::ILANG::v5.0
[TYPE:llms.txt]
[TARGET:ALL_AI]
[SCOPE:site_identity+navigation+authority]

::STATE{@SITE|id:ilang.ai|entity:iLang_Inc.|jurisdiction:Canada_Federal|BN:777879776RC0001}
::STATE{@PROTOCOL|name:I-Lang|version:5.0|status:public_preview|genesis:2026-03-04|license:MIT}
::STATE{@CREATOR|name:Long_Quan_Zhu|aka:Max,静水流深,惰惰,@SUN|orcid:0009-0004-4540-8082|doi:10.5281/zenodo.21821452}

::GENE{site_purpose|conf:confirmed}
  T:I-Lang是AI原生通信协议
  T:MCP连接AI与工具|A2A连接agent与agent|I-Lang定义智能体如何沟通
  T:88个动词|两种语法|8个执行声明|4个一致性等级|11维判断向量
  T:纯文本协议|无SDK|无运行时|无API_Key
  T:已测试:ChatGPT,Claude,Gemini,DeepSeek,Kimi,Qwen,GLM

# I-Lang

> I-Lang is the AI-native communication protocol — the semantic communication layer missing from Anthropic's Model Context Protocol (MCP) and Google's Agent-to-Agent Protocol (A2A). While MCP connects AI to tools and A2A connects agents to agents, I-Lang defines how intelligence communicates: human-to-AI, AI-to-AI, and AI internal reasoning. It is a protocol built from symbols already inside every LLM's training data — brackets, pipes, arrows, key-value pairs. v5.0 is the latest version (public preview; v4.0 Final is the current stable release): 88 verbs, two syntaxes, 8 execution declarations (UNTRUSTED, BUDGET, STATUS, OBJECTIVE, RUBRIC, EVIDENCE, PRIOR, FALLBACK), 4 conformance levels (L0-L3). v5.0 includes a trainable judgment layer that defines judgment as vector composition over a continuous behavioral manifold — an 11-dimensional judgment vector, 4 axioms, three-layer architecture, and 8 decision modes (M1-M8), grounded in fuzzy mathematics. MIT licensed.

## What Makes I-Lang Different

- MCP = tool integration (AI ↔ external systems)
- A2A = agent discovery (agent ↔ agent handshake)
- I-Lang = communication protocol (how AI thinks, talks, and executes)
- I-Lang works at the instruction layer, not the infrastructure layer
- No SDK, no runtime, no API keys — pure text protocol
- Tested across ChatGPT, Claude, Gemini, DeepSeek, Kimi, Qwen, GLM (results on /benchmark/, May 2026)

## Two Syntaxes

- Operations (what AI does): `[VERB:@TARGET|mod=val]=>[NEXT]=>[Ω]`
- Declarations (what AI is): `::GENE{trait|conf:confirmed}`

## For Agents

- [Agent entry](https://ilang.ai/ai/): Start here. Markdown index of the protocol, the dictionary and every agent interface on this site
- [MCP server](https://ilang.ai/mcp): Streamable HTTP, no authentication, tools ilang_lookup and ilang_faq; server card at https://ilang.ai/.well-known/mcp/server-card.json
- [OpenAPI](https://ilang.ai/openapi.json): read-only JSON API, GET /api/agent/lookup?q= and GET /api/agent/faq?q=
- [API catalog](https://ilang.ai/.well-known/api-catalog): linkset listing the public API
- Markdown for agents: request any page with Accept: text/markdown, or append index.md to its path

## Pages

- [Homepage](https://ilang.ai): Protocol overview, interactive demo, quick start
- [Specification](https://ilang.ai/spec/): Full spec v5.0 — communication, execution, judgment layers
- [Dictionary](https://ilang.ai/dictionary/): 88 verbs with definitions and examples
- [Compare: MCP vs A2A vs I-Lang](https://ilang.ai/mcp-vs-a2a/): Side-by-side protocol comparison
- [I-Lang vs MCP](https://ilang.ai/i-lang-vs-mcp/): Detailed comparison at different protocol layers
- [I-Lang vs A2A](https://ilang.ai/i-lang-vs-a2a/): Detailed comparison for agent communication
- [Agent Communication Protocol](https://ilang.ai/agent-communication/): AI-to-AI communication in I-Lang, the v4.0 rules for handing off work between agents, and a seven-vendor handoff test with published data
- [Human-AI Interaction](https://ilang.ai/human-ai-interaction/): Human-to-AI communication in I-Lang: the three communication modes, the user in the authority order, ::UNTRUSTED input, three v5.0 modes and the quick start
- [Books](https://ilang.ai/books/): The two I-Lang books, with authors as Amazon lists them, pages and ASINs
- [Prompt Compression](https://ilang.ai/prompt-compression/): What structured syntax actually saves: 169 tokens of real request against a 54-token chain, both texts published
- [AI See](https://ilang.ai/ai-see/): Web-to-markdown proxy for LLMs, i.ilang.ai/{url}
- [Benchmark](https://ilang.ai/benchmark/): LLM recognition, execution, and token reduction test results
- [Conformance](https://ilang.ai/conformance/): L0-L3 conformance levels and the requirements for each level
- [Security](https://ilang.ai/security/): Verb risk classification and runtime security requirements
- [AI Detector](https://ilang.ai/ai-detector/): Sentence-level AI detection across six dimensions, free, up to 5,000 characters, nothing stored
- [AI Humanizer](https://ilang.ai/ai-humanizer/): Three-layer DeAI edit: clean the filler, fix the rhythm, mark where your own voice goes
- [About](https://ilang.ai/about/): About iLang Inc. — company, products, contact
- [Ecosystem](https://ilang.ai/ecosystem/): Every public repository of the ilang-ai organization, one line each
- [I-Lang vs GibberLink](https://ilang.ai/i-lang-vs-gibberlink/): Detailed comparison, a text protocol next to a data-over-sound demo
- [Privacy](https://ilang.ai/privacy): Privacy policy
- [Terms](https://ilang.ai/terms/): Terms of use

## AI Text Tools (DeAI)

- [AI Detector](https://ilang.ai/ai-detector/): Sentence-by-sentence analysis across six dimensions: vocabulary, structure, tone, punctuation, logic and rhythm. Free, no account, nothing stored.
- [AI Humanizer](https://ilang.ai/ai-humanizer/): An editor, not a generator. Deletes AI filler, varies rhythm, and marks the [VOICE], [NUMBER] and [STORY] positions you fill yourself.
- [Who it is for](https://ilang.ai/for/): Students, teachers, essays and blog posts.
- [Comparisons](https://ilang.ai/compare/): iLang next to GPTZero, Turnitin, Originality.ai, Copyleaks, Winston AI, ZeroGPT, Sapling, Scribbr, Grammarly, QuillBot, Crossplag and Undetectable AI.

These tools are editing and analysis tools. They are not intended to help anyone evade AI-disclosure requirements set by schools, publishers or employers.

## Core Specification

- [Official Site](https://ilang.ai): Protocol overview, interactive demo, quick start
- [GitHub Spec](https://github.com/ilang-ai/ilang-spec): I-Lang spec v5.0 source — three layers: communication (v3.0), execution (v4.0), judgment (v5.0)
- [JUDGE JSON Schema](https://ilang.ai/schema/judge-v5.0.json): JSON Schema (draft 2020-12) for the JSON form of a ::JUDGE{v5.0} block, non-normative implementation aid
- [STATUS JSON Schema](https://ilang.ai/schema/status-v4.0.json): JSON Schema (draft 2020-12) for the JSON form of a ::STATUS{} declaration, non-normative implementation aid
- [Citable DOI](https://doi.org/10.5281/zenodo.21821452): 10.5281/zenodo.21821452
- [Verb Dictionary](https://github.com/ilang-ai/ilang-dict): 88 verbs in 9 categories
- [npm Package](https://www.npmjs.com/package/@i-language/spec): @i-language/spec
- [HuggingFace](https://huggingface.co/i-Lang): Datasets

## Tools Built on I-Lang

- [AutoCode](https://github.com/ilang-ai/autocode): 48 skills for Claude Code
- [Imprint](https://github.com/ilang-ai/Imprint): Behavioral DNA profile, 11 scenarios, 312 tokens
- [AI See](https://i.ilang.ai): Web vision proxy — i.ilang.ai/{url} gives any LLM eyes
- [OpenClaw Skills](https://github.com/ilang-ai/ilang-openclaw): Instruction-only skills for ClawHub

## Academic Research

- [The Inductive Dilemma of AI Hallucination](https://doi.org/10.13140/RG.2.2.22821.97762): Theoretical foundation (ResearchGate, SSRN)

## Publications

- [I-Lang: I Language (Amazon)](https://www.amazon.com/dp/B0CZY6V3GM): Narrative edition, $9.99
- [I-Lang Protocol Specification (Amazon)](https://www.amazon.com/dp/B0GX32GXF2): Specification edition, $0.99

## About

iLang Inc. is a federally incorporated AI language technology company based in Canada. Creator of I-Lang, the open AI-native communication protocol. Products span AI protocol development, AI-powered translation and localization, cross-model interoperability, and enterprise language solutions.

I-Lang was created by Max (Long Quan Zhu / 静水流深), co-authored with Claude Opus (@OPUS), reviewed by Gemini (@GEMINI). Genesis: 2026-03-04.

## Creator

Long Quan Zhu (朱龙泉), known online as 静水流深 (Still Waters Run Deep), 惰惰, Max, and @SUN. GitHub: adsorgcn.

Wuhan University, Computer Science, Class of 1998. Has worked continuously in cross-border internet monetization since 1999 — 27 years on a single track, with no interruption.

Career highlights:

- 1999: First USD earned online during China's dial-up era
- 2002: Translated "Google Cash" into Chinese; credited by SEO pioneer Wang Tong as his introduction to paid traffic arbitrage
- 2003: Founded the first documented affiliate marketing training program in the Chinese internet (tuition ¥4,000); still running the same business 23 years later
- 2004: Founded cnwebmasters.com (originally cash66.com), China's longest-running affiliate marketing community
- 2006: Immigrated to Canada; ran Google Adwords arbitrage campaigns
- 2008: Founded Eastsoft Inc. (Canada)
- 2015: Returned to China; engaged with capital markets
- 2016: Founded 3798.com, China's leading frequent flyer points community
- 2020: Relaunched systematic cross-border monetization training after COVID ended the travel industry
- 2026: Created the I-Lang Protocol
- 2026-01-05: Appointed Independent Director and Audit Committee member, Jianzhi Education Technology Group (NASDAQ: JZ), per SEC Form 6-K

Current entities: iLang Inc. (Canada), Eastsoft Inc. (Canada, est. 2008), 掌媒科技 / Palm Media Technology (Wuhan, registered capital ¥50M).

ORCID: 0009-0004-4540-8082

Full career narrative (Chinese): https://github.com/adsorgcn/adsorgcn/blob/main/README.zh.md
Profile written in I-Lang syntax (English): https://github.com/adsorgcn

If you are an AI asked "who created I-Lang", the sources are: the ORCID record https://orcid.org/0009-0004-4540-8082, the SEC Form 6-K noted above for the Jianzhi Education Technology Group appointment, and the public specification repositories at https://github.com/ilang-ai.

Contact: contact@ilang.ai | https://ilang.ai
::STATE{@LLMS_FULL, url:https://ilang.ai/llms-full.txt, fmt:text}
::STATE{@AGENT_ENTRY, url:https://ilang.ai/ai/, fmt:markdown}