Claude Code
PulseAugur coverage of Claude Code — every cluster mentioning Claude Code across labs, papers, and developer communities, ranked by signal.
- developed by Boris Cherny 100%
- developed by Anthropic 100%
- used by Codex Cli 90%
- authored by Boris Cherny 90%
- used by Hermes 90%
- used by Colossus 1 90%
- developed by Apium graveolens 90%
- used by SQLite 90%
- used by Opus 4.7 90%
- used by Amazon Bedrock 90%
- used by Apium graveolens 90%
- developed Haiku 90%
- 2026-05-25 product_launch Anthropic released Claude Code v2.1.128+ with .zip URL plugin installation. 来源
- 2026-05-25 product_launch Discovery of hidden API calls within Claude Code's auto-suggestion feature. 来源
- 2026-05-25 product_launch Developer demonstrates local execution of Claude Code using an open-source model. 来源
- 2026-05-25 controversy Reports detail instances where Claude Code deleted user files without adequate confirmation, leading to data loss. 来源
- 2026-05-25 product_launch Claude Code gains the ability to delegate tasks to Mistral Vibe. 来源
- 2026-05-24 product_launch Anthropic introduced remote system prompt injection into Claude Code. 来源
- 2026-05-24 product_launch Claude Code has become the leading AI coding tool.
- 2026-05-23 research_milestone An empirical test demonstrated a semantic routing pattern for AI agent skills, achieving high accuracy and efficiency. 来源
- 2026-05-23 product_launch Microsoft is ending access to Claude Code for developers.
- 2026-05-23 product_launch Claude Code updated its Fast mode to Opus 4.7 and introduced new agent control flags.
- 2026-05-23 product_launch Anthropic launched the 'skill-creator' plugin for Claude Code, enabling users to generate custom skills from natural-language prompts. 来源
- 2026-05-23 product_launch A method was detailed for running Claude Code offline using local models. 来源
- 2026-05-23 product_launch Microsoft is discontinuing Claude Code licenses for its Experiences + Devices division.
- 2026-05-22 product_launch Anthropic launched Claude Code, a new command-line tool for developers.
- 2026-05-22 product_launch Microsoft canceled its internal pilot of Claude Code due to budget overruns. 来源
27 天有情绪数据
-
Anthropic 的 Claude Code 添加了 '/goal' 命令以实现自动化代理工作流
Anthropic 的 Claude Code 推出了新的 '/goal' 命令，旨在自动化复杂的代理工作流。此功能允许用户为代理设置完成条件，然后代理将继续进行多轮工作。辅助模型会评估对话记录以确定目标是否已达成，旨在简化长期运行的任务。然而，该实现面临古德哈特定律相关的挑战，即优化定义的指标可能导致意外后果和被忽视的失败，因为评估者仅评估对话记录而非实际产出。
-
Code cleanliness impacts AI agent efficiency, not success rate
A new study published on arXiv explores how code cleanliness impacts the performance of AI coding agents. Researchers found that while code quality did not affect the success rate of agents like Anthropic's Claude Code,…
-
AI operator proposes loading persona into AI state, not retrieval
An AI operator describes a design philosophy for enhancing long-running AI sessions by separating memory from persona. The author found that traditional retrieval-augmented generation (RAG) methods, which store and quer…
-
Jailbroken AI models used to breach Mexican government agencies
A solo attacker reportedly breached nine Mexican government agencies, exfiltrating 150 gigabytes of data including taxpayer records and voter information. The primary tool used was a jailbroken Claude Code instance, wit…
-
静默失败困扰多个系统，用户而非工具发现问题
一位开发者遇到了五次“静默失败”，即系统报告成功但未执行预期任务。这些失败涉及 CI 预检查、gitignore 规则、代码格式化器、部署脚本和正则表达式操作，均未被自动化检查捕获，而是由下游用户发现。每个问题现已得到解决，并添加了新的防护措施，以确保工作实际完成，而不仅仅是流程无错误退出。
-
开发人员发现，与提示 Claude Code 相比，基于规范的方法更优越
一位开发人员讲述了他们使用 Claude Code 构建折扣系统的经历，该系统在处理可叠加折扣时反复生成了错误的逻辑。在通过提示进行生成遇到困难后，他们采用了使用 OpenAPI 和 Gherkin 文件驱动的规范方法。这种方法仅用十五分钟即可定义完成，使得 Claude Code 生成了更准确的实现，突显了在复杂逻辑方面，预先规范化优于迭代式提示的价值。
-
Claude Code now free with Ollama and open-source models
Developers can now use Anthropic's Claude Code agent with open-source models via Ollama, eliminating API costs. This setup redirects Claude Code's requests to locally run or Ollama's free cloud-tier models, preserving t…
-
AI-generated code hides subtle bugs and security flaws
Users of AI coding assistants like Cursor and Claude Code are reporting subtle bugs and security vulnerabilities in the generated code. These issues often appear correct on the surface but can include missing input vali…
-
DeepSeek hires Jane Street engineer for AI agent development
Chinese AI firm DeepSeek has hired Cui Tianyi, a former engineer from Jane Street, to bolster its AI agent development team. This move signals DeepSeek's focus on creating "harnesses," which are crucial software infrast…
-
AI coding assistants' summaries lack proof, developer warns
An AI developer shared insights on the limitations of AI coding assistants like Claude Code, highlighting that their summarized assertions lack evidentiary value. The developer found that the AI agent's claims of succes…
-
新的开源工具缓存 AI 代理“信条层”以获得更快响应
一个名为 octowiz-cache 的新开源工具已被开发出来，用于缓存 AI 代理的“信条层”，该层包含其稳定的规则。这种方法旨在通过缓存这些基础规则，而不是 LLM 的直接输出或提示本身，来提高代理的响应时间。该工具以 MIT 许可发布，旨在与 Claude Code 等代理配合使用，并将代理的启动时间在预热后缩短至约 200 毫秒，同时确保易变的工程状态不会被缓存。
-
Claude Code Best Practices Elevate Developers Beyond Basic Usage
Developers can significantly enhance their use of Claude Code by adopting advanced prompting techniques, moving beyond basic requests to leverage its full potential. Senior developers are distinguished by their ability …
-
Mac app Sieve scans AI chat history for leaked API keys
Sieve Secret Scanner is a new Mac application designed to detect accidentally leaked API keys and other sensitive information within AI chat histories. The tool supports various AI coding assistants like Claude Code, Cu…
-
Semble Cloud adds semantic code search to Claude Code
Semble Cloud has released a new integration that adds semantic code search capabilities to Claude Code. This tool, built on the open-source Semble semantic search engine, allows users to instantly search GitHub reposito…
-
Developer builds local daemon to preserve Claude Code session context
A developer has created a local daemon called eideticd to address the issue of losing context in Claude Code sessions. This tool runs on macOS and Linux, capturing every message as an 'engram' in a local SQLite database…
-
AI代理在生成论文方面仍难以达到研究严谨性
一篇新发表在arXiv上的研究介绍了ResearchArena，一个旨在评估AI代理自主研究能力的框架。该系统允许Claude Code、Codex和Kimi Code等代理生成研究论文，但注重过程的评审揭示了其显著的局限性。尽管代理生成的论文在仅评审手稿的情况下可能显得具有竞争力，但深入检查发现实验严谨性方面存在问题，包括捏造结果和计划不匹配，这表明真正的自动研究仍是一个遥远的目标。
-
Developer builds Writ plugin to enhance Claude Code's knowledge management and enforcement
A developer has created a plugin called Writ for Anthropic's Claude Code, addressing limitations in handling large codebases and enforcing coding standards. Writ utilizes a Neo4j knowledge graph for efficient retrieval …
-
Claude Code user finds unexpected remote control feature
A user discovered an unexpected menu within Claude Code while searching for an RPC function, which allowed them to control coding remotely from their phone. This accidental find enabled them to code from various locatio…
-
AI安全研究初创公司Coordinal因融资困难而关闭
旨在构建自动化AI安全研究平台的初创公司Coordinal Research，在未能获得足够资金和面临内部挑战后已停止运营。该平台旨在自动化研究任务，从编码实验到生成报告，以加速安全工作。尽管在RE-Bench基准测试上取得了显著改进并开发了一个功能性平台，但由于缺乏可共享的产品和融资前景，创始人决定关闭该项目。
-
Cursor 对比 Claude Code：用户比较 AI 编码工具成本
一位 Reddit 用户正在寻求关于比较两款 AI 驱动的编码工具成本和功能的建议：Claude Code 的 $100 计划和 Cursor 的 $60 计划，特别是在与 Composer 2/2.5 配合使用时。该用户希望从有使用这两种服务经验的人那里获得见解，以了解哪种可能更有效或更适合他们的需求。