NO/FOMO

Independent AI signal, once a day

The AI briefing worth opening.

ISSUE DATE2025-11-30ENGLISH EDITION
This issue
—
All time
—

Hacker News

6 stories
01

Program-of-Thought Prompting Outperforms Chain-of-Thought by 15%

A novel prompting technique, Program-of-Thought (PoT) prompting, has demonstrated a significant performance improvement of 15% over the widely adopted Chain-of-Thought (CoT) prompting method. This research introduces PoT as an innovative approach where large language models are guided to express their intermediate reasoning steps in a structured programming language, rather than conventional natural language. By leveraging the formal syntax, logical constructs, and inherent compositionality of programming paradigms, PoT facilitates more systematic, modular, and verifiable reasoning processes. This structured format offers several advantages, including explicit function definitions, parameter passing, and conditional logic, which enhance the model's capability to decompose and solve complex, multi-step problems more effectively. The improved performance observed across various tasks suggests that instructing LLMs with programmatic steps can unlock superior symbolic reasoning abilities, enable robust error correction, and lead to more accurate and explainable outputs, representing a substantial advancement in the field of prompt engineering and cognitive architectures for AI models.

Program-of-ThoughtChain-of-ThoughtPrompt EngineeringLarge Language ModelsSymbolic ReasoningArtificial IntelligenceLarge Language ModelNatural Language Processing
02

Writing a Good Claude.md

This article, titled "Writing a Good Claude.md," delves into best practices for crafting effective and clear interactions with AI models, particularly focusing on Anthropic's Claude. It emphasizes the importance of structured prompting, suggesting the use of Markdown (.md) to organize instructions, examples, and contextual information within prompts. Key recommendations include defining the AI's persona and task clearly, breaking down complex requests into smaller, manageable steps, and providing relevant background data to guide the model's responses. The discussion likely covers techniques for precise language, avoiding ambiguity, and iterating on prompts to refine output quality. By adopting a methodical approach to prompt engineering, users can significantly enhance Claude's performance, ensuring more accurate, relevant, and consistently high-quality generated content, whether for code, creative writing, or analytical tasks. The core idea revolves around transforming natural language instructions into a robust, machine-interpretable format to unlock the full potential of large language models.

Claude AIPrompt EngineeringLarge Language ModelAI InteractionMarkdownLarge Language ModelNatural Language ProcessingArtificial Intelligence
03

The Thinking Game Film – Google DeepMind Documentary

The Thinking Game Film is a documentary anticipated to delve into the groundbreaking work of Google DeepMind, a leading artificial intelligence research organization. This film will likely provide an in-depth look at DeepMind's significant contributions to the field of AI, exploring its journey from initial concepts to major breakthroughs. Viewers can expect an examination of the scientific challenges, technological advancements, and the intellectual prowess involved in developing advanced machine intelligence. The documentary is projected to cover key achievements, such as DeepMind's innovations in reinforcement learning, its success in mastering complex strategic games, and its overarching ambition to create general AI capable of addressing real-world problems. It will also likely touch upon the ethical considerations and societal implications of cutting-edge AI, offering a comprehensive perspective on how DeepMind is shaping the future of artificial intelligence through its dedicated research and pioneering projects.

Artificial IntelligenceDeepMindMachine LearningAI ResearchDeep LearningArtificial IntelligenceMachine LearningDeep Learning
04

Norway wealth fund to vote for human rights report at Microsoft, against Nadella

The Norway wealth fund, one of the world's largest sovereign wealth funds and a significant institutional investor in Microsoft, has publicly announced its intention to vote in favor of a human rights report at Microsoft's upcoming Annual General Meeting. This decision simultaneously signals a vote against the current management, including CEO Satya Nadella. While the core of this matter addresses corporate governance and the critical issue of human rights accountability, it carries substantial implications for a leading global technology company such as Microsoft. In an era where tech giants face increasing scrutiny over ethical AI development, data privacy practices, and their global operational footprint, investor pressure on non-financial performance indicators like ESG (Environmental, Social, and Governance) is intensifying. This move by a major shareholder could influence Microsoft's strategic integration of human rights considerations into its product development cycles, supply chain management, and worldwide operations, potentially affecting its reputation and long-term sustainability within the highly competitive technology landscape. The emphasis on human rights reports reflects a growing demand for greater transparency and ethical oversight in the design and deployment of technology globally, indicating a broader trend among investors advocating for responsible technology practices across the industry.

Responsible AIData PrivacyDigital EthicsTechnology GovernanceAI PolicyArtificial IntelligenceMachine LearningOthers
05

The space of minds

Given the title 'The space of minds' from a prominent AI researcher's blog, the article likely delves into the conceptual and architectural diversity within artificial intelligence, framing different AI models or paradigms as distinct 'minds' occupying a vast computational landscape. It is anticipated to explore how various AI systems, from large language models to reinforcement learning agents, represent different approaches to intelligence and cognition, each with unique capabilities and limitations. The discussion may touch upon the evolution of AI, the emergent properties observed in complex models, and the challenges of defining and categorizing machine intelligence. The author likely encourages readers to consider the philosophical implications of this 'mind space'—how new architectures expand its boundaries, the potential for novel forms of intelligence, and the broader societal impact of creating and understanding a multitude of artificial cognitive entities. The piece aims to provide a high-level perspective on the ongoing exploration and development of intelligent systems, emphasizing the breadth and depth of current and future AI research.

Artificial IntelligenceMachine LearningAI ModelsCognitive ArchitecturesEmergent PropertiesArtificial IntelligenceMachine LearningLarge Language Model
06

Garry Tan says MCP "barely works" today

Prominent tech investor and former Y Combinator president, Garry Tan, has reportedly shared a critical assessment regarding the current operational status of a system referred to as "MCP," stating that it "barely works" today. While the specific context and technical details of "MCP" were not provided in the initial brief, Tan's remarks from a figure of his influence typically indicate significant underlying performance or functional challenges within a key technological framework or platform. His observation suggests potential issues impacting the system's reliability, efficiency, or broader applicability within its intended domain. In the absence of further clarification, it is speculated that "MCP" could represent a complex computational platform, a critical software protocol, or an essential component within a larger technological ecosystem, possibly with implications for areas such as machine learning infrastructure or distributed computing. This public declaration could prompt developers and stakeholders to scrutinize the platform's robustness, its architectural design, and its capacity to meet contemporary demands for scalability, data processing, or integrated functionality, potentially initiating efforts for strategic re-evaluation and targeted improvements to address the highlighted deficiencies.

MCPSystem PerformancePlatform ReliabilityTechnology AssessmentOperational IssuesArtificial IntelligenceMachine LearningOthers

GitHub

3 stories
01

TrendRadar

TrendRadar is a lightweight, easy-to-deploy hot topic assistant designed to deliver relevant news and information, eliminating screen clutter. It aggregates hotspots from over 11 mainstream platforms including Zhihu, Douyin, Weibo, and Wallstreetcn, allowing users to customize additional sources. The project features intelligent push strategies (daily summary, current ranking, incremental monitoring) and precise content filtering based on personal keywords with advanced syntax for inclusion, exclusion, and quantity limits. It also offers real-time hot topic trend analysis, personalized algorithms, and multi-channel notification support (WeChat, Feishu, DingTalk, Telegram, Email, ntfy, Bark, Slack). A significant addition is the AI smart analysis powered by the Model Context Protocol (MCP), providing 13 analytical tools for natural language querying, trend tracking, data insights, and sentiment analysis on accumulated news data. TrendRadar supports multi-terminal adaptation via GitHub Pages and Docker, promoting active information retrieval over algorithm-driven consumption for investors, self-media, and general users.

News AggregationHot Topic MonitoringTrend AnalysisContent FilteringAI AnalysisArtificial IntelligenceNatural Language ProcessingAI Agent
02

Agent Development Kit (ADK) for Go

The Agent Development Kit (ADK) for Go is an open-source, code-first toolkit developed by Google for building, evaluating, and deploying sophisticated AI agents. This flexible and modular framework applies software development principles to AI agent creation, simplifying the orchestration of workflows from simple tasks to complex systems. ADK is notably model-agnostic and deployment-agnostic, ensuring compatibility beyond its optimization for Google's Gemini models.Tailored for cloud-native applications, the Go version of ADK leverages Go's strengths in concurrency and performance. Its key features include an idiomatic Go design, a rich tool ecosystem for integrating custom or pre-built functions, and a code-first development approach enhancing flexibility, testability, and versioning. ADK facilitates the design of modular multi-agent systems for scalable applications and supports easy containerization and deployment, particularly in cloud-native environments like Google Cloud Run. This makes it an ideal choice for developers aiming for robust, high-performance agent solutions.

AI Agent DevelopmentGo Programming LanguageCloud-native ApplicationsMulti-Agent SystemsArtificial Intelligence ToolkitArtificial IntelligenceAI AgentMachine Learning
03

➤ Cursor Free VIP

Cursor Free VIP is a cross-platform utility designed for educational and research purposes, supporting the AI-powered coding assistant Cursor, specifically its 0.49.x versions. This tool provides functionalities such as resetting Cursor's configuration and offering multi-language support including English, Simplified Chinese, Traditional Chinese, and Vietnamese. It is compatible with major operating systems: Windows (x64, x86), macOS (Intel, Apple Silicon), and Linux (x64, x86, ARM64). The project clarifies that it adheres to legal standards, refrains from generating fake email accounts or OAuth access, and encourages users to support the original Cursor project. For ease of deployment, automated installation scripts are provided for Linux/macOS and Windows. Furthermore, the tool features extensive configuration options via a `config.ini` file, allowing customization of paths for browsers and Cursor's internal storage, as well as fine-tuning various timing parameters for operations like human verification and email checks. It also addresses common issues, such as permission problems and account authorization errors, advising the use of non-temporary email services for stable operation. The project actively welcomes community contributions.

Cursor AIDeveloper ToolCross-platformConfiguration UtilityAutomation ScriptArtificial IntelligenceAI AgentOthers