily Brief: Start your day on the right foot We’re introducing Daily Brief, an agent that gives you a personalized morning digest that’s designed to be your first stop every day. Built on the success of our recent Google Labs experiment CC, Daily Brief gives you a seamless, intuitive entry point into the world of AI agents. Once you opt in, Gemini works across your connected apps in the background. It gathers urgent updates from your Gmail inbox, tracks upcoming events from your Calendar and compiles relevant follow-up details into a skimmable briefing. It goes far beyond a simple summary. Daily Brief actively organizes and prioritizes based on your specific goals, even suggesting immediate next steps. You can easily steer it by giving responses a quick thumbs up or down over time. Daily Brief is rolling out to Google AI Plus, Pro and Ultra subscribers, starting in the U.S. ## Gemini Spark: From information to action We’re also introducing Gemini Spark, a 24/7 personal AI agent that helps you navigate your digital life. Spark represents a big shift for Gemini, transforming it from an assistant that can answer your questions into an active partner that does real work on your behalf and under your direction. Gemini Spark runs on Gemini 3.5 and uses the Antigravity harness. It’s deeply integrated with the Workspace tools you rely on daily, like Gmail, Docs, Slides and more. Even better, because it is a cloud-based agent, Spark keeps working in the background even when you close your laptop or lock your phone. That combination means Spark is ready to take complex tasks off your plate so you can be more present for what matters most. With Gemini Spark, you can: - Set recurring tasks or triggers: Automatically parse monthly credit card statements to flag new or hidden subscription fees. - Teach it new skills: Direct it to check your inbox for ongoing updates from your kids' school, extract critical deadlines and send a consolidated daily digest to you and your partner. - Create complete workflows: Ask it to synthesize raw meeting notes across emails and chats, create polished Google Docs with its findings and even draft the companion email kicking off a project. This is just the beginning. We’ve got a packed roadmap of features shipping over the summer. We’re expanding our list of Gemini connected apps with new MCP connections to Canva, OpenTable and Instacart launching today, and a full list of more partners are integrating now. In the coming weeks, Spark will be able to use these MCP connections to get things done for you. We'll also be adding new abilities, including texting and emailing Spark, creating custom sub-agents and operating your local browser. Spark operates under your direction. You choose whether to turn it on and what apps it connects to, and it’s designed to ask you first before performing high-stakes actions like spending money or sending emails. Gemini Spark will roll out to trusted testers this week, and we're planning to roll it out as a Beta for U.S. Google AI Ultra subscribers next week. ## Gemini app for macOS: Take control of your desktop We’re working on big updates to the Gemini app for macOS. We’ll be bringing Gemini Spark to the Gemini desktop app this summer so it can help with tasks involving your local files and automate workflows across your desktop. We’re also innovating on new voice experiences in the macOS app, similar to what we previewed at The Android Show. You won’t have to worry about all the “ums” or “what abouts” that happen as you think aloud. Using the context from your screen, Gemini can turn your free-flowing speech into precise drafts, instantly reformatting the text to capture your intent, right where your cursor is. The macOS app is available to download today for all users, with Gemini Spark and the new voice features will roll out later this summer. All of these updates get us closer to our vision of a truly universal assistant that’s personal, proactive and powerful. So whether you’re a busy student, parent or small business owner, we look forward to what you can do with Gemini.

译谷歌宣布Gemini应用月活用户已突破9亿。在此次更新中，Gemini正从工具演变为更主动的个人AI代理。主要更新包括新一代模型Gemini 3.5 Flash、全新的“Neural Expressive”设计语言，以及能将提示转化为高质量视频的Gemini Omni模型。核心亮点是两项代理功能：“Daily Brief”提供个性化每日简报，“Gemini Spark”则作为24/7的个人代理，在用户授权下主动管理任务与数字生活。这些更新标志着AI助手向更主动、更整合的方向发展。

Google DeepMind@GoogleDeepMind · 5月23日52

SynthID, our imperceptible watermark for AI-generated content, is expanding to more partners. We’re also adding new ways to find out if content was generated using AI - just ask in the @GeminiApp or in @Google Search.

译SynthID，我们为AI生成内容打造的不可见水印技术，正在扩展至更多合作伙伴。 我们还新增了识别AI生成内容的方式——只需在@GeminiApp或@Google搜索中提问即可。

Ethan Mollick@emollick · 5月23日76

I think people don't realize why Gemini Omni is different than other video AIs. It is fully multimodal, so it can edit video natively, too I took the famous "train " movie from 1896 & made it a bullet train, LEGO, added a time traveler, a centipede, muppets... (see reflections?)

译我认为人们没有意识到Gemini Omni与其他视频AI的不同之处。它是完全多模态的，因此也能原生编辑视频。 我拿了1896年著名的“火车”电影，把它变成了高铁、乐高，加入了时间旅行者、蜈蚣、布偶……（看到倒影了吗？）

Rohan Paul@rohanpaul_ai · 5月23日55

Sundar Pichai: - At the frontier labs competition is fierce - Only few labs are really at the frontier & then there is a big gap. - If recursive self-improvement emerges, we need more seriousness & it then becomes a societal issue, not one company’s call

译Sundar Pichai: - 在前沿实验室，竞争非常激烈 - 真正处于前沿的实验室很少，之后存在巨大差距 - 如果递归自我改进出现，我们需要更严肃对待，这将成为社会问题，而非一家公司能决定

Google AI@GoogleAI · 5月23日70

http://x.com/i/article/2057866361636888576 # I/O Developer News Recap ICYMI: here’s a quick rundown based on your vibe. (Pro tip: Bookmark this post!) For Building with Agents in Google @Antigravity: — Antigravity 2.0: A new standalone desktop application built for the next era of how we build, work, and automate with agents — Antigravity CLI: For those who prefer the terminal, this delivers a lightweight, high-velocity surface to invoke, monitor, and interact with agents — Antigravity SDK: Programmatic access to the Antigravity coding agent so you can customize agent behavior and deploy it on your own infrastructure of choice — Increased usage limits in Antigravity: We’ve upped the limits to keep developers in flow state For Bringing Ideas to Life in Google AI Studio: — Vibe Code Android Apps: @googleaistudio now has native Kotlin support to build full-stack Android apps. Test via Android Emulator and publish straight to the Test Track in @GooglePlayStore, all from AI Studio — Seamless Context Export to Antigravity: The new export flow brings your code and your context straight into Antigravity — AI Studio Mobile App: Bring your ideas to life while on-the-go with the new iOS and Android mobile app coming soon! For Building on Android & Web: — Managed Agents in the Gemini API: Build production-ready custom agents with the Antigravity agent harness and our Gemini models, fully hosted on our infrastructure. With one API call, you get the agent, environment and tools — Android CLI and skills: The @AndroidDev team introduced an Android CLI and open-sourced skills to make it easier to create high-quality, native Android apps with coding agents — WebMCP: The open web standard that allows you to expose web tools to browser-based agents is available in original trial in Chrome 149 — Chrome DevTools for agents: AI agents can now autonomously use DevTools to scale workflows for verifying, debugging, and optimizing code in real-time For Enterprise & More: — Google Cloud Availability: Antigravity can now be connected directly to Google Cloud projects. Antigravity will be supported in the coming months for existing Gemini Enterprise customers — Specialized Skills: Introducing new domain-specific skill bundles, starting with Science Skills from Google DeepMind to equip agents with specific science primitives to accelerate health, biology, and scientific research workflows.

译谷歌在I/O开发者大会宣布，系统性构建面向AI代理（Agent）的开发与部署工具链。核心更新包括：独立桌面应用Antigravity 2.0及其命令行工具、SDK面世；Google AI Studio新增Kotlin支持，可一键开发安卓应用并发布，同时推出移动端App。此外，Gemini API推出托管代理服务，实现一键部署；WebMCP作为开放标准在Chrome 149中推出，允许网页向代理暴露工具；Chrome DevTools也开放给AI代理以自动化调试。企业级客户可直接连接Google Cloud项目，而DeepMind的科学技能包则加速特定领域研究。此举标志着谷歌正全面打造从开发、接口到部署的完整AI代理生态系统。

Google Gemini@GeminiApp · 5月23日45

Built for the AI era. ✨ Meet Neural Expressive, Gemini’s new design language featuring fluid animations, vibrant colors, updated typography, haptic feedback, and visual responses.

译为AI时代而生。✨ 介绍Neural Expressive，Gemini的全新设计语言，其特点包括流畅的动画、鲜艳的色彩、更新的字体、触觉反馈和视觉响应。

Rohan Paul@rohanpaul_ai · 5月22日57

World models are moving into wild territory. Project Genie just turned Google Maps Street View into a promptable world simulator. Google AI Ultra users can now turn any real U.S. locations into interactive AI-made scenes.

译世界模型正进入未知领域。 Project Genie刚刚将谷歌地图街景转变为一个可通过提示词操控的世界模拟器。 谷歌AI Ultra用户现在可以将任何美国真实地点转化为交互式AI生成场景。

Google DeepMind@GoogleDeepMind · 5月22日67

Project Genie 🤝 @GoogleMaps Street View You can now take real U.S. places and transform them into new, interactive worlds. 🌍

译Project Genie 🤝 @GoogleMaps Street View 你现在可以将真实的美国地点转化为全新的交互式世界。🌍

Chubby♨️@kimmonismus · 5月22日41

June will be huge. -Gemini 3.5 pro (confirmed) -GPT-5.6 (rumored but pretty confident for a release) Still waiting for annoucements Claude Sonnet 4.8 (Claude-Code-/Source-Map-Leak)

译六月将是巨大的一个月。 - Gemini 3.5 pro（已确认） - GPT-5.6（传闻中，但发布可能性很高） 仍在等待官方公告 Claude Sonnet 4.8（Claude-Code-/Source-Map-Leak）

Rohan Paul@rohanpaul_ai · 5月22日60

Google was not far from launching its own ChatGPT-like product back in 2022. Sundar Pichai talks about how Google had already built a almost equvalent product around LaMDA, but its raw behavior was too toxic for Google standards, and unreliable to ship broadly. Pichai: "I think as a company, you had this search quality bias, and so, you know, we had a higher bar, for what we thought was an acceptable product quality to go out." The LaMDA model of Google was not simply a case of brilliant research sitting on a shelf, because Google had already started turning the model into a product and even exposed a constrained version through AI Test Kitchen. Pichai is also saying hindsight makes the miss look cleaner than it felt at the time, because even OpenAI’s move was not instantly obvious as the winning form factor. --- From @collision podast with @sundarpichai (full video link in comment)

译谷歌CEO桑达尔·皮查伊透露，公司在2022年已基于LaMDA模型开发出接近ChatGPT的产品，但因输出内容存在毒性且不符合内部质量标准而未能广泛发布。皮查伊指出，公司存在“搜索质量偏见”，对产品的可靠性和安全性设有更高门槛。LaMDA模型并非仅停留在研究阶段，谷歌曾通过AI Test Kitchen进行有限测试。他补充道，从后视镜看，OpenAI的成功在当时并非显而易见，这使得谷歌的决策过程更为复杂。

Rohan Paul@rohanpaul_ai · 5月22日70

Great news for devs from Google. Every paid Antigravity tier now gets permanent 3x Gemini rate limits.

译Google为开发者带来好消息。 所有付费Antigravity层级现在都将获得永久的3倍Gemini速率限制。

Google AI Developers@googleaidevs · 5月22日58

Update: *Weekly* Gemini quotas on all paid @Antigravity tiers have just been tripled (again!), and quotas are officially reset (again!).

译为回应用户对每周配额用尽的担忧，Antigravity平台已将所有付费层级的Gemini每周配额再次提升至三倍，并已完成配额重置。此前该平台曾三倍提高限制以支持用户更持续地开发，此次更新旨在进一步保障用户的使用空间。

Google AI Developers@googleaidevs · 5月22日70

Update: *Weekly* Gemini quotas on all paid @Antigravity tiers have just been tripled, and quotas are officially reset.

译更新：所有付费@Antigravity层级的*每周*Gemini配额已再次提升至三倍，且配额已正式重置。

Sundar Pichai@sundarpichai · 5月22日63

Excited to see what you are building with Antigravity, We just 3xed the Antigravity limits again, but this time, the weekly quotas. Don't stop building!

译该通知宣布Antigravity的限额再次提升至原有标准的三倍，此次调整覆盖每周配额上限。同时，针对用户反馈的“几轮工作后可能触及每周配额上限”的顾虑，平台已为所有付费计划进一步三倍提升每周Gemini配额，并同步重置了当前配额。此举旨在为开发者提供更充足的使用空间，鼓励其持续投入项目构建。

Google Gemini@GeminiApp · 5月22日82

Gemini Omni is here, and we’ve been seeing amazing creations all week. Here are some standouts 👇

译Gemini Omni来了，我们本周看到了许多令人惊叹的创作。以下是一些杰出作品 👇

Berryxia.AI@berryxia · 5月22日49

早上刚睁眼，你还没来得及刷通知和邮件。 Gemini 已经主动把今天最该关注的内容，整理成一份简短清晰的 Daily Brief 发给你。 它会扫描你的邮件、日程和最新动态，挑出真正重要的事项，告诉你今天该把精力放在哪里。 你边喝咖啡边看这份清单，就已经对一天的重点心里有数。 这功能现在对 Google AI Plus、Pro 和 Ultra 用户开放，美国地区先上线。 牛马们需要邮件助理嘛😂

译Google Gemini上线Daily Brief功能，能在早晨主动扫描用户的邮件、日历与最新动态，提炼出当日需优先处理的事项，生成一份简洁的待办清单，帮助用户快速把握全天重点。该功能目前面向Google AI Plus、Pro和Ultra用户在美国地区开放。

Orange AI@oran_ge · 5月22日35

我想吐槽一下这些模型厂的发布，能不能把基本的东西都直接写出来 比如 qwen max 3.7 max 多大参数，价格多少 比如 gemini omni 价格多少 这些最基本的东西都不写，谷歌也搜不到，还要去官网查半天

译用户批评部分AI模型厂商在发布新模型时，未公开关键基本信息，如Qwen Max 3.7 Max的参数规模、定价，以及Gemini Omni的价格。这些基础信息缺乏公开透明，导致用户难以通过常规搜索获取，需额外花费时间查询官网，影响了信息获取效率与体验。

🚨 AI News | TestingCatalog@testingcatalog · 5月22日65

GOOGLE 🔥: Pomelli can now generate and host a landing page website based on your Brand DNA in 2 clicks! Brand Books generation is now available as well. Testing time 👀

译Google为Pomelli平台引入了“代理”能力，旨在帮助企业更便捷地创建符合品牌调性的内容。用户现在可以从现有文件和图片中提取信息来构建品牌的“DNA”，并基于此在两次点击内生成并托管一个着陆页网站。此外，平台还新增了品牌手册的生成功能。这些更新标志着品牌内容创作正朝着更自动化、个性化的方向发展。

Google AI Developers@googleaidevs · 5月22日58

Rate limits for Gemini on all paid @Antigravity tiers just permanently tripled. We also hit refresh on everyone's weekly quotas. ⤵️

译所有付费Antigravity层级的Gemini速率限制已永久提升三倍。我们还重置了所有人的每周配额。⤵️

Josh Woodward@joshwoodward · 5月22日18

What a ride for @GoogleLabs at IO! Hope you're enjoying lots of the new stuff!

译@GoogleLabs在IO大会上真是大放异彩！希望你们喜欢这些新功能！

Josh Woodward@joshwoodward · 5月22日13

♥️ ♥️♥️ iOS, Android, and Web - so fun to see people loving the new Neural Expressive design!

译♥️ ♥️♥️ iOS、Android和Web平台——看到大家喜欢新的神经表达式设计真是太棒了！ [引用 @immasiddx]：Gemini的iOS应用看起来比ChatGPT和Claude好多了。 Google这次绝对搞砸了。🔥

Google Gemini@GeminiApp · 5月22日67

Gemini can now connect to even more apps, including @OpenTable, @Canva, and @Instacart. Whether you're booking a table at a restaurant, creating a flyer, or ordering groceries, Gemini doesn’t just find info, it helps you take action seamlessly with connected apps.

译Gemini现在可以连接更多应用，包括@OpenTable、@Canva和@Instacart。 无论您是预订餐厅、制作传单还是订购杂货，Gemini不仅能查找信息，还能通过连接的应用帮助您无缝采取行动。

Sundar Pichai@sundarpichai · 5月22日72

Really enjoyed this conversation with @MatthewBerman at I/O 👇

译谷歌CEO Sundar Pichai在I/O大会上与MatthewBerman进行了深入对话，全面探讨了当前人工智能发展的多个核心方面。Pichai重点关注了AI代理如何可能取代部分互联网功能，并改变用户的信息获取方式，甚至威胁原始互联网的形态。对话还深入讨论了开源AI模型面临的商业模式困境、中美在AI领域的激烈竞争格局，以及网络安全威胁，包括AI可能引发的攻击和谷歌的应对策略。此外，Pichai指出谷歌正面临巨大的AI算力需求，计算资源已成为主要瓶颈，并探讨了发布强大AI模型的阈值判断问题。他强调了谷歌致力于开发低成本、高效率的AI模型的原因，以应对资源挑战。整体而言，这次对话覆盖了AGI竞赛、AI代理、信息饮食等关键议题，深入分析了AI发展的挑战、机遇和未来趋势。

Google Gemini@GeminiApp · 5月22日63

Get a head start on your day with Daily Brief. Gemini can now proactively flag what matters most in an easily digestible to-do list, so you’re ready for the day before you even finish breakfast.

译用Daily Brief开启高效一天。 Gemini现在能主动将最重要的事项整理成易于理解的待办清单，让你在早餐结束前就为一天做好准备。

歸藏(guizang.ai)@op7418 · 5月22日52

剪映海外版宣布跟 Gemini 合作了 未来可以在这边里使用剪映的编辑功能去编辑图像视频，但是没有说具体的合作方式和上线时间

Berryxia.AI@berryxia · 5月22日51

兄弟们，终于Google AI Studio 手机版来了。 官方已经上架至iOS APP Store商单。（非国区） 以及Google Play 商店即将上架。 预计上架时间7月1日。 iOS地址：https://apps.apple.com/us/app/google-ai-studio/id6761388855

译Google正式推出了AI Studio的原生移动应用，同时支持iOS和Android平台。目前iOS版本已上架App Store（非中国区），Android版本预计于7月1日在Google Play上架。此次推出旨在将AI开发体验带给更广泛的移动用户，并重塑了其“vibe coding”体验，使产品界面更加简洁易用。