Articles

Anthropic has acquired Bun, the high-performance JavaScript runtime, which they previously adopted for their AI coding assistant, Claude Code. This acquisition follows Claude Code's impressive achievement of $1 billion in run-rate revenue within six months of public availability. Bun's founder, Jarred Sumner, explained that joining Anthropic allows Bun to bypass the challenges of monetization as a VC-backed startup and instead focus on developing the best JavaScript tooling, specifically for AI-driven software like Claude Code and future AI coding products. The move solidifies Bun's role as critical infrastructure for Anthropic's growing AI ecosystem and ensures its continued development without immediate revenue pressure.

This article deeply explores the core mechanisms and industry status of Interleaved Thinking in Large Language Models. It traces the evolution of Interleaved Thinking, from Anthropic's Extended Thinking to implementations by models such as OpenAI, MiniMax, Kimi, and DeepSeek. The core principle is that the model retains and accumulates reasoning states during multi-turn tool calling and reasoning, instead of simply discarding them. This significantly improves the performance of complex Agent tasks. The article points out that the lack of native support for Thinking fields in the OpenAI API has led the industry into a lack of standardization. Finally, the article affirms MiniMax's contributions to promoting Interleaved Thinking standardization and ecosystem building, discussing it within the broader context of Agentic AI development.

This article delves into an innovative and practical workflow designed to address the common issue of generic, AI-driven aesthetics in AI-generated UIs and guides readers on how to generate beautiful UIs using Claude Code. The author proposes a process comprising five core steps: First, guide Claude Code to generate an initial HTML structure by providing detailed screenshots and HTML/CSS style information of the target UI. Second, optimize the UI's details and presentation through refined prompts and iterative adjustments to specific element CSS. The crucial third step involves extracting a detailed `STYLE_GUIDE.md` from the optimized UI. This guide covers all design specifications, including colors, fonts, spacing, component styles, and shadows. This style guide serves as the context for subsequent AI-generated UIs, significantly reducing the information gap. Finally, the article demonstrates how to guide AI in constructing pixel-perfect HTML prototypes based on this guide and ultimately convert them into clear, maintainable Next.js applications. This method provides developers with an effective path to improve the quality of AI-assisted UI design.

This article announces the public beta of LangSmith Agent Builder, a no-code platform designed to democratize the creation and deployment of production-ready AI agents. It clearly differentiates agents from traditional, fixed-path workflows, emphasizing agents' superior ability to dynamically determine steps, iteratively call tools until tasks are complete, and continuously improve through user feedback and memory. Agent Builder itself functions as an agent, leveraging best practices from millions of developers to guide users via a chat interface from an initial idea to a fully deployed agent. The beta release introduces significant enhancements, including 'bring your own tools' for integrating external APIs, 'workspace agents' for seamless team collaboration and cloning, multi-model support (OpenAI, Anthropic), programmatic invocation via API for embedding agents into existing systems, and a streamlined UI within LangSmith. The platform aims to resolve common AI adoption challenges by offering an intuitive agent creation experience for all employees, while simultaneously providing technical teams with robust governance over tools. Practical use cases are extensively detailed, spanning role-specific research (e.g., sales, marketing, recruiting), transforming ambient information into tracked projects (e.g., bug tickets, customer support), and developing time-saving communication assistants (e.g., email, calendar, Slack management).

The article details the open-source AI Workflow core engine Flowra, jointly released by the ModelScope community and the WULI team. Flowra aims to address the challenges such as complex model integration, multi-modal data handling, debugging inefficiencies, and scaling difficulties in traditional AI Workflow development. Its core highlights include a comprehensive type system supporting diverse data types including images and videos; a DAG-based execution engine that provides intelligent caching, parallel scheduling, and distributed support; and a complete set of node package development toolchains. Flowra is also deeply integrated with ModelScope, supporting one-line code management of AI Models, and provides a flexible storage backend. The article demonstrates Flowra's powerful features and ease of use through a quick-start guide and a practical YOLO object detection example, aiming to reduce the barrier to entry for AI Workflow development, empowering more developers to rapidly create AI applications.

The article details how ByteDance's DataMind project builds an "AI + Data" unified data engine based on Apache Doris. The DataMind project addresses the limitations of traditional solutions in handling unstructured data and integrating with AI models. It integrates hybrid search capabilities, including Faiss-based vector indexing and Tablet-level BM25 scoring functions. At the same time, it complements AI functions such as AI_QUERY and TEXT_EMBEDDING, and supports Python UDF to meet custom model requirements. The article also delves into the implementation of GraphRAG (Graph Retrieval-Augmented Generation) construction and query mechanisms on DataMind, enhancing complex knowledge retrieval capabilities. Finally, it shares the implementation of an enterprise AI-powered data query platform, effectively solving practical challenges such as massive data volume, security, and query latency by optimizing data agent routing and connecting data lake permission systems, providing a solid foundation for the application of AI in enterprise data consumption.

The article introduces LatentMAS, a multi-agent inference framework proposed by Princeton University and other institutions, aimed at addressing the issues of low efficiency and information loss in traditional text-based multi-agent systems. LatentMAS innovatively allows agents to collaborate directly in the latent space. This is achieved by exchanging hidden layer representations and KV-cache working memory, rather than relying on text communication. Experiments demonstrate that LatentMAS achieves the highest accuracy improvement of 14.6% compared to single models and traditional text-based MAS on nine benchmark tasks, including mathematics, scientific reasoning, and code generation, while reducing output token usage by 70.8–83.7% and accelerating end-to-end inference speed by 4–4.3 times. This framework requires no additional training, is versatile, compatible with any HuggingFace model, and reveals that latent thinking has richer and more expressive semantic capabilities than discrete tokens.

The article summarizes Alibaba Group Chairman Joe Tsai's speech at the University of Hong Kong. He reviewed Alibaba's development from B2B to C2C to cloud computing, driven by customer demand, emphasizing the criticality of technological self-reliance for both the nation and enterprises, partly in response to current geopolitical dynamics. Tsai highlighted China's three major competitive advantages in AI: low energy costs due to large-scale energy investments and clean energy development, overall cost advantages in data center construction and model development, and a large pool of engineering talent. He emphasized the importance of the open-source strategy for accelerating AI adoption and promoting social benefit, and explained Alibaba's business logic of monetizing through cloud services rather than directly profiting from AI itself. Finally, Tsai provided young people with career advice, recommending hard and soft skills such as programming, data analysis, psychology, and materials science, and urging them to be prepared to seize opportunities with asymmetric risk-reward profiles.

This article tests and evaluates the various functions of Zhipu AI's newly launched GLM-4.6V multimodal large model. Due to Google AI Studio tightening access for free users to Gemini models, the author sought alternatives and was deeply impressed by GLM-4.6V's performance. Through detailed case studies, the article showcases GLM-4.6V's capabilities in replicating web screenshots into code, extracting structured information from mixed data (such as HTML tables and JSON seal information), comparatively analyzing academic papers, retrieving information from a 114-page document, and summarizing and analyzing video content. The author believes that GLM-4.6V performs excellently in practical applications, boasting native multimodal advantages and strong Grounding capabilities, and emphasizes the importance of domestic AI models in solving users' usage limit concerns.

The article details the latest EyeReal glasses-free 3D display technology from a Chinese team appearing in Nature. By deeply integrating computational optics and artificial intelligence, this technology successfully overcomes the physical limitations of traditional glasses-free 3D displays in terms of screen size and viewing angle. EyeReal achieves a desktop monitor-sized 3D image without glasses, covering a wide viewing angle of over 100 degrees, and provides a continuous and smooth viewing experience. Its core innovation lies in the dynamic space-bandwidth product (SBP) utilization strategy, combined with real-time eye tracking, which precisely projects limited optical information to the observer's eyes. Technically, it transforms complex viewing angle changes into standardized planar images through physically accurate binocular geometric modeling and eye geometry encoding, and then uses a lightweight fully convolutional neural network (FCN) for reverse rendering to control the pixel phase of three common TFT-LCD panels, using Malus's Law to synthesize a precise light field. EyeReal achieves 'full parallax' and 'focus parallax', effectively solving the '3D sickness' problem that plagues users, and provides a low-cost, high-quality new solution for immersive experiences. The research was led by Ma Weijie, a doctoral student at Fudan University, and was jointly trained by institutions such as the Shanghai AI Laboratory.