Emphasize LLM-supervised positioning in README

Lead with the technical differentiator (comparison table + architecture
diagram) before the value proposition. Add feature summary to tagline.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: Grivn

README.md

@@ -4,7 +4,7 @@
 
 # Mnemon
 
-**Persistent memory for LLM agents.**
+**LLM-supervised persistent memory for AI agents.**
 
 [![Go 1.24+](https://img.shields.io/badge/Go-1.24%2B-00ADD8?logo=go&logoColor=white)](https://go.dev/)
 [![CI](https://github.com/mnemon-dev/mnemon/actions/workflows/ci.yml/badge.svg)](https://github.com/mnemon-dev/mnemon/actions/workflows/ci.yml)
@@ -15,34 +15,32 @@
 
 LLM agents forget everything between sessions. Context compaction drops critical decisions, cross-session knowledge vanishes, and long conversations push early information out of the window.
 
-Mnemon gives your agent persistent, cross-session memory — with a single binary and one setup command.
-
-<p align="center">
-  <img src="docs/diagrams/10-knowledge-graph.jpg" width="720" alt="Knowledge Graph — 87 insights connected by temporal, entity, semantic, and causal edges" />
-  <br />
-  <sub>A real knowledge graph built by Mnemon — 87 insights, 2150 edges across four graph types.</sub>
-</p>
+Mnemon gives your agent persistent, cross-session memory — a four-graph knowledge store with intent-aware recall, importance decay, and automatic deduplication. Single binary, zero API keys, one setup command.
 
 ### Why Mnemon?
 
-Memory has a **compound interest effect** — the longer it accumulates, the greater its value. LLM engines iterate constantly, skill files cost nearly nothing to write, but memory is a private asset that grows with the user. It is the only component in the agent ecosystem worth deep investment.
-
-Mnemon is built on one core belief: **the LLM itself is the best orchestrator.** Rather than embedding a small LLM inside the pipeline, Mnemon lets your host LLM — the one already holding full conversation context — act as supervisor. The binary is the organ (deterministic storage, graph indexing, search, decay); the LLM is the brain (decides what to remember, how to link, when to forget). The skill file is the textbook that teaches the protocol.
-
-This means: **memory management logic moves from prompt to code — deterministic, testable, portable.** The same binary + skill can run on Claude Code, Cursor, or any LLM CLI that reads markdown.
+Most memory tools embed their own LLM inside the pipeline. Mnemon takes a different approach: **your host LLM is the supervisor.** The binary handles deterministic computation (storage, graph indexing, search, decay); the LLM makes judgment calls (what to remember, how to link, when to forget). No middleman, no extra inference cost.
 
 | Pattern | LLM Role | Representative |
 |---|---|---|
 | **LLM-Embedded** | Executor inside the pipeline | Mem0, Letta |
 | **MCP Server** | Tool provider via MCP protocol | claude-mem |
-| **LLM-Supervised** | External supervisor of a standalone binary | Mnemon |
+| **LLM-Supervised** | External supervisor of a standalone binary | **Mnemon** |
 
 <p align="center">
   <img src="docs/diagrams/llm-supervised-concept.jpg" width="720" alt="LLM-Supervised Architecture — three patterns compared, with detailed Mnemon implementation showing hooks, brain/organ split, and sub-agent delegation" />
   <br />
   <sub>The LLM-Supervised pattern: hooks drive the lifecycle, the host LLM makes judgment calls, the binary handles deterministic computation.</sub>
 </p>
 
+Memory has a **compound interest effect** — the longer it accumulates, the greater its value. LLM engines iterate constantly, skill files cost nearly nothing to write, but memory is a private asset that grows with the user. It is the only component in the agent ecosystem worth deep investment.
+
+<p align="center">
+  <img src="docs/diagrams/10-knowledge-graph.jpg" width="720" alt="Knowledge Graph — 87 insights connected by temporal, entity, semantic, and causal edges" />
+  <br />
+  <sub>A real knowledge graph built by Mnemon — 87 insights, 2150 edges across four graph types.</sub>
+</p>
+
 See [Design & Architecture](docs/DESIGN.md) for details.
 
 ## Quick Start

docs/zh/README.md

@@ -14,34 +14,32 @@
 
 LLM 智能体在会话之间会遗忘一切。上下文压缩丢失关键决策，跨会话知识消失，长对话将早期信息推出窗口。
 
-Mnemon 为你的 LLM 提供持久的跨会话记忆 — 只需一个 Go 二进制文件和一条 setup 命令。
-
-<p align="center">
-  <img src="../diagrams/10-knowledge-graph.jpg" width="720" alt="知识图谱 — 87 条洞察通过时序、实体、语义和因果边连接" />
-  <br />
-  <sub>Mnemon 构建的真实知识图谱 — 87 条洞察，2150 条边，横跨四种图类型。</sub>
-</p>
+Mnemon 为你的 LLM 提供持久的跨会话记忆 — 四图知识存储、意图感知检索、重要度衰减、自动去重。单一二进制，零 API 密钥，一条命令完成部署。
 
 ### 为什么选择 Mnemon？
 
-记忆具有**复利效应** — 积累越久，价值越大。LLM 引擎不断迭代，技能文件几乎零成本编写，但记忆是随用户一起增长的私有资产。它是智能体生态中唯一值得深度投入的组件。
-
-Mnemon 基于一个核心理念：**LLM 本身就是最好的编排器。** 不在管线内嵌入小型 LLM，而是让你的宿主 LLM — 对话中已有完整上下文的那个 — 充当监督者。二进制是器官（确定性存储、图索引、搜索、衰减）；LLM 是大脑（决定记什么、怎么关联、何时遗忘）。技能文件是教授协议的教科书。
-
-这意味着：**记忆管理逻辑从提示词迁移到代码 — 确定性的、可测试的、可移植的。** 同一套二进制 + 技能文件可在 Claude Code、Cursor 或任何读取 markdown 的 LLM CLI 上运行。
+多数记忆工具在管线内嵌入自己的 LLM。Mnemon 采用不同路线：**你的宿主 LLM 就是监督者。** 二进制处理确定性计算（存储、图索引、搜索、衰减）；LLM 做判断（记什么、怎么关联、何时遗忘）。没有中间人，没有额外推理开销。
 
 | 模式 | LLM 角色 | 代表项目 |
 |---|---|---|
 | **LLM-Embedded** | 管线内部的执行者 | Mem0, Letta |
 | **MCP Server** | 通过 MCP 协议提供工具 | claude-mem |
-| **LLM-Supervised** | 独立二进制的外部监督者 | Mnemon |
+| **LLM-Supervised** | 独立二进制的外部监督者 | **Mnemon** |
 
 <p align="center">
   <img src="../diagrams/llm-supervised-concept.jpg" width="720" alt="LLM 监督式架构 — 三种模式对比，及 Mnemon 实现细节：钩子、大脑/器官分离、Sub-agent 委派" />
   <br />
   <sub>LLM 监督式模式：钩子驱动生命周期，宿主 LLM 做判断，二进制处理确定性计算。</sub>
 </p>
 
+记忆具有**复利效应** — 积累越久，价值越大。LLM 引擎不断迭代，技能文件几乎零成本编写，但记忆是随用户一起增长的私有资产。它是智能体生态中唯一值得深度投入的组件。
+
+<p align="center">
+  <img src="../diagrams/10-knowledge-graph.jpg" width="720" alt="知识图谱 — 87 条洞察通过时序、实体、语义和因果边连接" />
+  <br />
+  <sub>Mnemon 构建的真实知识图谱 — 87 条洞察，2150 条边，横跨四种图类型。</sub>
+</p>
+
 详见 [设计与架构](DESIGN.md)。
 
 ## 快速开始