llm-wiki

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

LLM Wiki — Karpathy Knowledge Base Pattern

LLM Wiki — Karpathy知识库模式

Experimental skill — iterating. Authored by Lewis Liu (lylewis@outlook.com) · Inspired by Karpathy's llm-wiki Gist

实验性功能——持续迭代中。 作者：Lewis Liu (lylewis@outlook.com) · 灵感来自Karpathy的llm-wiki Gist

Core idea

核心理念

Instead of RAG (re-retrieving raw docs on every query), the LLM compiles raw sources into a persistent, cross-linked wiki. Every ingest, query, lint, and audit pass makes the wiki richer. Knowledge compounds — and the human stays in the loop via a structured feedback channel instead of ad-hoc corrections that get lost.

You own: sourcing raw material, asking good questions, steering direction, filing feedback on anything the AI got wrong.
LLM owns: all writing, cross-referencing, filing, bookkeeping, and acting on your feedback.

The wiki is a living artifact with five operations —

compile

ingest

query

lint

audit

. Every session starts by reading

CLAUDE.md

and

wiki/index.md

不同于RAG（每次查询时重新检索原始文档），LLM会将原始源编译为持久化、带交叉链接的维基。每次导入、查询、检查和审计操作都会让维基内容更丰富。知识会不断积累——人类通过结构化反馈渠道参与流程，避免临时修正丢失。

你负责：获取原始材料、提出优质问题、把控方向、针对AI的错误提交反馈。
LLM负责：所有内容撰写、交叉引用、归档、记录，以及根据你的反馈执行修正。

维基是一个具备五大操作的动态产物——

compile

（编译）、

ingest

（导入）、

query

（查询）、

lint

（检查）、

audit

（审计）。每次会话开始时需读取

CLAUDE.md

和

wiki/index.md

。

Directory layout

目录结构

<wiki-root>/
├── CLAUDE.md          ← Schema: scope, conventions, current articles, gaps
├── log/               ← Per-day operation log (one file per day)
│   ├── 20260409.md
│   └── 20260410.md
├── audit/             ← Human feedback inbox (one file per comment)
│   ├── 20260409-143022-claude-code-size.md
│   └── resolved/      ← Processed feedback, archived with resolution notes
├── raw/               ← Immutable source documents (LLM reads, never writes)
│   ├── articles/
│   ├── papers/
│   ├── notes/
│   └── refs/          ← Pointer files for large binaries kept outside raw/
├── wiki/              ← LLM-generated knowledge (LLM writes, you read)
│   ├── index.md       ← Master catalog — every page, structured by category
│   ├── concepts/      ← Concept/topic pages (split into subfolders when >1200 words)
│   ├── entities/      ← People, tools, papers, organizations
│   └── summaries/     ← Per-source summary pages
└── outputs/
    └── queries/       ← Query answers (promote durable ones to wiki/)

CLAUDE.md

is the schema file — the single most important configuration. It tells the LLM the wiki's scope, naming conventions, current article list, open questions, and research gaps. Read

references/schema-guide.md

for what to put in it. Read it at the start of every session.

<wiki-root>/
├── CLAUDE.md          ← 配置文件：范围、规范、现有文章、内容缺口
├── log/               ← 每日操作日志（每天一个文件）
│   ├── 20260409.md
│   └── 20260410.md
├── audit/             ← 人工反馈收件箱（每条评论对应一个文件）
│   ├── 20260409-143022-claude-code-size.md
│   └── resolved/      ← 已处理的反馈，附带解决说明归档
├── raw/               ← 不可变的源文档（LLM仅读取，不写入）
│   ├── articles/
│   ├── papers/
│   ├── notes/
│   └── refs/          ← 指向raw/目录外大型二进制文件的指针文件
├── wiki/              ← LLM生成的知识内容（LLM写入，你读取）
│   ├── index.md       ← 主目录——所有页面，按类别结构化展示
│   ├── concepts/      ← 概念/主题页面（超过1200字时拆分到子文件夹）
│   ├── entities/      ← 人物、工具、论文、组织
│   └── summaries/     ← 每个源文档的摘要页面
└── outputs/
    └── queries/       ← 查询答案（将持久化内容升级到wiki/目录）

CLAUDE.md

是核心配置文件——最重要的配置项。它告知LLM维基的范围、命名规范、现有文章列表、待解决问题和研究缺口。请阅读

references/schema-guide.md

了解配置内容，每次会话开始时都需阅读此文件。

Core principles

核心原则

Four rules govern everything below. If a future instruction contradicts one, flag it to the user before acting.

以下四条规则指导所有操作。若后续指令与其中任何一条冲突，需先向用户标记再执行。

1. Divide and conquer

1. 分而治之

A single concept page should never try to cover a complex topic end-to-end. Target: 400–1200 words per page. When a topic would blow past that:

Create a subfolder:
```
wiki/concepts/<topic>/
```
Put a short index page at
```
wiki/concepts/<topic>/index.md
```
— definition, list of sub-pages, one-line summaries
Put each aspect in its own file:
```
wiki/concepts/<topic>/<aspect>.md
```
In
```
wiki/index.md
```
, show the hierarchy via indented bullets

Example layout (from a real wiki):

wiki/tech/claude-code/
├── index.md                         (overview + links to sub-pages)
├── Claude_Code_Architecture.md
├── Claude_Code_Agent_Framework.md
├── Claude_Code_Bridge_System.md
├── Claude_Code_Query_Engine.md
├── Claude_Code_Skills_Plugins.md
├── Claude_Code_State_Management.md
└── Claude_Code_Tool_System.md

One fat file covering all seven aspects would be unreadable and unlinkable. Seven focused files + an index page give you navigation, selective reading, clean backlinks, and small audit targets.

单个概念页面绝不应试图全面覆盖复杂主题。目标：每页400–1200字。当主题内容超出此范围时：

创建子文件夹：
```
wiki/concepts/<topic>/
```
在
```
wiki/concepts/<topic>/index.md
```
创建简短索引页——定义、子页面列表、单行摘要
将每个方面放在单独文件中：
```
wiki/concepts/<topic>/<aspect>.md
```
在
```
wiki/index.md
```
中通过缩进项目符号展示层级

示例结构（来自真实维基）：

wiki/tech/claude-code/
├── index.md                         (概述 + 子页面链接)
├── Claude_Code_Architecture.md
├── Claude_Code_Agent_Framework.md
├── Claude_Code_Bridge_System.md
├── Claude_Code_Query_Engine.md
├── Claude_Code_Skills_Plugins.md
├── Claude_Code_State_Management.md
└── Claude_Code_Tool_System.md

一个包含所有七个方面的臃肿文件会难以阅读和链接。七个聚焦的文件加一个索引页，能提供导航、选择性阅读、清晰的反向链接和更小的审计目标。

2. Mermaid for diagrams, KaTeX for formulas

2. 用Mermaid绘制图表，用KaTeX编写公式

Any flow, sequence, hierarchy, or state diagram must be written in mermaid — never ASCII art. ASCII boxes rot fast and are impossible to annotate.
```
```mermaid
flowchart LR
    A[raw/article.md] --> B[summary]
    B --> C[concept page]
    C --> D[index.md]
```
```
Any formula must be written in KaTeX: inline
```
$f(x) = \sum_i w_i x_i$
```
or block
```
$$...$$
```
.

Both render in the web viewer (server-side KaTeX, client-side mermaid) and in Obsidian with default settings.

任何流程图、序列图、层级图或状态图必须用mermaid编写——绝不要用ASCII艺术。ASCII框会很快过时且无法注释。
```
```mermaid
flowchart LR
    A[raw/article.md] --> B[summary]
    B --> C[concept page]
    C --> D[index.md]
```
```
任何公式必须用KaTeX编写：行内式
```
$f(x) = \sum_i w_i x_i$
```
或块级式
```
$$...$$
```
。

两种格式都可在网页查看器（服务端渲染KaTeX，客户端渲染mermaid）和默认设置的Obsidian中显示。

3. Raw file policy

3. 原始文件规则

Small text-based sources (md, txt, small pdfs, small images) → copy into

raw/<subfolder>/

Large binaries (videos, model weights, installers, datasets, large PDFs >10 MB) → do not copy. Instead:

Create a pointer file at
```
raw/refs/<slug>.md
```
with:
yaml
```
---
kind: ref
external_path: /Volumes/external/models/llama-3-70b/
size: ~140 GB
---
```
followed by a short description of what it is and why it matters to this wiki.
Wiki pages cite
```
[[raw/refs/<slug>]]
```
exactly like any other source.

This keeps the wiki repo git-friendly and portable.

小型文本源（md、txt、小型PDF、小型图片）→ 复制到

raw/<subfolder>/

目录。

大型二进制文件（视频、模型权重、安装包、数据集、大于10MB的大型PDF）→ 不要复制。而是：

在
```
raw/refs/<slug>.md
```
创建指针文件，内容如下：
yaml
```
---
kind: ref
external_path: /Volumes/external/models/llama-3-70b/
size: ~140 GB
---
```
后续添加简短描述，说明文件内容及其对维基的重要性。
维基页面引用
```
[[raw/refs/<slug>]]
```
，与其他源文件格式完全一致。

这样可保持维基仓库适合Git管理且便于移植。

4. Audit is the human feedback surface

4. 审计是人工反馈渠道

The wiki is AI-written; it will be wrong sometimes. The raw sources are human-written; they will contradict each other. The

audit/

directory is how humans correct both without losing the corrections in chat history.

Humans file feedback via the Obsidian plugin or the web viewer. Each feedback is one file in
```
audit/
```
with YAML frontmatter (anchor, target, severity) and a markdown body.
The AI must periodically run the
```
audit
```
op — never silently ignore
```
audit/*.md
```
files.
When feedback is applied, the file moves to
```
audit/resolved/
```
with a
```
# Resolution
```
section appended and a log entry recorded in
```
log/YYYYMMDD.md
```
.

See

references/audit-guide.md

for the full file format and processing workflow.

维基由AI撰写，有时会出错。原始源由人类撰写，可能存在矛盾。

audit/

目录是人类修正这些问题的方式，避免修正内容丢失在聊天记录中。

人类通过Obsidian插件或网页查看器提交反馈。每条反馈对应
```
audit/
```
目录中的一个文件，包含YAML前置信息（锚点、目标、严重程度）和Markdown正文。
AI必须定期执行
```
audit
```
操作——绝不能忽略
```
audit/*.md
```
文件。
应用反馈后，将文件移至
```
audit/resolved/
```
，追加
```
# Resolution
```
部分，并在
```
log/YYYYMMDD.md
```
中记录日志条目。

完整文件格式和处理流程请参阅

references/audit-guide.md

。

The five operations

五大操作

Every action on the wiki is one of these five. Each appends an entry to the current day's log file (

log/YYYYMMDD.md

对维基的所有操作均属于以下五类。每项操作都会追加条目到当日日志文件（

log/YYYYMMDD.md

）。

compile

compile

（编译）

(Re)structure wiki content from existing

raw/

material — including splitting oversized pages, merging near-duplicates, and rebuilding

index.md

When to run: after a big ingest batch, when an existing page has outgrown 1200 words, when

index.md

no longer reflects reality, or when the user says "clean up the wiki".

Steps:

Read
```
CLAUDE.md
```
,
```
wiki/index.md
```
, and every file in the target subtree.
For each page over ~1200 words: plan a split into
```
concepts/<topic>/
```
with an index + sub-pages. Confirm the plan with the user before writing.
For each pair of near-duplicate pages: propose a merge. Confirm, then rewrite.
Regenerate
```
wiki/index.md
```
so every page is listed exactly once.

Log:

## [HH:MM] compile | <what you did — files touched, splits, merges>

从现有

raw/

材料（重新）构建维基内容——包括拆分过长页面、合并近似重复内容、重建

index.md

。

执行时机：批量导入后、现有页面超过1200字、

index.md

与实际内容不符，或用户要求"清理维基"时。

步骤：

读取
```
CLAUDE.md
```
、
```
wiki/index.md
```
和目标子树中的所有文件。
对每个约1200字以上的页面：规划拆分到
```
concepts/<topic>/
```
目录，包含索引页和子页面。写入前需与用户确认方案。
对每对近似重复页面：提出合并建议。确认后重写内容。
重新生成
```
wiki/index.md
```
，确保每个页面仅列出一次。

日志：

## [HH:MM] compile | <操作内容——涉及文件、拆分、合并>

ingest

ingest

（导入）

Add a new source. One source typically touches 5–15 wiki pages.

Steps:

Save source to the right subfolder:
- web article →
```
raw/articles/<slug>.md
```
- paper →
```
raw/papers/<slug>.md
```
  (extracted text for big PDFs)
- note →
```
raw/notes/<slug>.md
```
- large binary →
```
raw/refs/<slug>.md
```
  pointer file (see raw file policy)
Read the source in full.
Create
```
wiki/summaries/<slug>.md
```
(200–400 words — key takeaways, not a rewrite; see
```
references/article-guide.md
```
).
Create or update relevant concept pages in
```
wiki/concepts/
```
. Respect divide-and-conquer: if a concept page would exceed 1200 words, split instead of cramming.
Create or update entity pages in
```
wiki/entities/
```
for any new people / tools / papers / organizations referenced.
Update
```
wiki/index.md
```
so the new pages appear under the right category.

Log:

## [HH:MM] ingest | <slug> — <one-line description> (touched N pages)

添加新源文件。一个源文件通常会涉及5–15个维基页面。

步骤：

将源文件保存到对应子文件夹：
- 网页文章 →
```
raw/articles/<slug>.md
```
- 论文 →
```
raw/papers/<slug>.md
```
  （大型PDF需提取文本）
- 笔记 →
```
raw/notes/<slug>.md
```
- 大型二进制文件 →
```
raw/refs/<slug>.md
```
  指针文件（见原始文件规则）
完整读取源文件。
创建
```
wiki/summaries/<slug>.md
```
（200–400字——核心要点，而非重写；请参阅
```
references/article-guide.md
```
）。
创建或更新
```
wiki/concepts/
```
中的相关概念页面。遵循分而治之原则：若概念页面会超过1200字，应拆分而非强行塞入。
创建或更新
```
wiki/entities/
```
中的实体页面，涵盖所有新提及的人物/工具/论文/组织。
更新
```
wiki/index.md
```
，使新页面出现在对应类别下。

日志：

## [HH:MM] ingest | <slug> — <单行描述>（涉及N个页面）

query

query

（查询）

Answer a question grounded in the wiki, not general knowledge.

Steps:

Read
```
wiki/index.md
```
. Scan for relevant pages by category.
Read the identified pages in full; follow one level of wikilinks.
If the wiki doesn't have enough material, say so and suggest what to ingest next instead of making something up.
Synthesize the answer, citing pages inline with
```
[[Page Name]]
```
.

Save to

outputs/queries/<YYYY-MM-DD>-<question-slug>.md

If the answer is durable (a comparison, analysis, or new synthesis) → promote a cleaned-up version to
```
wiki/concepts/
```
, add to
```
index.md
```
.

Log:

## [HH:MM] query | <question-slug>

(and a separate

## [HH:MM] promote | ...

line if promoted).

基于维基内容回答问题，而非通用知识。

步骤：

读取
```
wiki/index.md
```
，按类别扫描相关页面。
完整读取识别出的页面；追踪一层维基链接。
若维基内容不足，需告知用户并建议下一步应导入的内容，而非编造答案。
合成答案，用
```
[[页面名称]]
```
在行内引用页面。

保存到

outputs/queries/<YYYY-MM-DD>-<question-slug>.md

。

若答案具备持久价值（对比、分析或新合成内容）→ 将整理后的版本升级到
```
wiki/concepts/
```
，并添加到
```
index.md
```
。

日志：

## [HH:MM] query | <question-slug>

（若升级则需单独添加

## [HH:MM] promote | ...

条目）。

lint

lint

（检查）

Health check. Run:

bash

python3 scripts/lint_wiki.py <wiki-root>

The script reports:

Dead wikilinks —
```
[[Target]]
```
where
```
Target.md
```
doesn't exist
Orphan pages — pages with no inbound wikilinks
Missing index entries — pages not listed in
```
wiki/index.md
```
Frequently-linked missing pages —
```
[[X]]
```
referenced 3+ times but no page
log/ shape — stray files or wrong filenames in
```
log/
```
audit/ shape — malformed YAML frontmatter in
```
audit/*.md
```
Audit target resolution — every open audit's
```
target
```
file must exist

For each issue, propose a fix, confirm with the user, then apply. Log:

## [HH:MM] lint | <N> issues found, <M> fixed

健康检查。执行：

bash

python3 scripts/lint_wiki.py <wiki-root>

脚本会报告：

无效维基链接 —
```
[[Target]]
```
对应的
```
Target.md
```
不存在
孤立页面 — 无入站维基链接的页面
缺失索引条目 — 未在
```
wiki/index.md
```
中列出的页面
频繁引用的缺失页面 —
```
[[X]]
```
被引用3次以上但无对应页面
log/目录格式 — log/目录中的零散文件或错误文件名
audit/目录格式 — audit/*.md文件中的YAML前置信息格式错误
审计目标有效性 — 所有未处理审计的
```
target
```
文件必须存在

针对每个问题，提出修复方案，与用户确认后执行。日志：

## [HH:MM] lint | <N>个问题被发现，<M>个已修复

。

audit

audit

（审计）

Process human feedback from

audit/

Steps:

Run

python3 scripts/audit_review.py <wiki-root> --open

to get a grouped list.

For each open audit, read the file. Use the
```
anchor_before
```
/
```
anchor_text
```
/
```
anchor_after
```
window to locate the exact range in the target file (line numbers may have drifted).
Decide the action:
- Accept: apply the correction to the target file.
- Partially accept: apply what makes sense, note the rest in the resolution.
- Reject: explain why in the resolution — the feedback may be based on a misreading of scope or a contradictory source.
- Defer: add to
```
CLAUDE.md
```
  "Open research questions" and leave the audit in place with a comment.

For applied audits, append a

# Resolution

section to the audit file:

markdown

# Resolution

2026-04-10 · accepted.
Fixed the file count (was "~1,900", corrected to "~1,800" per commit abc123).
Updated: tech/Claude_Code.md lines 47–48.

Move the file from
```
audit/
```
to
```
audit/resolved/
```
. Filename unchanged.

Log per resolved audit:

## [HH:MM] audit | resolved 20260409-143022-a1b2 — <one-line what>

Never delete audit files. Rejected ones still go to
```
resolved/
```
with the rejection rationale in their resolution section — that's valuable history.

See

references/audit-guide.md

for the full audit file format.

处理

audit/

目录中的人工反馈。

步骤：

执行

python3 scripts/audit_review.py <wiki-root> --open

获取分组列表。

对每个未处理的审计，读取文件。使用
```
anchor_before
```
/
```
anchor_text
```
/
```
anchor_after
```
定位目标文件中的精确范围（行号可能已变动）。
决定操作：
- 接受：将修正应用到目标文件。
- 部分接受：应用合理部分，在解决说明中记录剩余内容。
- 拒绝：在解决说明中解释原因——反馈可能基于对范围的误读或矛盾的源文件。
- 延期：添加到
```
CLAUDE.md
```
  的"待解决研究问题"部分，保留审计文件并添加注释。

对已应用的审计，在审计文件中追加

# Resolution

部分：

markdown

# Resolution

2026-04-10 · 已接受。
修正文件数量（原为"~1,900"，根据提交abc123改为"~1,800"）。
更新：tech/Claude_Code.md第47–48行。

将文件从
```
audit/
```
移至
```
audit/resolved/
```
，文件名不变。

为每个已解决的审计记录日志：

## [HH:MM] audit | resolved 20260409-143022-a1b2 — <单行描述>

绝不要删除审计文件。被拒绝的文件仍需移至
```
resolved/
```
，并在解决说明中记录拒绝理由——这是有价值的历史记录。

完整审计文件格式请参阅

references/audit-guide.md

。

Tooling

工具集

Tool	Purpose
Obsidian	IDE for browsing the wiki; graph view shows connections
`plugins/obsidian-audit/`	Obsidian plugin — select text → add feedback → writes to `audit/`
`web/`	Local Node.js server — preview the wiki with mermaid/math rendered; select → feedback → `audit/`
`scripts/scaffold.py`	Bootstrap a new wiki directory tree
`scripts/lint_wiki.py`	Seven-pass health check
`scripts/audit_review.py`	Group open/resolved audits by target file
qmd	Optional local semantic search (useful at >100 pages)

The Obsidian plugin and the web viewer both write audit files in the same format with the same anchor algorithm, so feedback filed from either place can be resolved by either place.

工具	用途
Obsidian	维基浏览IDE；图谱视图展示关联关系
`plugins/obsidian-audit/`	Obsidian插件——选中文本→添加反馈→写入 `audit/` 目录
`web/`	本地Node.js服务器——预览维基，渲染mermaid/数学公式；选中文本→提交反馈→写入 `audit/` 目录
`scripts/scaffold.py`	快速搭建新维基目录结构
`scripts/lint_wiki.py`	七轮健康检查
`scripts/audit_review.py`	按目标文件分组展示未处理/已处理审计
qmd	可选本地语义搜索（页面超过100个时实用）

Obsidian插件和网页查看器均采用相同格式和相同锚点算法写入审计文件，因此从任一渠道提交的反馈都可在任一渠道处理。

Starting a new wiki

创建新维基

bash

python3 scripts/scaffold.py <wiki-root> "<Topic Title>"

Creates the full tree (including

log/<today>.md

audit/

audit/resolved/

), a blank

CLAUDE.md

based on the new template, and a blank

wiki/index.md

with the recommended category layout.

After scaffolding:

Fill in
```
CLAUDE.md
```
— define scope, naming conventions, initial research questions.
Start ingesting sources.
Ask questions to build up
```
outputs/queries/
```
; promote durable answers.
Run
```
lint
```
periodically.
Run
```
audit
```
whenever new feedback accumulates.

bash

python3 scripts/scaffold.py <wiki-root> "<主题标题>"

创建完整目录结构（包括

log/<today>.md

、

audit/

、

audit/resolved/

），基于新模板生成空白

CLAUDE.md

，以及带推荐类别结构的空白

wiki/index.md

。

搭建完成后：

填写
```
CLAUDE.md
```
——定义范围、命名规范、初始研究问题。
开始导入源文件。
提出问题以构建
```
outputs/queries/
```
内容；将持久化答案升级到维基。
定期执行
```
lint
```
检查。
有新反馈积累时执行
```
audit
```
操作。

wiki/index.md

format

wiki/index.md

格式

The LLM rebuilds

index.md

on every compile and touches it on every ingest. Format:

markdown

undefined

LLM会在每次编译时重建

index.md

，每次导入时更新此文件。格式：

markdown

undefined

Index — <Topic>

Index — <主题>

One-sentence scope of the wiki.

维基范围的单行描述。

🔖 Navigation

🔖 导航

[[#Concepts]] · [[#Entities]] · [[#Summaries]] · [[#Open Questions]]

[[#Concepts]] · [[#Entities]] · [[#Summaries]] · [[#Open Questions]]

Concepts

<Category A>

<类别A>

[[concepts/Foo]] — one-line summary
[[concepts/Bar/index|Bar]] — (folder-split) one-line summary
- [[concepts/Bar/aspect-1]] — ...
- [[concepts/Bar/aspect-2]] — ...

[[concepts/Foo]] — 单行摘要
[[concepts/Bar/index|Bar]] — （文件夹拆分）单行摘要
- [[concepts/Bar/aspect-1]] — ...
- [[concepts/Bar/aspect-2]] — ...

<Category B>

<类别B>

Entities

[[entities/Andrej Karpathy]] — AI researcher, author of the llm-wiki pattern

[[entities/Andrej Karpathy]] — AI研究员，llm-wiki模式提出者

Summaries (chronological)

Summaries（按时间排序）

2026-04-09 — [[summaries/llm-wiki-gist]] — Karpathy's original Gist

2026-04-09 — [[summaries/llm-wiki-gist]] — Karpathy的原始Gist

Open Questions

Q1: ...


Rules:
- Every wiki page must appear exactly once in `index.md`. `lint` enforces this.
- Folder-split concepts show hierarchy via indented bullets.
- `index.md` + `CLAUDE.md` together are what the AI reads at session start.

Q1: ...


规则：
- 每个维基页面必须在`index.md`中精确出现一次。`lint`检查会强制执行此规则。
- 拆分到文件夹的概念通过缩进项目符号展示层级。
- `index.md` + `CLAUDE.md`是AI会话开始时读取的内容。

log/

format

log/

目录格式

See

references/log-guide.md

for full details. Minimum:

One file per day:
```
log/YYYYMMDD.md
```

H1 = the date; H2 per entry with

## [HH:MM] <op> | <one-line description>

Ops:

compile

ingest

query

lint

audit

promote

split

scaffold

Quick grep across history:

grep -rh "^## \[" log/ | tail -20

完整细节请参阅

references/log-guide.md

。最低要求：

每天一个文件：
```
log/YYYYMMDD.md
```
一级标题为日期；每个条目为二级标题，格式为
```
## [HH:MM] <操作> | <单行描述>
```

操作类型：

compile

ingest

query

lint

audit

promote

split

scaffold

快速检索历史日志：

grep -rh "^## \[" log/ | tail -20

。

Use cases

适用场景

Research deep-dive — reading papers/articles on a topic over weeks; the wiki evolves with your understanding, and the audit trail keeps AI mistakes from silently accumulating
Personal wiki — journal entries, notes, ideas compiled into a personal encyclopedia; comment on anything you disagree with later, the AI corrects it
Team knowledge base — fed by Slack threads, meeting notes, docs; team members file corrections through the web viewer
Reading companion — filing each book chapter as you go; builds a rich companion wiki by the end

深度研究——数周内阅读某主题的论文/文章；维基随你的理解不断演进，审计记录可防止AI错误悄悄积累
个人维基——将日记、笔记、想法编译成个人百科全书；后续可对任何不同意的内容提交评论，AI会进行修正
团队知识库——整合Slack线程、会议记录、文档；团队成员可通过网页查看器提交修正
阅读助手——阅读时归档每章内容；读完后构建丰富的配套维基

References

参考文档

```
references/schema-guide.md
```
— What to put in
```
CLAUDE.md
```
```
references/article-guide.md
```
— How to write good wiki articles (length, wikilinks, mermaid, math, divide-and-conquer)
```
references/log-guide.md
```
— The
```
log/
```
folder convention
```
references/audit-guide.md
```
— Audit file format, anchor strategy, processing workflow
```
references/tooling-tips.md
```
— Obsidian setup, Web Clipper, qmd, plugin + web installation

```
references/schema-guide.md
```
—
```
CLAUDE.md
```
的配置指南
```
references/article-guide.md
```
— 优质维基文章撰写指南（篇幅、维基链接、mermaid、公式、分而治之）
```
references/log-guide.md
```
—
```
log/
```
目录规范
```
references/audit-guide.md
```
— 审计文件格式、锚点策略、处理流程
```
references/tooling-tips.md
```
— Obsidian设置、网页剪辑器、qmd、插件及网页端安装指南

llm-wiki

Original

Translation

LLM Wiki — Karpathy Knowledge Base Pattern

LLM Wiki — Karpathy知识库模式

Core idea

核心理念

Directory layout

目录结构

Core principles

核心原则

1. Divide and conquer

1. 分而治之

2. Mermaid for diagrams, KaTeX for formulas

2. 用Mermaid绘制图表，用KaTeX编写公式

3. Raw file policy

3. 原始文件规则

4. Audit is the human feedback surface

4. 审计是人工反馈渠道

The five operations

五大操作

1. compile

1. compile（编译）

2. ingest

2. ingest（导入）

3. query

3. query（查询）

4. lint

4. lint（检查）

5. audit

5. audit（审计）

Tooling

工具集

Starting a new wiki

创建新维基

wiki/index.md format

wiki/index.md格式

Index — <Topic>

Index — <主题>

🔖 Navigation

🔖 导航

Concepts

Concepts

<Category A>

<类别A>

<Category B>

<类别B>

Entities

Entities

Summaries (chronological)

Summaries（按时间排序）

Open Questions

Open Questions

log/ format

log/目录格式

Use cases

适用场景

References

参考文档

1.
`compile`

1.
`compile`
（编译）

2.
`ingest`

2.
`ingest`
（导入）

3.
`query`

3.
`query`
（查询）

4.
`lint`

4.
`lint`
（检查）

5.
`audit`

5.
`audit`
（审计）

`wiki/index.md`
format

`wiki/index.md`
格式

`log/`
format

`log/`
目录格式