Discover the AI Tools

腾讯AI创意智能体平台Miora正式全面上线。该平台突破传统单点AI工具局限，首次实现长期记忆、深度需求理解与多智能体协作能力，旨在为创意工作者提供从灵感到成品的全流程解决方案。其核心优势在于“全链条”服务能力，配备多个垂直领域专业智能体，可灵活调度，覆盖品牌设计、影视创意、电商广告等多种场景。

Anthropic's Claude Code now integrates directly with the iOS simulator, letting developers build, run, and test apps with AI assistance. The tool launches apps in a dedicated panel, monitors the screen, and iterates code automatically—no complex permissions needed. A public beta is open for macOS users with Xcode.

Tencent has launched Miora, a new AI creative agent that can handle complex design tasks from start to finish. Just tell it what you want, and it breaks down the work, coordinates multiple AI specialists, and delivers polished visuals—even if you have zero design experience. New users get 1,000 free points to try it out.

A new study reveals a significant gender gap in generative AI usage, with women 22% less likely to use the technology than men. The gap is linked to risk perception, lack of representation, and time constraints. Experts call for inclusive training and flexible learning to bridge the divide.

Horizon and Volkswagen are deepening their AI partnership through joint venture Coretech. The key? A white-box licensing model that gives Volkswagen access to Horizon's AI foundation models, not just a black-box solution. This tech will integrate with Coretech's C7H chip and GAIA world model platform to accelerate L3/L4 autonomous driving for the Volkswagen Group in China.

At the World Internet Conference in Xi'an, UBTech founder Wang Xingxing predicted that humanoid robots will reach a 'ChatGPT moment' within two to three years. He envisions robots working in unfamiliar environments without pre-programmed actions or remote control, marking a leap from demonstration to practical application.

GitMind, a multimodal AI mind mapping tool, has launched a lifetime subscription for $49.99. It can generate structured charts from PDFs, YouTube videos, audio, and more in seconds. This deal aims to simplify knowledge organization for students and professionals alike.

Samsung is reportedly considering a €1 billion investment in French AI startup Mistral AI, signaling growing confidence in Europe's ability to compete in the global AI race. The move comes as the US tightens regulations, making Mistral's open-weight approach a strategic asset for regions seeking technological sovereignty. Other investors, including Sweden's EQT, are also circling.

Substack has launched a new feature that lets readers check if an article was written by a human or an AI. Using the Pangram detection tool, the platform aims to boost transparency and help audiences navigate the growing wave of AI-generated content. Creators can still opt out or add notes about their writing process.

As AI-powered search becomes the norm, traditional media and online platforms are seeing a sharp drop in traffic. Reddit is reconsidering its $60 million deal with Google, while outlets like Reuters and Politico are blocking crawlers or hiding ads. The battle for content control is heating up.

NVIDIA has launched a synthetic video detection service called SVD that analyzes each frame to identify AI-generated content. It achieves 92% accuracy on uncompressed videos and processes a 1080p clip in just 22 milliseconds on consumer GPUs. The tool is now part of NVIDIA's inference micro-platform, offering a fast and reliable way to combat deepfakes.

Elon Musk is at it again, this time feeding SpaceX's engineering data into his Grok AI model. The new model, boasting a staggering 2 trillion parameters, will be trained on everything from manufacturing processes to Starlink hardware design—minus the rocket engine secrets, of course. It's a classic Musk move: turning his empire's core assets into an AI moat.

OpenRouter 在同一个 API 里集成了语音转录功能，开发者无需再为聊天和转录分别对接不同服务。新端点支持 Whisper 和基于 Token 的 STT 模型，兼容 OpenAI 格式，还能自动负载均衡。这下，给应用加上语音能力，门槛又低了一截。

A federal judge has approved a $1.5 billion settlement in the Anthropic book copyright case, paying authors about $3,000 per book. The ruling distinguishes between lawful fair use training and illegal data acquisition, setting a key precedent for AI companies.

据《The Information》报道，微软正在测试月之暗面（Moonshot AI）的最新模型Kimi K3，评估其接入Microsoft Copilot的可能性，并准备将其部署到Azure平台。这并非微软首次引入月之暗面模型——此前Kimi K2.7Code已登陆GitHub Copilot。若K3正式上线，企业用户将能通过微软云基础设施直接使用该模型，无需再单独对接月之暗面服务。

Samsung is in talks to invest up to €1 billion in Mistral AI, a French startup challenging US tech giants. The deal, which could value Mistral at €20 billion, reflects Europe's push for AI independence and growing demand for computing power. Mistral's open-source models offer a customizable alternative to US-controlled AI.

Meta is testing a new AI app called StoryKit that lets parents create personalized bedtime stories for their kids. Users can snap photos of toys or people to generate characters, choose settings and moral lessons, and even add background music. The app, rolling out in select countries, is for adults only and has no social features. It's Meta's latest push to bring generative AI into everyday family life, sparking conversations about technology's role in parenting and child development.

AliHealth's medical AI platform, Hydrogen Ion, has secured full-text content authorization from NEJM, JAMA, and BMJ, making it the first platform in China to achieve this. This integration allows Chinese doctors to access top-tier medical evidence directly, with AI-powered Q&A and bilingual support, streamlining research and clinical decision-making.

第67届国际数学奥林匹克竞赛（IMO2026）结果揭晓，小红书大模型dots-note-3.0以满分42分夺得金牌，成为中国首个获IMO官方金牌认证的大模型。相比谷歌Gemini去年仅答对5题，这次小红书不仅满分，其第三题的解法更被两位CMO金牌得主盛赞“正确而优雅”。

AI voice cloning is now so advanced that just a few seconds of audio can replicate your voice with alarming accuracy. The Ministry of State Security warns of three major risks: fraud, celebrity impersonation, and fake official statements. With ordinary people failing to detect fake voices over half the time, it's time to rethink trust in what we hear.