由《Mastering Emacs》一书的作者开发并正式发布的 AI 编码工具“DreamCoder”,旨在解决现有终端式人工智能代理在界面交互上的痛点 1。该项目源于作者在2025年5月安装 Claude Code 后,因不满其终端交互方式及 GUI 的局限性,决定自行构建更好的编码框架 1。经过15个月的开发和测试,该工具最终发布正式版 1。
DreamCoder 的核心设计理念借鉴了 Emacs 的“建议系统”(Advice System)和可塑性(Ductility),强调工具应适应人类而非相反 1。在功能实现上,该软件采用了类似 Emacs 的平铺窗口管理设计,支持键盘操作、多工作区并行管理及本地开源模型运行 1。具体特性包括平铺窗口管理器、工作区(Workspaces)、小地图监控状态、Avy风格的键盘选择器、MCP服务器支持以及后台子代理等 1。此外,DreamCoder 支持本地运行 Qwen 3.8 27B 等开源模型,作者认为本地模型能力正在迅速提升且成本更低 1。展望未来,开发者计划在该平台中集成世界级的渗透测试和网络安全套件 1。
Michael, the author of Mastering Emacs, has officially released DreamCoder, an AI coding harness designed to address limitations found in existing terminal-based AI agents 1. The development journey began in May 2025, when Michael installed Claude Code but grew dissatisfied with its interactive approach and graphical user interface constraints 1. Following 15 months of rigorous testing and development, the final version was launched to provide a more efficient human-AI collaboration experience compared to traditional IDEs or simple chat interfaces 1.
DreamCoder adopts design principles inspired by Emacs, particularly its Advice System and ductility, emphasizing that tools should adapt to humans rather than the reverse 1. Key features include a tiling window manager, workspaces for parallel task management, minimap status monitoring, Avy-style keyboard selectors, MCP server support, and background sub-agents 1. The software supports running open-source models locally, such as Qwen 3.8 27B, reflecting Michael's view that local model capabilities are rapidly improving while offering lower costs 1. Future plans for the project include the integration of world-class penetration testing and cybersecurity suites 1.
评论
还没有评论,欢迎留下第一条。