From 7ba607f78833c320877c08aeb22619ec9c1cf803 Mon Sep 17 00:00:00 2001 From: HanZijie Date: Mon, 19 Oct 2026 20:26:27 +0800 Subject: [PATCH] README: reframe the intro around Agent OS and MiniApps Replace the opening of the product vision with the Agent OS framing (plugins as software, MiniApps as professional workspaces built on AI as an atomic capability), in both English and Chinese. Drop the paragraph about MiniApp content entering the input box before sending. --- README.md | 6 ++---- README.zh-CN.md | 6 ++---- 2 files changed, 4 insertions(+), 8 deletions(-) diff --git a/README.md b/README.md index 2f48cc7b..7bdf6c4a 100644 --- a/README.md +++ b/README.md @@ -14,9 +14,9 @@ Packages are organized by author under `plugins///`. ## Pursuing Human in the Loop, Toward Agent OS -**Delivery between AI and knowledge workers should not be one-shot.** Code delivery between a coding agent and a programmer is not completed in one pass. Yet many AI products try to deliver a lawyer's, analyst's, or operator's work as a single final answer. Serious work moves from draft to final through human–AI collaboration and repeated refinement; that loop is still too fragmented. Human–AI collaboration is still at a very early stage. +We believe the Agent is evolving into the next-generation operating system (OS). In this paradigm shift, plugins are no longer mere feature extensions: plugins are software, and plugins are apps (MiniApps). -Programmers write in an IDE; lawyers, analysts, and operators write in their own professional environments. Real work, like software development, advances continuously and gets refined until it is stable and ready to deliver. An AI that writes once and walks away is irresponsible in serious work, because **the goal is not text generation, but delivery.** +At the same time, real productivity in every industry remains rooted in highly specialized work environments. A conversational interface (Chat UI) alone cannot replace the precise control and rich views of professional tools. Professional workflows of the future should not ask users to leave their workbench to accommodate AI. Instead, professional environments and AI should evolve in deep symbiosis, with AI internalized as a foundational, atomic capability available on demand—like power and water, compute, and storage. By building professional-grade MiniApps that grow on top of the Agent OS, we aim to connect the specialized interfaces of vertical domains with the AI-native compute beneath them, and to reshape how every industry works. > *Self-drive Route Planner case: an Agent-assisted weekend trip around Shanghai, with the final itinerary and synchronized route workspace visible together.* This case shows the kind of professional workspace a MiniApp can provide: the Agent conversation and complete itinerary are on the left, while the route summary, interactive map, and editable trip workspace are on the right. @@ -42,8 +42,6 @@ In the way of working we envision, a lawyer would not submit an AI-written plead Instead, they would ask AI to research background, organize material, prepare a first draft, improve the expression, and find omissions—then return to their professional workspace to **review, edit, confirm, and let AI continue.** -MiniApps enforce this boundary in the product: when content from a MiniApp page is handed to the Agent, it first enters the user's input box. The user can edit and confirm it before sending. AI never presses the Send button on the user's behalf. - ## Architecture MiniApps run inside the MiniMax Code desktop host through four cooperating layers: diff --git a/README.zh-CN.md b/README.zh-CN.md index 54f9077e..dd0c744f 100644 --- a/README.zh-CN.md +++ b/README.zh-CN.md @@ -14,9 +14,9 @@ ## 追求 Human in the Loop，迈向 Agent OS -**AI 与办公用户之间的交付，不应该是一次性的。** Coding Agent 和程序员之间的代码交付并非一次完成；但如今市面上的 AI 产品，却往往试图一次性交付律师、分析师、运营人员的工作成果。从草稿到定稿，本应是人机协作、反复打磨的过程，这一过程却被割裂了。人与 AI 的协作，仍停留在非常初级的阶段。 +我们相信，Agent 正在演进为下一代操作系统（OS）。在这一范式转移下，插件不再只是简单的功能扩展，插件即软件，插件即 App（MiniApp）。 -程序员在 IDE 里码字；律师、分析师、运营人员，也在各自的领域里码字。真正的工作和代码开发一样，是一个持续推进、不断打磨的过程，最终形成稳定、可交付的成果。一个码完字就走的 AI，在任何严肃工作中都是不负责任的，因为**工作要的不是码字，而是交付。** +与此同时，各行各业的真实生产力依然扎根于高度专业化的工作环境中。单纯的“对话式交互（Chat UI）”无法替代专业工具的精准控制与复杂视图。未来的专业工作流，不应是让用户离开工作台去迁就 AI，而是让专业工作环境与 AI 深度共生——将 AI 内化为像水电、计算、存储一样可随时调用的底层原子能力。通过构建生长在 Agent OS 之上的专业级 MiniApps，我们致力于连接垂直领域的专业交互界面与底层的 AI 原生算力，重塑各行各业的工作方式。 > *自驾规划案例：在 Agent 协助下规划上海周边周末自驾，并同时查看最终行程和同步后的路线工作台。* 这个案例展示了 MiniApp 可以提供的专业工作界面：左侧是 Agent 对话与完整行程，右侧是路线摘要、交互式地图和可编辑的行程工作台。 @@ -43,8 +43,6 @@ 但他们会让 AI 帮自己研究背景、整理材料、完成初稿、完善表达、发现遗漏，然后回到自己的专业工作台中，**审阅、修改、确认，再让 AI 继续推进。** -在产品里，这条边界是这样守的：MiniApp 页面上的内容要交给 Agent 时，会先进入你的输入框，由你修改、确认后再发送。AI 不会替你按下发送键。 - ## 架构 MiniApp 运行在 MiniMax Code 桌面宿主中，由四个相互配合的层组成：