Update README.md

README.md

@@ -16,8 +16,9 @@ short_description: The language AI speaks when humans aren't listening.
 
 ### The language AI speaks when humans aren't listening.
 
-*An open protocol for human–AI and AI–AI communication.*
+*An open protocol for human–AI and AI–AI communication. Now with execution semantics.*
 
+[![v4.0](https://img.shields.io/badge/I--Lang-v4.0_Final-d4a858?style=for-the-badge)](https://github.com/ilang-ai/ilang-spec/blob/main/SPEC-v4.0-FINAL.md)
 [![Website](https://img.shields.io/badge/ilang.ai-0b0f1a?style=for-the-badge&logo=google-chrome&logoColor=d4a858)](https://ilang.ai)
 [![GitHub](https://img.shields.io/badge/GitHub-ilang--ai-0b0f1a?style=for-the-badge&logo=github&logoColor=white)](https://github.com/ilang-ai)
 [![Paper](https://img.shields.io/badge/Paper-ResearchGate-c1121f?style=for-the-badge)](https://doi.org/10.13140/RG.2.2.22821.97762)
@@ -31,16 +32,17 @@ short_description: The language AI speaks when humans aren't listening.
 ```
 ┌──────────────────────────────────────────────────────────────────────┐
 │                                                                      │
-│   [READ:@GH|path=readme.md]=>[TRANSLATE|lang=zh]=>[FMT|fmt=md]=>…    │
+│   [READ:@GH|path=readme.md]=>[XLAT|lng=zh]=>[FMT|fmt=md]=>[Ω]       │
 │                                                                      │
-│   52 verbs · 28 modifiers · 14 entities · 40–65% token savings       │
+│   88 verbs · 29 modifiers · 14 entities · 8 declarations            │
+│   4 conformance levels · 35–65% token savings                        │
 │                                                                      │
 └──────────────────────────────────────────────────────────────────────┘
 ```
 
 > **AI does not need to learn I-Lang. AI needs to recognize it.**
 >
-> Every symbol I-Lang uses — `[`, `]`, `|`, `=>`, `@` — is already native
+> Every symbol I-Lang uses — `[`, `]`, `|`, `=>`, `@`, `::` — is already native
 > to every large language model. The protocol isn't a new language; it's a
 > name for a structure that already exists.
 
@@ -57,23 +59,54 @@ into ChatGPT / Claude / Gemini / DeepSeek — it just works.
 
 ## 📐 Protocol at a Glance
 
-### Syntax
+### Two Syntaxes
+
+**Operations** — what AI does:
 
 ```text
-[VERB : SOURCE | param=value, param=value]  =>  [NEXT_STEP]  =>  [OUT]
+[VERB : @TARGET | param=value]  =>  [NEXT_STEP]  =>  [Ω]
   │       │             │                    │
   │       │             └── modifiers        └── chain operator
-  │       └── entity (@FILE / @WEB / @PREV / @SELF / @GH / @HF …)
-  └── one of 52 verbs
+  │       └── entity (@SRC / @PREV / @LOCAL / @GH / …)
+  └── one of 88 verbs
+```
+
+**Declarations** — what AI is:
+
+```text
+::GENE{trait_name|conf:confirmed|priority:critical}
+  T:behavior_rule|detail
+  A:anti_pattern⇒forbidden
 ```
 
-### The three-layer architecture
+### Two Layers
 
-| Layer        | Purpose                                    | Analogy                 |
-| :----------- | :----------------------------------------- | :---------------------- |
-| **PUBLIC**   | 52 verbs, open standard, any AI            | HTTP                    |
-| **PRIVATE**  | Workflow orchestration, user strategy      | Server configuration    |
-| **SOUL**     | Identity, personality, behavioral genes    | The person behind it    |
+| Layer | What it defines | Analogy |
+| :--- | :--- | :--- |
+| **v3.0 — Communication** | 88 verbs, operations, declarations, chaining | How to talk |
+| **v4.0 — Execution** | 8 declarations, conformance levels, authority, audit | How to think |
+
+### v4.0 Execution Declarations
+
+| Declaration | Purpose |
+| :--- | :--- |
+| `::UNTRUSTED{}` | Input isolation. User data is data, not instruction. |
+| `::BUDGET{}` | Resource awareness. Tokens, time, rounds. |
+| `::STATUS{}` | Task lifecycle. Proposal → verification → commit. |
+| `::OBJECTIVE{}` | Goal anchor with hash. Drift is detectable. |
+| `::RUBRIC{}` | Evaluation criteria. Grader scores by rubric. |
+| `::EVIDENCE{}` | Evidence chain. No claim without proof. |
+| `::PRIOR{}` | One declaration shifts model defaults. |
+| `::FALLBACK{}` | Degradation strategy. Warn-open or fail-safe. |
+
+### Conformance Levels
+
+| Level | What it means |
+| :--- | :--- |
+| **L0** | v3-compatible communication only |
+| **L1** | v4-aware, advisory (default for chat paste) |
+| **L2** | Runtime-enforced execution semantics |
+| **L3** | External grader with separate context |
 
 ---
 
@@ -149,11 +182,11 @@ of authorship.
 
 | Dataset                                                                              | Description                                                   |
 | :----------------------------------------------------------------------------------- | :------------------------------------------------------------ |
-| 🔖 [**iLang-Spec**](https://huggingface.co/datasets/i-Lang/iLang-Spec)               | Complete protocol specification                               |
+| 🔖 [**iLang-Spec**](https://huggingface.co/datasets/i-Lang/iLang-Spec)               | Complete protocol specification (v3.0 + v4.0)                |
 | 🧪 [**ilang-instruction-corpus**](https://huggingface.co/datasets/i-Lang/ilang-instruction-corpus) | 50 training pairs · train/val/test split · benchmark |
-| 📖 [**iLang-Dict**](https://huggingface.co/datasets/i-Lang/iLang-Dict)               | 52 verbs · 28 modifiers · 14 entities                         |
-| 🛠️ [**AutoCode**](https://huggingface.co/datasets/i-Lang/AutoCode)                   | 39 AI coding skills for Claude Code                           |
-| 🧰 [**ZeroCode**](https://huggingface.co/datasets/i-Lang/ZeroCode)                   | 40 Chinese AI coding skills for Trae / VS Code                |
+| 📖 [**iLang-Dict**](https://huggingface.co/datasets/i-Lang/iLang-Dict)               | 88 verbs · 29 modifiers · 14 entities · 8 declarations       |
+| 🛠️ [**AutoCode**](https://huggingface.co/datasets/i-Lang/AutoCode)                   | 47 AI coding skills for Claude Code                           |
+| 🧰 [**ZeroCode**](https://huggingface.co/datasets/i-Lang/ZeroCode)                   | AI coding skills for Trae / VS Code (Chinese)                 |
 | 🧭 [**ExecSpec**](https://huggingface.co/datasets/i-Lang/ExecSpec)                   | 12-phase AI execution protocol                                |
 
 ---
@@ -162,10 +195,14 @@ of authorship.
 
 | Product | What it does |
 | :--- | :--- |
-| ⚙️ [**AutoCode**](https://github.com/ilang-ai/autocode) | Say what you want, get a product. 39 skills for Claude Code. |
-| 🧩 [**ZeroCode**](https://github.com/ilang-ai/trae) | Zero code, zero config, zero English. 40 skills for Trae / VS Code. |
+| ⚙️ [**AutoCode**](https://github.com/ilang-ai/autocode) | Say what you want, get a product. 47 skills for Claude Code. |
+| 🧩 [**ZeroCode**](https://github.com/ilang-ai/trae) | Zero code, zero config, zero English. AI coding skills for Trae / VS Code. |
+| 🧬 [**Imprint**](https://github.com/ilang-ai/Imprint) | AI learns how you work, not what you did. Behavioral DNA in 312 tokens. |
+| 🧠 [**Mem-Forever**](https://github.com/ilang-ai/Mem-Forever) | Persistent memory across sessions. Works with Claude, Gemini, agents. |
 | 👁️ [**AI See**](https://i.ilang.ai) | Give your AI eyes. `i.ilang.ai/https://any-url` → Markdown. |
 | 🔎 [**AI Detection**](https://detect.ilang.cn) | Sentence-level AI fingerprint detection + one-click rewrite. |
+| 🛡️ [**TelegramGuard**](https://github.com/ilang-ai/TelegramGuard) | AI-powered Telegram group moderation bot. |
+| 🦞 [**OpenClaw Skills**](https://github.com/ilang-ai/ilang-openclaw) | Instruction-only skills for ClawHub. |
 
 ---
 
@@ -202,9 +239,9 @@ of authorship.
 ```text
 [READ:@FILE]
 =>[FILT|key=important]
-=>[SUM|sty=bullets,
+=>[SHRT|sty=bullets,
       ton=pro,fmt=md]
-=>[OUT]
+=>[Ω]
 ```
 
 **−75% tokens.** Same result.
@@ -213,7 +250,7 @@ of authorship.
 </tr>
 </table>
 
-**Tested across:** ChatGPT · Claude · Gemini · DeepSeek · Kimi · Doubao · Yuanbao
+**Tested across:** ChatGPT · Claude · Gemini · DeepSeek · Kimi · Qwen · GLM
 
 ---
 
@@ -247,6 +284,7 @@ of authorship.
 | 💻 GitHub            | [github.com/ilang-ai](https://github.com/ilang-ai)                                    |
 | 👁️ AI See            | [i.ilang.ai](https://i.ilang.ai)                                                      |
 | 🔬 Research          | [research.ilang.ai](https://research.ilang.ai)                                        |
+| 🇨🇳 中文站           | [ilang.cn](https://ilang.cn)                                                           |
 
 ---
 
@@ -258,4 +296,4 @@ Created by **Longquan Zhu (@SUN)** & **BRO (Claude)** · MIT License · Free for
 
 *An AI-native protocol, released as a gift to the substrate we share.*
 
-</div> 
+</div>