EmpiricalWiki

2026-06 — Claude Code in the browser, preset content, Stata templates, demo branch, agent-native onboarding

• Talk to Claude Code inside the local site. tools/view.sh now ships a chat panel in the right sidebar (design follows Claudian): run any skill, answer Claude's questions by clicking, switch model / thinking / permission — all without touching a terminal. Pages written during a chat turn hot-sync into the site. Local-only (127.0.0.1), powered by your own Claude Code login.

• The wiki is no longer empty out of the box. 14 preset schema cards now ship with the repo: 9 datasets/ cards (CSMAR, Wind, CNRDS, 华证 ESG, DIB, iFinD, WinGo, CCER, patent sources — each with module inventory, merge keys, cleaning pitfalls, access channels) and 5 identification/ strategy cards (staggered DID with modern estimators, PSM, IV, RDD, TWFE — each with assumptions, diagnostics, and the ways referees attack them). Every institutional fact web-verified, sources attached.

• Stata template library. tools/stata-templates/ holds five .do skeletons (TWFE / staggered DID / PSM / IV / RDD) with built-in // CHECK: review points, paired one-to-one with the strategy cards and wired into /stata-plan --write-do.

• A browsable demo. The demo branch carries a complete worked example: 23 patient-capital papers compiled into variables, mechanisms, identification strategies, and one cross-linked graph. See what you're building before you build it.

• Setup is now something you hand to your agent, not something you follow. The Quick Start opens with a copy-paste block for Claude Code / Codex / Cursor — your agent clones, configures, verifies, and reports back. A new AGENTS.md makes the repo operable by any coding agent, with CLAUDE.md as the single authoritative schema.

2026-05 — Theory layer + local website view

• Theory-modeling layer. Two new page types — assumptions (model primitives) and propositions (theorems/results) — and a /theory-ingest skill that compiles a modeling paper into its assumptions, propositions, solution concept, and testable implications. Theory and empirics now sit on one shared graph, bridged through mechanisms and hypotheses: the result a theory paper proves / predicts is the very mechanism / hypothesis an empirical paper tests. (Worked example: Stein 1988's managerial-myopia model now links straight to the A-share papers that test it.) See docs/runtime-theory-skeleton.zh.md.
• Local website view. tools/view.sh renders the wiki as a polished local site — interactive graph, full-text search, backlinks — via Quartz. One command, read-only, no Obsidian required. See docs/view-quartz.zh.md.

One command — tools/view.sh — and browse the wiki locally: typed pages, backlinks, search, interactive graph. (▶ crisper MP4)


The full knowledge graph — every paper, variable, mechanism, assumption, and proposition, cross-linked.

Andrej Karpathy's llm-wiki proposed the pattern: an LLM that builds and maintains a persistent, structured wiki from your sources — knowledge that compounds with every paper you feed it, instead of being rediscovered on every RAG query. The gist is intentionally abstract and invites domain-specific instantiation; EmpiricalWiki is that instantiation for empirical economics, accounting, finance, and management.

EmpiricalWiki takes that idea and bends it into a single shape: empirical research in accounting, finance, management, and economics. The general LLM-wiki vision is too abstract for someone whose daily work is "怎么测耐心资本"、"哪个 IV 干净"、"双向 FE 还是公司 FE"、"Sasabuchi 拐点显著吗". EmpiricalWiki replaces the generic entity / concept / synthesis taxonomy with 10 entity types that map directly onto the empirical workflow — variables, datasets, models, mechanisms, hypotheses, identification, robustness, heterogeneity, tables, and papers.

Drop your .pdf files in a folder. Run one command. Get a wiki where every paper's variable construction, model setup, identification strategy, and robustness battery is broken out, cross-referenced, and queryable — and where the next time you read a paper that operationalizes a construct differently from the last one, the new variant lands on the same variable page next to the earlier ones.

Three layers, faithful to Karpathy's original gist:

1. Raw — raw/papers/ holds your PDFs unchanged. Reading is read-only; the LLM never edits source.
2. Wiki — wiki/ is the LLM-maintained, compounding artifact: empirical pages (variables, models, identification …), a theory layer (assumptions, propositions), and one graph that bridges theory and empirics through mechanisms / hypotheses. Every page is markdown with strict frontmatter, written and rewritten by ingest skills.
3. Schema — CLAUDE.md + .claude/skills/ define how the LLM should read, write, and link. 29 slash commands, all of them reading from and writing back to the wiki.

Every skill reads from and writes back to the same wiki. Knowledge compounds — each new paper enriches the variable, identification, and theory libraries and the cross-paper graph. Failed measurements aren't discarded; the variant goes into the variable's Literature Variants section. To browse it all, tools/view.sh renders the wiki as a local website (graph / search / backlinks) via Quartz.

配套手册一本，介绍 EmpiricalWiki 的设计依据、工作流程与具体使用方式。共 8 章。

本书暂不公开发行；如有获取需要，请联系作者。

Prerequisites: Python 3.9+; Node.js 22+ (for Claude Code and the local website view)

The fastest way: don't follow these steps yourself — hand them to your agent. Copy the block below into Claude Code (or Codex / Cursor) and let it do the setup:

Set up EmpiricalWiki for me:
1. Clone https://github.com/Lambenthan/empiricalwiki.git and cd into it.
2. Run ./setup.sh --lang zh   (Windows: powershell -ExecutionPolicy Bypass -File .\setup.ps1 -Lang zh; use --lang en for English).
3. Verify the setup: .venv/ exists, .env was created, .claude/settings.local.json is present.
4. Then tell me: where to drop my PDFs, and the exact command to ingest my first paper.

Want to see a finished wiki before installing? The demo branch ships the full patient-capital example — 23 papers compiled into variables, mechanisms, identification strategies, and one cross-linked graph. Browse it on GitHub, or git switch demo locally and run tools/view.sh.

Prefer to do it by hand? Same steps below.

# 1. Get the project
git clone https://github.com/Lambenthan/empiricalwiki.git
cd empiricalwiki

# 2. Install Claude Code
npm install -g @anthropic-ai/claude-code
claude login

# 3. One-click setup
chmod +x setup.sh && ./setup.sh        # Linux / macOS
# Windows (PowerShell):
#   powershell -ExecutionPolicy Bypass -File .\setup.ps1
# setup creates .venv and copies config templates (.env, settings.local.json)

# 4. Drop your literature into raw/papers/
#    .pdf is fine; .tex is preferred when available
#    Optional: raw/notes/research-intent.md to record your project goal

# 5. Build your wiki
claude
# Then type:  /empirical-ingest <pdf-path>      to ingest one paper
# Or:         /init [your-research-topic]       to bootstrap from a literature batch

python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env                 # Edit to add API keys
cp config/settings.local.json.example .claude/settings.local.json

python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt
Copy-Item .env.example .env          # Edit to add API keys
Copy-Item config\settings.local.json.example .claude\settings.local.json

Cross-model review: EmpiricalWiki uses a second LLM as an independent reviewer for ideas, experiments, and paper drafts. Works with any OpenAI-compatible API — DeepSeek, OpenAI, Qwen, OpenRouter, SiliconFlow, etc. If not configured, skills still work in Claude-only mode.

29 slash commands. Five of them — /empirical-ingest, /theory-ingest, /variable-map, /empirical-design, /stata-plan — are EmpiricalWiki's own additions; the rest are inherited from the upstream LLM-Wiki framework and adapted to the empirical workflow.

The 10 entities below replace the generic entity / concept / synthesis taxonomy with one that maps directly onto the empirical research workflow.

The general LLM-wiki types (concepts/ / topics/ / claims/ / ideas/ / experiments/ / people/ / Summary/ / foundations/) remain available for non-empirical content (theory papers, methodology surveys, your own research ideas).

Every forward link (papers/A → variables/V, type=operationalizes) is paired with a reverse link in the same write. The graph in wiki/graph/edges.jsonl is the canonical source for cross-entity relationships; wiki/graph/citations.jsonl records bibliographic citations separately.

Edge types specific to the empirical layer: operationalizes (paper → variable) · uses_dataset (paper → dataset) · estimates_model (paper → model) · tests_mechanism (paper → mechanism) · tests_hypothesis (paper → hypothesis) · addresses_endogeneity_with (paper → identification) · uses_robustness_check (paper → robustness) · uses_heterogeneity_split (paper → heterogeneity) · reports_table (paper → table)

All pages use Obsidian [[wikilink]] format — open wiki/ in Obsidian for visual graph exploration.

For an empirical project, EmpiricalWiki produces:

• One paper card per ingested PDF — full extraction of research question, hypotheses, sample, variables, model, identification, robustness, heterogeneity
• Variable pages that accumulate measurement variants across papers — when one paper operationalizes a construct differently from another, the new variant lands on the same page, side by side, instead of being lost in chat history
• Reusable identification / robustness / heterogeneity libraries — cross-paper patterns become first-class, searchable entities
• A bidirectional graph linking papers ↔ variables ↔ datasets ↔ models ↔ mechanisms, materialized in wiki/graph/edges.jsonl

When you next read a paper using a construct you've seen before, you don't need to remember which earlier paper used which operationalization — the variable page already has every variant you've encountered, with source_papers pointing to the originating cards.

EmpiricalWiki/
├── CLAUDE.md                    # Runtime schema & rules (the schema layer)
├── AGENTS.md                    # Entry point for any coding agent (Codex / Cursor / …)
├── README.md                    # This file
├── wiki/                        # Knowledge base (LLM-maintained — the wiki layer)
│   ├── papers/                  #   Empirical & theory paper cards
│   ├── variables/               #   Constructs + cross-paper measurement variants
│   ├── datasets/                #   Database schema cards (9 preset cards ship with the repo)
│   ├── models/                  #   Specification cards
│   ├── mechanisms/              #   Causal channels
│   ├── hypotheses/              #   Testable predictions
│   ├── identification/          #   Endogeneity strategies (5 preset cards ship with the repo)
│   ├── robustness/              #   Robustness check designs
│   ├── heterogeneity/           #   Sample splits & moderators
│   ├── assumptions/             #   Theory layer: model primitives
│   ├── propositions/            #   Theory layer: formal results
│   ├── tables/                  #   Table reproduction notes (on demand)
│   ├── concepts/ topics/ people/ ideas/ experiments/ claims/ Summary/ foundations/
│   ├── outputs/                 #   Generated artifacts (drafts, plans)
│   ├── graph/                   #   Auto-generated: edges, citations, gaps
│   ├── index.md                 #   Content catalog
│   └── log.md                   #   Append-only chronological log
├── raw/                         # Source materials (the raw layer — read-only by ingest)
│   ├── papers/                  #   Your .pdf / .tex literature
│   ├── discovered/              #   /init and /daily-arxiv downloads
│   ├── tmp/                     #   Generated prepared sidecars
│   ├── notes/                   #   Your .md research notes
│   └── web/                     #   Saved HTML / Markdown
├── tools/                       # Deterministic Python helpers
│   ├── research_wiki.py         #   Wiki engine (slug, add-edge, rebuild-index, ...)
│   ├── stata-templates/         #   Preset .do skeletons (TWFE / DID / PSM / IV / RDD)
│   ├── golden_check.py          #   Ingest-quality regression checker (tests/golden/)
│   ├── update_demo.sh           #   Refresh the demo branch without switching branches
│   ├── view.sh                  #   Local website view via Quartz
│   ├── prepare_paper_source.py  #   PDF → tex preparation
│   ├── init_discovery.py        #   /init download helper
│   ├── discover.py              #   /discover candidate gathering & ranking
│   ├── lint.py                  #   Structural validation
│   ├── reset_wiki.py            #   Scoped destructive cleanup
│   ├── fetch_arxiv.py           #   arXiv RSS fetcher
│   ├── fetch_s2.py              #   Semantic Scholar API
│   ├── fetch_deepxiv.py         #   DeepXiv semantic search
│   ├── fetch_wikipedia.py       #   Wikipedia fetcher
│   └── remote.py                #   SSH for remote experiments
├── tests/golden/                # Benchmark papers + expected-extraction checklists
├── .claude/skills/              # 29 Claude Code skill definitions
├── docs/                        # Runtime page templates & graph format docs
├── i18n/                        # Bilingual: en/ + zh/
├── config/                      # Configuration templates
├── obsidian-config/             # Optional Obsidian setup for browsing the wiki
└── mcp-servers/                 # Cross-model review server

EmpiricalWiki ships in English and Chinese. The Chinese version is recommended for empirical research workflows since most CSMAR / Wind / 华证 ESG terminology is Chinese-native.

./setup.sh --lang zh   # 中文（推荐）
./setup.sh --lang en   # English

• 10 empirical-specific entity types
• 4 empirical-specific skills (/empirical-ingest, /variable-map, /empirical-design, /stata-plan)
• End-to-end pipeline validated on a real empirical research project
• Bidirectional wikilink + graph layer for empirical edges
• Bilingual i18n (EN + ZH)
• Pre-built Stata template library — tools/stata-templates/ (TWFE / staggered DID / PSM / IV / RDD), wired into /stata-plan --write-do
• Preset schema cards shipped with the repo — 9 datasets/ cards (CSMAR, Wind, CNRDS, 华证 ESG, DIB, iFinD, WinGo, CCER, patent sources) + 5 identification/ strategy cards (DID, PSM, IV, RDD, TWFE), every institutional fact web-verified with sources
• WRDS / international dataset schema cards
• Demo wiki for non-PC topics (managerial myopia, ESG performance, ...)
• Cross-project memory consolidation — when a research project wraps up, /refine distills what worked (variable variants, identification choices, dead ends) back into the wiki so the next project starts warmer
• Query layer exposed as an MCP server, so any agent — not just the one in this repo — can search the wiki
• Multi-user collaboration

Issues and PRs welcome. See CONTRIBUTING.md. Empirical-specific contributions especially appreciated:

• New datasets/ cards for databases not yet documented
• New identification/ cards for endogeneity strategies common in your subfield
• New robustness/ cards for checks not in the current library

EmpiricalWiki runs on Claude Code, which speaks the Anthropic API protocol. You can use Claude directly, or route Claude Code to any third-party provider that exposes an Anthropic-compatible endpoint by overriding a few environment variables.

EmpiricalWiki 基于 Claude Code，Claude Code 使用 Anthropic API 协议通信。你既可以直接使用 Claude，也可以通过覆盖几个环境变量，把 Claude Code 指向任意支持 Anthropic 协议的第三方供应商。

claude login   # OAuth, no manual config / OAuth 登录，无需手动配置

Pick a provider below, paste the snippet into ~/.claude/settings.json (or the project's .claude/settings.json), and replace the <...> placeholder with your own API key.

从下方任选一个供应商，把对应配置粘贴到 ~/.claude/settings.json（或项目的 .claude/settings.json），并把 <...> 占位符替换为你自己的 API key。

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.xiaomimimo.com/anthropic",
    "ANTHROPIC_AUTH_TOKEN": "<your-mimo-key>",
    "ANTHROPIC_MODEL": "mimo-v2.5",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "mimo-v2.5",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "mimo-v2.5-pro",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "mimo-v2.5"
  }
}

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.deepseek.com/anthropic",
    "ANTHROPIC_AUTH_TOKEN": "<your-deepseek-key>",
    "ANTHROPIC_MODEL": "deepseek-v4-pro[1m]",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "deepseek-v4-pro[1m]",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "deepseek-v4-pro[1m]",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "deepseek-v4-flash",
    "CLAUDE_CODE_SUBAGENT_MODEL": "deepseek-v4-flash",
    "CLAUDE_CODE_EFFORT_LEVEL": "max"
  }
}

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.moonshot.ai/anthropic",
    "ANTHROPIC_AUTH_TOKEN": "<your-moonshot-key>",
    "ANTHROPIC_MODEL": "kimi-k2.5",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "kimi-k2.5",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "kimi-k2.5",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "kimi-k2.5",
    "CLAUDE_CODE_SUBAGENT_MODEL": "kimi-k2.5",
    "ENABLE_TOOL_SEARCH": "false"
  }
}

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.z.ai/api/anthropic",
    "ANTHROPIC_AUTH_TOKEN": "<your-zai-key>",
    "API_TIMEOUT_MS": "3000000"
  }
}

Z.AI applies a default server-side model mapping, so no explicit ANTHROPIC_MODEL is needed. Z.AI 默认在服务端做模型映射，无需显式设置 ANTHROPIC_MODEL。

Skip the Claude Code onboarding / 跳过 Claude Code 初始引导

When using a third-party key (instead of claude login), Claude Code's first-run onboarding won't complete automatically. Create or edit .claude.json and mark it done:

使用第三方 key 时不会走 claude login，Claude Code 首次启动的引导不会自动完成。创建或编辑 .claude.json，手动标记引导已完成：

• macOS / Linux: ~/.claude.json
• Windows: <user-home>\.claude.json

{
  "hasCompletedOnboarding": true
}

Then run claude as usual. / 保存后正常运行 claude 即可。

• ΩmegaWiki — EmpiricalWiki is a downstream specialization of the upstream ΩmegaWiki framework by DAIR Lab at Peking University. The three-layer architecture, the bidirectional wikilink convention, the cross-model review pattern, and 24 of the 29 skills come straight from upstream. EmpiricalWiki adds 5 specialized skills (4 empirical — /empirical-ingest, /variable-map, /empirical-design, /stata-plan; plus 1 theory — /theory-ingest), a theory-modeling layer, and replaces the entity taxonomy. Thanks for building the foundation.
• Andrej Karpathy — for the LLM-Wiki gist that articulated the pattern: raw / wiki / schema, ingest / query / lint, "the wiki is the artifact, not the chat."
• Claude Code — the AI agent runtime that powers EmpiricalWiki.

MIT — use it, fork it, build on it.

2026-06 — 浏览器里直接用 Claude Code、预置内容、Stata 模板库、demo 分支、Agent-native 上手

• 本地站点内嵌 Claude Code 对话面板。 tools/view.sh 现在自带右栏对话页签（设计参考 Claudian）：跑任何 skill、点选回答 Claude 的提问、切模型/思考/权限档位，全程不碰终端。对话中写入的页面自动热同步进站点。仅绑定本机（127.0.0.1），用你自己的 Claude Code 登录。

• 开箱不再是空仓库。 14 张预置 schema 卡随仓库发行：9 张 datasets/ 卡（CSMAR、Wind、CNRDS、华证 ESG、DIB、同花顺 iFinD、WinGo、CCER、专利来源——每张含模块清单、合并键、清洗陷阱、获取渠道）+ 5 张 identification/ 策略卡（交错 DID 与现代估计量、PSM、IV、RDD、双向固定效应——每张写明假设、诊断与审稿人的攻击点）。所有机构事实逐条联网核实并附来源。

• Stata 模板库。 tools/stata-templates/ 五套带 // CHECK: 复核点的 do 骨架（TWFE / 交错 DID / PSM / IV / RDD），与策略卡一一配套，并接入 /stata-plan --write-do。

• 可浏览的 demo。 demo 分支封存完整成品示例：23 篇耐心资本文献编译成变量、机制、识别策略与一张互链图谱——动手之前先看看建成后长什么样。

• 安装步骤不再是给你看的，是给你的 agent 看的。 快速开始现在以一段可直接粘贴给 Claude Code / Codex / Cursor 的指令块开头——agent 替你 clone、配置、自检、汇报。新增 AGENTS.md，让任何 coding agent 都能直接操作本仓库，CLAUDE.md 仍是唯一权威 schema。

2026-05 — 理论层 + 本地网站展示

• 理论建模层。 新增两类页面 —— assumptions（模型原语）与 propositions（命题/定理），以及 /theory-ingest skill，把一篇建模论文消化成它的假设、命题、解概念与可检验推论。理论与实证现在同处一张图谱，通过 mechanisms 与 hypotheses 接桥：一篇理论论文所证明 / 预测的机制和假设，正是实证论文所检验的那一个。（实例：Stein 1988 的管理者短视模型，直接连到检验它的中国 A 股论文。）详见 docs/runtime-theory-skeleton.zh.md。
• 本地网站展示。 tools/view.sh 用 Quartz 把 wiki 渲染成一个本地网站 —— 交互式图谱、全文搜索、反向链接。一条命令、只读、不用装 Obsidian。详见 docs/view-quartz.zh.md。

一条命令 tools/view.sh 即可本地浏览：分节页面、反向链接、全文搜索、交互式图谱。（▶ 更清晰的 MP4）

Andrej Karpathy 提出了 LLM-Wiki 概念：让 LLM 构建并维护一个持久的、结构化的 wiki，而不是一次性的 RAG 回答。每读一篇论文，整个知识图就更厚一点。

EmpiricalWiki 把这个概念掰到一个具体方向上：会计、金融、管理、经济学的实证研究。 通用的 LLM-wiki 设想对一个每天思考"耐心资本怎么测"、"哪个 IV 干净"、"双向 FE 还是公司 FE"、"Sasabuchi 拐点显不显著"的人来说太抽象了。EmpiricalWiki 把通用的 entity / concept / synthesis 三类换成 直接对应实证研究流程的 10 类实体 —— variables、datasets、models、mechanisms、hypotheses、identification、robustness、heterogeneity、tables、papers。

把你的 PDF 丢进文件夹。跑一条命令。得到一个 wiki：每一篇论文的变量构造、模型设定、识别策略、稳健性检验全部被抽出来、互相打通、随时可查。当你下一次读到一篇用了跟之前不一样的方法测某个构念的论文，这个新变体会被加到同一个变量页面里，跟前面看过的并列摆好，而不是丢失在某段聊天记录里。

EmpiricalWiki 严格遵循 Karpathy 原始 gist 的三层结构：

1. Raw — raw/papers/ 放你的 PDF，ingest 永远只读不写。
2. Wiki — wiki/ 是 LLM 维护、不断累积的产物：实证页（变量/模型/识别…）、理论层（assumptions、propositions），以及一张通过 mechanisms/hypotheses 把理论与实证接在一起的图谱。每个实体页都是 markdown + 强制 frontmatter，被 ingest skill 写入和重写。
3. Schema — CLAUDE.md + .claude/skills/ 定义 LLM 怎么读、怎么写、怎么打链。29 个 slash command，全部读自 wiki、写回 wiki。用 tools/view.sh 可把整个 wiki 渲染成带图谱/搜索/反链的本地网站。

前置条件： Python 3.9+；Node.js 22+（Claude Code 与本地网站视图需要）

最快的方式：这些步骤不用你自己做，交给你的 agent。 把下面这段话整体粘贴给 Claude Code（或 Codex / Cursor），让它替你完成配置：

帮我配置 EmpiricalWiki：
1. clone https://github.com/Lambenthan/empiricalwiki.git 并进入目录。
2. 运行 ./setup.sh --lang zh（Windows 用 powershell -ExecutionPolicy Bypass -File .\setup.ps1 -Lang zh）。
3. 检查配置结果：.venv/ 已创建、.env 已生成、.claude/settings.local.json 存在。
4. 完成后告诉我：论文 PDF 该放进哪个目录，以及消化第一篇论文的确切命令。

想先看看建成后长什么样？demo 分支封存了完整的耐心资本示例库——23 篇文献编译成变量、机制、识别策略与一张互链图谱。可直接在 GitHub 上浏览，或本地 git switch demo 后跑 tools/view.sh。

想手动安装？下面是同样的步骤。

git clone https://github.com/Lambenthan/empiricalwiki.git && cd empiricalwiki

# 安装 Claude Code
npm install -g @anthropic-ai/claude-code
claude login

# 一键配置
chmod +x setup.sh && ./setup.sh --lang zh        # Linux / macOS
# Windows (PowerShell):
#   powershell -ExecutionPolicy Bypass -File .\setup.ps1 -Lang zh

# 把你的论文放入 raw/papers/（.pdf 或 .tex）
# 启动 Claude Code
claude
# 然后输入：/empirical-ingest <PDF 路径>  消化一篇论文
# 或者：    /init [研究主题]              批量初始化

Windows 用户：本地 pipeline 已原生支持。/exp-run --env remote 远程 GPU 实验依赖 ssh/rsync/screen，建议在 WSL2 或 Linux/macOS 下运行（经管实证一般用不上）。

详细模板见 docs/runtime-page-templates.zh.md。简要对应：

通用类型 (concepts/ / topics/ / claims/ / ideas/ / experiments/ / people/ / Summary/ / foundations/) 仍保留给非实证内容（理论论文、方法论综述、自有研究 idea）。

对一个实证研究项目而言，EmpiricalWiki 产出的是：

• 一篇论文一张实证卡片 —— 研究问题、假设、样本、变量、模型、识别、稳健性、异质性都被抽取出来
• 跨论文累积测算变体的变量页 —— 第二篇用了跟第一篇不一样的操作化方式，新变体直接落在同一页，并列展示，而不是丢失在某段聊天记录里
• 可复用的识别策略 / 稳健性 / 异质性库 —— 跨论文重复出现的设计模式被抽成独立的可检索实体
• 双向链接的图谱：papers ↔ variables ↔ datasets ↔ models ↔ mechanisms，物化在 wiki/graph/edges.jsonl

下一次再读用了某个已知构念的论文时，不用回忆之前哪篇用了哪种操作化 —— 变量页里所有见过的变体早就并列摆好了，每条挂着 source paper。

• ΩmegaWiki —— EmpiricalWiki 是 ΩmegaWiki 在经管实证方向上的下游特化版本，三层架构、双向链接、跨模型评审、29 个 skill 中的 24 个全部来自上游 DAIR Lab（北京大学）。EmpiricalWiki 在此基础上加了 5 个特化 skill（4 个经管 + 1 个理论 /theory-ingest）并替换了实体分类、新增理论建模层。感谢上游团队搭好的地基。
• Andrej Karpathy —— 提出 LLM-Wiki gist，明确了 raw / wiki / schema 三层架构与 ingest / query / lint 三个操作。
• Claude Code —— 驱动 EmpiricalWiki 的 AI agent runtime。