Discover quality AI tools
Bring together the world’s leading artificial intelligence tools to help improve work efficiency
Articles published by JoseJackson
QwenWork launched QwenNote Eva, its first screenless desktop robot, at the 2026 CloudIn Conference. Priced at 899 yuan, it integrates QwenWork with DingTalk to offer intelligent collaboration. Key features include natural language command processing, proactive task execution, and deep DingTalk ecosystem integration for meetings and messaging. The device supports Android, iOS, and HarmonyOS. It features gaze and voice activation, over 100 emotional expressions, and long-term memory for personalized interaction. Hardware includes a 3-axis gimbal, RGB lighting, and 65W fast charging via dual Type-C and magnetic wireless. It aims to provide an AI assistant experience combining productivity with emotional companionship.
QwenWork launched QwenNote Eva, its first screenless desktop robot, at the 2026 CloudIn Conference. Priced at 899 yuan, it integrates QwenWork with DingTalk to offer intelligent collaboration. Key features include natural language command processing, proactive task execution, and deep DingTalk ecosystem integration for meetings and messaging. The device supports Android, iOS, and HarmonyOS. It features gaze and voice activation, over 100 emotional expressions, and long-term memory for personalized interaction. Hardware includes a 3-axis gimbal, RGB lighting, and 65W fast charging via dual Type-C and magnetic wireless. It aims to provide an AI assistant experience combining productivity with emotional companionship.
Ollmo, a production-grade RAG and Agent visualization platform for private deployment and team collaboration, has released version 0.1.0. It allows users to obtain precise answer sources by uploading documents and asking questions, and set up a private knowledge base Q&A system out of the box. The platform features a front-end and back-end separation design using Go and Fiber for the backend and Next.js App Router with shadcn-ui for the front-end, supporting one-click Docker deployment. It offers two service models (private deployment and SaaS) and integrates a hybrid search engine combining Milvus dense vectors and MySQL full-text search with RRF score fusion and semantic weight adjustment. Full-chain logical isolation of MySQL, Milvus, and MinIO ensures multi-tenant safety. The platform includes tools such as asynchronous batch document upload, real-time answer generation with citations, manual annotation to reduce hallucinations, workflow orchestration via React-Flow, GraphRAG for entity extraction, plus usage statistics dashboards, team permission management, open APIs, and bilingual support to meet enterprise team needs.
Ollmo, a production-grade RAG and Agent visualization platform for private deployment and team collaboration, has released version 0.1.0. It allows users to obtain precise answer sources by uploading documents and asking questions, and set up a private knowledge base Q&A system out of the box. The platform features a front-end and back-end separation design using Go and Fiber for the backend and Next.js App Router with shadcn-ui for the front-end, supporting one-click Docker deployment. It offers two service models (private deployment and SaaS) and integrates a hybrid search engine combining Milvus dense vectors and MySQL full-text search with RRF score fusion and semantic weight adjustment. Full-chain logical isolation of MySQL, Milvus, and MinIO ensures multi-tenant safety. The platform includes tools such as asynchronous batch document upload, real-time answer generation with citations, manual annotation to reduce hallucinations, workflow orchestration via React-Flow, GraphRAG for entity extraction, plus usage statistics dashboards, team permission management, open APIs, and bilingual support to meet enterprise team needs.
Douyin upgraded its minor mode by integrating multimodal large language models (MLLM) for age-appropriate content recommendations. The system filters content via MLLM, human expert review, and age classification, then matches minors' preferences. Users activate the mode by entering birth date. The algorithm adjusts to cognitive stages and allows real-time feedback.
Douyin upgraded its minor mode by integrating multimodal large language models (MLLM) for age-appropriate content recommendations. The system filters content via MLLM, human expert review, and age classification, then matches minors' preferences. Users activate the mode by entering birth date. The algorithm adjusts to cognitive stages and allows real-time feedback.
AI detection tools mistakenly flag classic Chinese literature like Zhu Ziqing's Moonlight on the Lotus Pond as AI-generated with high probability due to their extensive use in training AI models. Experts clarify that the percentage reflects the likelihood the entire text is AI-generated, not the portion written by AI, and warn these tools are unreliable for judging originality, as results vary widely by platform and text length. They advise maintaining a unique writing style to counter such algorithmic errors.
AI detection tools mistakenly flag classic Chinese literature like Zhu Ziqing's Moonlight on the Lotus Pond as AI-generated with high probability due to their extensive use in training AI models. Experts clarify that the percentage reflects the likelihood the entire text is AI-generated, not the portion written by AI, and warn these tools are unreliable for judging originality, as results vary widely by platform and text length. They advise maintaining a unique writing style to counter such algorithmic errors.
Chinese local services giant Meituan outlines its three core AI development directions for 2026. The strategy includes integrating AI into existing operations like delivery routing and merchant tools, developing native AI products such as the Miyou community, and building an open AI ecosystem with APIs and a developer skill market. These efforts aim to digitally transform and reshape local life services.
Chinese local services giant Meituan outlines its three core AI development directions for 2026. The strategy includes integrating AI into existing operations like delivery routing and merchant tools, developing native AI products such as the Miyou community, and building an open AI ecosystem with APIs and a developer skill market. These efforts aim to digitally transform and reshape local life services.
Researchers introduced HarmonyGNN, a new training technique for graph neural networks (GNNs) that improves accuracy on heterogeneous graph data without relying on labeled nodes. Tests on 11 benchmark graphs showed state-of-the-art performance on seven homogeneous graphs and accuracy gains of 1.27% to 9.6% on four heterogeneous graphs. The framework also boosts computational efficiency. The work will be presented at the International Conference on Learning Representations in 2026.
Researchers introduced HarmonyGNN, a new training technique for graph neural networks (GNNs) that improves accuracy on heterogeneous graph data without relying on labeled nodes. Tests on 11 benchmark graphs showed state-of-the-art performance on seven homogeneous graphs and accuracy gains of 1.27% to 9.6% on four heterogeneous graphs. The framework also boosts computational efficiency. The work will be presented at the International Conference on Learning Representations in 2026.
MiniMax launches Expert 2.0 AI agent platform and cloud assistant MaxClaw enabling no-code professional agent creation and seamless integration with office tools DeepSeek V4 trillion-parameter multimodal AI model with 1M context window begins internal testing prioritizing compatibility with domestic chips KFC China introduces AI ordering assistant Xiao K powered by Alibabas Tongyi Qianwen handling complex orders via natural language Shenzhen Huaqiangbei AI glasses sales surge 80 on Amazon challenging Meta with low-cost alternatives Perplexity releases Computer a general AI agent integrating multiple models to execute complex workflows autonomously Samsung Galaxy S26 series debuts with over ten upgraded AI features including enhanced input and image processing Sanqi Interactive implements its small seven AI model across 85 of products boosting game exports and cultural preservation
MiniMax launches Expert 2.0 AI agent platform and cloud assistant MaxClaw enabling no-code professional agent creation and seamless integration with office tools DeepSeek V4 trillion-parameter multimodal AI model with 1M context window begins internal testing prioritizing compatibility with domestic chips KFC China introduces AI ordering assistant Xiao K powered by Alibabas Tongyi Qianwen handling complex orders via natural language Shenzhen Huaqiangbei AI glasses sales surge 80 on Amazon challenging Meta with low-cost alternatives Perplexity releases Computer a general AI agent integrating multiple models to execute complex workflows autonomously Samsung Galaxy S26 series debuts with over ten upgraded AI features including enhanced input and image processing Sanqi Interactive implements its small seven AI model across 85 of products boosting game exports and cultural preservation
Amazon plans a content licensing marketplace connecting publishers and AI firms to legally sell data for AI training This move addresses industry copyright disputes and follows similar initiatives like Microsofts PCM aiming to leverage AWS infrastructure for publisher revenue
Amazon plans a content licensing marketplace connecting publishers and AI firms to legally sell data for AI training This move addresses industry copyright disputes and follows similar initiatives like Microsofts PCM aiming to leverage AWS infrastructure for publisher revenue
OpenAI launches GPT-5.2, its most advanced frontier model designed for developers and enterprise use. It comes in three variants: Instant for speed, Thinking for complex tasks like coding, and Pro for maximum accuracy. The release intensifies competition with Google's Gemini 3 as OpenAI aims to reclaim market leadership and boost its enterprise offerings.
OpenAI launches GPT-5.2, its most advanced frontier model designed for developers and enterprise use. It comes in three variants: Instant for speed, Thinking for complex tasks like coding, and Pro for maximum accuracy. The release intensifies competition with Google's Gemini 3 as OpenAI aims to reclaim market leadership and boost its enterprise offerings.
DS-STAR a new AI agent automates complex data science tasks by analyzing diverse file formats and using iterative planning with LLM verification It achieves top performance on benchmarks including DABStep significantly outperforming existing agents
DS-STAR a new AI agent automates complex data science tasks by analyzing diverse file formats and using iterative planning with LLM verification It achieves top performance on benchmarks including DABStep significantly outperforming existing agents
