Jun 9, 2026, 3:59 PM
This commit is contained in:
@@ -1,4 +1,4 @@
|
||||
# LLM Wiki
|
||||
# LLM Knowledge Base
|
||||
|
||||
A pattern for building personal knowledge bases using LLMs.
|
||||
|
||||
@@ -8,18 +8,18 @@ This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.
|
||||
|
||||
Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.
|
||||
|
||||
The idea here is different. Instead of just retrieving from raw documents at query time, the LLM **incrementally builds and maintains a persistent wiki** — a structured, interlinked collection of markdown files that sits between you and the raw sources. When you add a new source, the LLM doesn't just index it for later retrieval. It reads it, extracts the key information, and integrates it into the existing wiki — updating entity pages, revising topic summaries, noting where new data contradicts old claims, strengthening or challenging the evolving synthesis. The knowledge is compiled once and then *kept current*, not re-derived on every query.
|
||||
The idea here is different. Instead of just retrieving from raw documents at query time, the LLM **incrementally builds and maintains a persistent knowledge** — a structured, interlinked collection of markdown files that sits between you and the raw sources. When you add a new source, the LLM doesn't just index it for later retrieval. It reads it, extracts the key information, and integrates it into the existing knowledge — updating entity pages, revising topic summaries, noting where new data contradicts old claims, strengthening or challenging the evolving synthesis. The knowledge is compiled once and then *kept current*, not re-derived on every query.
|
||||
|
||||
This is the key difference: **the wiki is a persistent, compounding artifact.** The cross-references are already there. The contradictions have already been flagged. The synthesis already reflects everything you've read. The wiki keeps getting richer with every source you add and every question you ask.
|
||||
This is the key difference: **the knowledge is a persistent, compounding artifact.** The cross-references are already there. The contradictions have already been flagged. The synthesis already reflects everything you've read. The knowledge keeps getting richer with every source you add and every question you ask.
|
||||
|
||||
You never (or rarely) write the wiki yourself — the LLM writes and maintains all of it. You're in charge of sourcing, exploration, and asking the right questions. The LLM does all the grunt work — the summarizing, cross-referencing, filing, and bookkeeping that makes a knowledge base actually useful over time. In practice, I have the LLM agent open on one side and Obsidian open on the other. The LLM makes edits based on our conversation, and I browse the results in real time — following links, checking the graph view, reading the updated pages. Obsidian is the IDE; the LLM is the programmer; the wiki is the codebase.
|
||||
You never (or rarely) write the knowledge yourself — the LLM writes and maintains all of it. You're in charge of sourcing, exploration, and asking the right questions. The LLM does all the grunt work — the summarizing, cross-referencing, filing, and bookkeeping that makes a knowledge base actually useful over time. In practice, I have the LLM agent open on one side and Obsidian open on the other. The LLM makes edits based on our conversation, and I browse the results in real time — following links, checking the graph view, reading the updated pages. Obsidian is the IDE; the LLM is the programmer; the knowledge is the codebase.
|
||||
|
||||
This can apply to a lot of different contexts. A few examples:
|
||||
|
||||
- **Personal**: tracking your own goals, health, psychology, self-improvement — filing journal entries, articles, podcast notes, and building up a structured picture of yourself over time.
|
||||
- **Research**: going deep on a topic over weeks or months — reading papers, articles, reports, and incrementally building a comprehensive wiki with an evolving thesis.
|
||||
- **Reading a book**: filing each chapter as you go, building out pages for characters, themes, plot threads, and how they connect. By the end you have a rich companion wiki. Think of fan wikis like [Tolkien Gateway](https://tolkiengateway.net/wiki/Main_Page) — thousands of interlinked pages covering characters, places, events, languages, built by a community of volunteers over years. You could build something like that personally as you read, with the LLM doing all the cross-referencing and maintenance.
|
||||
- **Business/team**: an internal wiki maintained by LLMs, fed by Slack threads, meeting transcripts, project documents, customer calls. Possibly with humans in the loop reviewing updates. The wiki stays current because the LLM does the maintenance that no one on the team wants to do.
|
||||
- **Research**: going deep on a topic over weeks or months — reading papers, articles, reports, and incrementally building a comprehensive knowledge with an evolving thesis.
|
||||
- **Reading a book**: filing each chapter as you go, building out pages for characters, themes, plot threads, and how they connect. By the end you have a rich companion knowledge. Think of fan knowledges like [Tolkien Gateway](https://tolkiengateway.net/knowledge/Main_Page) — thousands of interlinked pages covering characters, places, events, languages, built by a community of volunteers over years. You could build something like that personally as you read, with the LLM doing all the cross-referencing and maintenance.
|
||||
- **Business/team**: an internal knowledge maintained by LLMs, fed by Slack threads, meeting transcripts, project documents, customer calls. Possibly with humans in the loop reviewing updates. The knowledge stays current because the LLM does the maintenance that no one on the team wants to do.
|
||||
- **Competitive analysis, due diligence, trip planning, course notes, hobby deep-dives** — anything where you're accumulating knowledge over time and want it organized rather than scattered.
|
||||
|
||||
## Architecture
|
||||
@@ -28,42 +28,42 @@ There are three layers:
|
||||
|
||||
**Raw sources** — your curated collection of source documents. Articles, papers, images, data files. These are immutable — the LLM reads from them but never modifies them. This is your source of truth.
|
||||
|
||||
**The wiki** — a directory of LLM-generated markdown files. Summaries, entity pages, concept pages, comparisons, an overview, a synthesis. The LLM owns this layer entirely. It creates pages, updates them when new sources arrive, maintains cross-references, and keeps everything consistent. You read it; the LLM writes it.
|
||||
**The knowledge** — a directory of LLM-generated markdown files. Summaries, entity pages, concept pages, comparisons, an overview, a synthesis. The LLM owns this layer entirely. It creates pages, updates them when new sources arrive, maintains cross-references, and keeps everything consistent. You read it; the LLM writes it.
|
||||
|
||||
**The schema** — a document (e.g. CLAUDE.md for Claude Code or AGENTS.md for Codex) that tells the LLM how the wiki is structured, what the conventions are, and what workflows to follow when ingesting sources, answering questions, or maintaining the wiki. This is the key configuration file — it's what makes the LLM a disciplined wiki maintainer rather than a generic chatbot. You and the LLM co-evolve this over time as you figure out what works for your domain.
|
||||
**The schema** — a document (e.g. CLAUDE.md for Claude Code or AGENTS.md for Codex) that tells the LLM how the knowledge is structured, what the conventions are, and what workflows to follow when ingesting sources, answering questions, or maintaining the knowledge. This is the key configuration file — it's what makes the LLM a disciplined knowledge maintainer rather than a generic chatbot. You and the LLM co-evolve this over time as you figure out what works for your domain.
|
||||
|
||||
## Operations
|
||||
|
||||
**Ingest.** You drop a new source into the raw collection and tell the LLM to process it. An example flow: the LLM reads the source, discusses key takeaways with you, writes a summary page in the wiki, updates the index, updates relevant entity and concept pages across the wiki, and appends an entry to the log. A single source might touch 10-15 wiki pages. Personally I prefer to ingest sources one at a time and stay involved — I read the summaries, check the updates, and guide the LLM on what to emphasize. But you could also batch-ingest many sources at once with less supervision. It's up to you to develop the workflow that fits your style and document it in the schema for future sessions.
|
||||
**Ingest.** You drop a new source into the raw collection and tell the LLM to process it. An example flow: the LLM reads the source, discusses key takeaways with you, writes a summary page in the knowledge, updates the index, updates relevant entity and concept pages across the knowledge, and appends an entry to the log. A single source might touch 10-15 knowledge pages. Personally I prefer to ingest sources one at a time and stay involved — I read the summaries, check the updates, and guide the LLM on what to emphasize. But you could also batch-ingest many sources at once with less supervision. It's up to you to develop the workflow that fits your style and document it in the schema for future sessions.
|
||||
|
||||
**Query.** You ask questions against the wiki. The LLM searches for relevant pages, reads them, and synthesizes an answer with citations. Answers can take different forms depending on the question — a markdown page, a comparison table, a slide deck (Marp), a chart (matplotlib), a canvas. The important insight: **good answers can be filed back into the wiki as new pages.** A comparison you asked for, an analysis, a connection you discovered — these are valuable and shouldn't disappear into chat history. This way your explorations compound in the knowledge base just like ingested sources do.
|
||||
**Query.** You ask questions against the knowledge. The LLM searches for relevant pages, reads them, and synthesizes an answer with citations. Answers can take different forms depending on the question — a markdown page, a comparison table, a slide deck (Marp), a chart (matplotlib), a canvas. The important insight: **good answers can be filed back into the knowledge as new pages.** A comparison you asked for, an analysis, a connection you discovered — these are valuable and shouldn't disappear into chat history. This way your explorations compound in the knowledge base just like ingested sources do.
|
||||
|
||||
**Lint.** Periodically, ask the LLM to health-check the wiki. Look for: contradictions between pages, stale claims that newer sources have superseded, orphan pages with no inbound links, important concepts mentioned but lacking their own page, missing cross-references, data gaps that could be filled with a web search. The LLM is good at suggesting new questions to investigate and new sources to look for. This keeps the wiki healthy as it grows.
|
||||
**Lint.** Periodically, ask the LLM to health-check the knowledge. Look for: contradictions between pages, stale claims that newer sources have superseded, orphan pages with no inbound links, important concepts mentioned but lacking their own page, missing cross-references, data gaps that could be filled with a web search. The LLM is good at suggesting new questions to investigate and new sources to look for. This keeps the knowledge healthy as it grows.
|
||||
|
||||
## Indexing and logging
|
||||
|
||||
Two special files help the LLM (and you) navigate the wiki as it grows. They serve different purposes:
|
||||
Two special files help the LLM (and you) navigate the knowledge as it grows. They serve different purposes:
|
||||
|
||||
**index.md** is content-oriented. It's a catalog of everything in the wiki — each page listed with a link, a one-line summary, and optionally metadata like date or source count. Organized by category (entities, concepts, sources, etc.). The LLM updates it on every ingest. When answering a query, the LLM reads the index first to find relevant pages, then drills into them. This works surprisingly well at moderate scale (~100 sources, ~hundreds of pages) and avoids the need for embedding-based RAG infrastructure.
|
||||
**index.md** is content-oriented. It's a catalog of everything in the knowledge — each page listed with a link, a one-line summary, and optionally metadata like date or source count. Organized by category (entities, concepts, sources, etc.). The LLM updates it on every ingest. When answering a query, the LLM reads the index first to find relevant pages, then drills into them. This works surprisingly well at moderate scale (~100 sources, ~hundreds of pages) and avoids the need for embedding-based RAG infrastructure.
|
||||
|
||||
**log.md** is chronological. It's an append-only record of what happened and when — ingests, queries, lint passes. A useful tip: if each entry starts with a consistent prefix (e.g. `## [2026-04-02] ingest | Article Title`), the log becomes parseable with simple unix tools — `grep "^## \[" log.md | tail -5` gives you the last 5 entries. The log gives you a timeline of the wiki's evolution and helps the LLM understand what's been done recently.
|
||||
**log.md** is chronological. It's an append-only record of what happened and when — ingests, queries, lint passes. A useful tip: if each entry starts with a consistent prefix (e.g. `## [2026-04-02] ingest | Article Title`), the log becomes parseable with simple unix tools — `grep "^## \[" log.md | tail -5` gives you the last 5 entries. The log gives you a timeline of the knowledge's evolution and helps the LLM understand what's been done recently.
|
||||
|
||||
## Optional: CLI tools
|
||||
|
||||
At some point you may want to build small tools that help the LLM operate on the wiki more efficiently. A search engine over the wiki pages is the most obvious one — at small scale the index file is enough, but as the wiki grows you want proper search. [qmd](https://github.com/tobi/qmd) is a good option: it's a local search engine for markdown files with hybrid BM25/vector search and LLM re-ranking, all on-device. It has both a CLI (so the LLM can shell out to it) and an MCP server (so the LLM can use it as a native tool). You could also build something simpler yourself — the LLM can help you vibe-code a naive search script as the need arises.
|
||||
At some point you may want to build small tools that help the LLM operate on the knowledge more efficiently. A search engine over the knowledge pages is the most obvious one — at small scale the index file is enough, but as the knowledge grows you want proper search. [qmd](https://github.com/tobi/qmd) is a good option: it's a local search engine for markdown files with hybrid BM25/vector search and LLM re-ranking, all on-device. It has both a CLI (so the LLM can shell out to it) and an MCP server (so the LLM can use it as a native tool). You could also build something simpler yourself — the LLM can help you vibe-code a naive search script as the need arises.
|
||||
|
||||
## Tips and tricks
|
||||
|
||||
- **Obsidian Web Clipper** is a browser extension that converts web articles to markdown. Very useful for quickly getting sources into your raw collection.
|
||||
- **Download images locally.** In Obsidian Settings → Files and links, set "Attachment folder path" to a fixed directory (e.g. `raw/assets/`). Then in Settings → Hotkeys, search for "Download" to find "Download attachments for current file" and bind it to a hotkey (e.g. Ctrl+Shift+D). After clipping an article, hit the hotkey and all images get downloaded to local disk. This is optional but useful — it lets the LLM view and reference images directly instead of relying on URLs that may break. Note that LLMs can't natively read markdown with inline images in one pass — the workaround is to have the LLM read the text first, then view some or all of the referenced images separately to gain additional context. It's a bit clunky but works well enough.
|
||||
- **Obsidian's graph view** is the best way to see the shape of your wiki — what's connected to what, which pages are hubs, which are orphans.
|
||||
- **Marp** is a markdown-based slide deck format. Obsidian has a plugin for it. Useful for generating presentations directly from wiki content.
|
||||
- **Dataview** is an Obsidian plugin that runs queries over page frontmatter. If your LLM adds YAML frontmatter to wiki pages (tags, dates, source counts), Dataview can generate dynamic tables and lists.
|
||||
- The wiki is just a git repo of markdown files. You get version history, branching, and collaboration for free.
|
||||
- **Obsidian's graph view** is the best way to see the shape of your knowledge — what's connected to what, which pages are hubs, which are orphans.
|
||||
- **Marp** is a markdown-based slide deck format. Obsidian has a plugin for it. Useful for generating presentations directly from knowledge content.
|
||||
- **Dataview** is an Obsidian plugin that runs queries over page frontmatter. If your LLM adds YAML frontmatter to knowledge pages (tags, dates, source counts), Dataview can generate dynamic tables and lists.
|
||||
- The knowledge is just a git repo of markdown files. You get version history, branching, and collaboration for free.
|
||||
|
||||
## Why this works
|
||||
|
||||
The tedious part of maintaining a knowledge base is not the reading or the thinking — it's the bookkeeping. Updating cross-references, keeping summaries current, noting when new data contradicts old claims, maintaining consistency across dozens of pages. Humans abandon wikis because the maintenance burden grows faster than the value. LLMs don't get bored, don't forget to update a cross-reference, and can touch 15 files in one pass. The wiki stays maintained because the cost of maintenance is near zero.
|
||||
The tedious part of maintaining a knowledge base is not the reading or the thinking — it's the bookkeeping. Updating cross-references, keeping summaries current, noting when new data contradicts old claims, maintaining consistency across dozens of pages. Humans abandon knowledges because the maintenance burden grows faster than the value. LLMs don't get bored, don't forget to update a cross-reference, and can touch 15 files in one pass. The knowledge stays maintained because the cost of maintenance is near zero.
|
||||
|
||||
The human's job is to curate sources, direct the analysis, ask good questions, and think about what it all means. The LLM's job is everything else.
|
||||
|
||||
@@ -72,4 +72,4 @@ The idea is related in spirit to Vannevar Bush's Memex (1945) — a personal, cu
|
||||
|
||||
## Note
|
||||
|
||||
This document is intentionally abstract. It describes the idea, not a specific implementation. The exact directory structure, the schema conventions, the page formats, the tooling — all of that will depend on your domain, your preferences, and your LLM of choice. Everything mentioned above is optional and modular — pick what's useful, ignore what isn't. For example: your sources might be text-only, so you don't need image handling at all. Your wiki might be small enough that the index file is all you need, no search engine required. You might not care about slide decks and just want markdown pages. You might want a completely different set of output formats. The right way to use this is to share it with your LLM agent and work together to instantiate a version that fits your needs. The document's only job is to communicate the pattern. Your LLM can figure out the rest.@_in
|
||||
This document is intentionally abstract. It describes the idea, not a specific implementation. The exact directory structure, the schema conventions, the page formats, the tooling — all of that will depend on your domain, your preferences, and your LLM of choice. Everything mentioned above is optional and modular — pick what's useful, ignore what isn't. For example: your sources might be text-only, so you don't need image handling at all. Your knowledge might be small enough that the index file is all you need, no search engine required. You might not care about slide decks and just want markdown pages. You might want a completely different set of output formats. The right way to use this is to share it with your LLM agent and work together to instantiate a version that fits your needs. The document's only job is to communicate the pattern. Your LLM can figure out the rest.@_in
|
||||
|
||||
@@ -4,7 +4,7 @@ type: "concept"
|
||||
tags: [AI, automatyzacja, machine-learning, karpathy]
|
||||
created: 2026-05-14
|
||||
updated: 2026-05-14
|
||||
sources: ["[[summaries/how-to-deploy-autoresearch]]", "[[summaries/karpathy-llm-wiki-breakdown]]"]
|
||||
sources: ["[[summaries/how-to-deploy-autoresearch]]", "[[summaries/karpathy-llm-knowledge-breakdown]]"]
|
||||
confidence: high
|
||||
---
|
||||
|
||||
@@ -42,4 +42,5 @@ System działa w pętli zamkniętej:
|
||||
|
||||
## Sources
|
||||
- [[summaries/how-to-deploy-autoresearch]]
|
||||
- [[summaries/karpathy-llm-wiki-breakdown]]
|
||||
- [[summaries/karpathy-llm-knowledge-breakdown]]
|
||||
|
||||
|
||||
@@ -4,14 +4,14 @@ type: "concept"
|
||||
tags: [produktywność, AI, wiedza, karpathy]
|
||||
created: 2026-05-14
|
||||
updated: 2026-05-14
|
||||
sources: ["[[summaries/karpathy-llm-wiki-breakdown]]"]
|
||||
sources: ["[[summaries/karpathy-llm-knowledge-breakdown]]"]
|
||||
confidence: high
|
||||
---
|
||||
|
||||
# Kumulowanie Wiedzy (Compounding Knowledge)
|
||||
|
||||
## Definition
|
||||
Właściwość bazy wiedzy (Wiki), dzięki której z każdym nowym dodanym źródłem staje się ona nie tylko większa (więcej plików), ale przede wszystkim gęstsza i bogatsza w powiązania.
|
||||
Właściwość bazy wiedzy (Knowledge Base), dzięki której z każdym nowym dodanym źródłem staje się ona nie tylko większa (więcej plików), ale przede wszystkim gęstsza i bogatsza w powiązania.
|
||||
|
||||
## How It Works
|
||||
Podczas dodawania nowego źródła, LLM nie tworzy izolowanej notatki, ale aktywnie aktualizuje istniejące strony encji i koncepcji. Nowe informacje wzmacniają lub podważają stare twierdzenia, a automatyczne linkowanie krzyżowe buduje gęstą sieć skojarzeń.
|
||||
@@ -33,4 +33,5 @@ Podczas dodawania nowego źródła, LLM nie tworzy izolowanej notatki, ale aktyw
|
||||
- [[feynman_problems]] - Narzędzie do ukierunkowania kumulacji wiedzy na konkretne problemy.
|
||||
|
||||
## Sources
|
||||
- [[summaries/karpathy-llm-wiki-breakdown]]
|
||||
- [[summaries/karpathy-llm-knowledge-breakdown]]
|
||||
|
||||
|
||||
@@ -4,17 +4,17 @@ type: "concept"
|
||||
tags: [AI, wiedza, architektura, karpathy]
|
||||
created: 2026-05-14
|
||||
updated: 2026-05-14
|
||||
sources: ["[[summaries/karpathy-llm-wiki-breakdown]]"]
|
||||
sources: ["[[summaries/karpathy-llm-knowledge-breakdown]]"]
|
||||
confidence: high
|
||||
---
|
||||
|
||||
# Kompilacja Wiedzy (Knowledge Compilation)
|
||||
|
||||
## Definition
|
||||
Analogia zapożyczona z inżynierii oprogramowania, gdzie surowe dokumenty źródłowe (PDFy, notatki, artykuły) są traktowane jako "kod źródłowy", który LLM "kompiluje" do postaci trwałej Wiki ("pliku binarnego").
|
||||
Analogia zapożyczona z inżynierii oprogramowania, gdzie surowe dokumenty źródłowe (PDFy, notatki, artykuły) są traktowane jako "kod źródłowy", który LLM "kompiluje" do postaci trwałej Knowledge Base ("pliku binarnego").
|
||||
|
||||
## How It Works
|
||||
Zamiast przeszukiwać surowe fragmenty tekstów przy każdym pytaniu, LLM przetwarza je raz w momencie ingestji, ekstrahując esencję i integrując ją z istniejącą strukturą. Wynikowa Wiki jest sformatowana tak, aby była optymalna dla przyszłych zapytań i syntez.
|
||||
Zamiast przeszukiwać surowe fragmenty tekstów przy każdym pytaniu, LLM przetwarza je raz w momencie ingestji, ekstrahując esencję i integrując ją z istniejącą strukturą. Wynikowa Knowledge Base jest sformatowana tak, aby była optymalna dla przyszłych zapytań i syntez.
|
||||
|
||||
## Key Parameters
|
||||
- **Niezmienność źródeł (Immutability)**: Oryginalne pliki pozostają nienaruszone jako "ground truth".
|
||||
@@ -26,11 +26,12 @@ Zamiast przeszukiwać surowe fragmenty tekstów przy każdym pytaniu, LLM przetw
|
||||
|
||||
## Risks & Pitfalls
|
||||
- **Koszt Ingestji**: Proces kompilacji jest droższy (wymaga więcej tokenów) niż proste indeksowanie RAG.
|
||||
- **Ryzyko "utrwalenia" błędów**: Halucynacja podczas kompilacji może zostać zapisana w Wiki jako fakt.
|
||||
- **Ryzyko "utrwalenia" błędów**: Halucynacja podczas kompilacji może zostać zapisana w Knowledge Base jako fakt.
|
||||
|
||||
## Related Concepts
|
||||
- [[concepts/compounding-knowledge]] - Rezultat skutecznej kompilacji.
|
||||
- [[concepts/rag-vs-wiki]] - Alternatywne podejście "interpretowane".
|
||||
- [[concepts/rag-vs-knowledge]] - Alternatywne podejście "interpretowane".
|
||||
|
||||
## Sources
|
||||
- [[summaries/karpathy-llm-wiki-breakdown]]
|
||||
- [[summaries/karpathy-llm-knowledge-breakdown]]
|
||||
|
||||
|
||||
+11
-10
@@ -1,27 +1,27 @@
|
||||
---
|
||||
title: "RAG vs LLM Wiki"
|
||||
title: "RAG vs LLM Knowledge Base"
|
||||
type: "concept"
|
||||
tags: [AI, architektura, RAG, wiki]
|
||||
tags: [AI, architektura, RAG, knowledge]
|
||||
created: 2026-05-14
|
||||
updated: 2026-05-14
|
||||
sources: ["[[summaries/karpathy-llm-wiki-breakdown]]"]
|
||||
sources: ["[[summaries/karpathy-llm-knowledge-breakdown]]"]
|
||||
confidence: high
|
||||
---
|
||||
|
||||
# RAG vs LLM Wiki
|
||||
# RAG vs LLM Knowledge Base
|
||||
|
||||
## Definition
|
||||
Porównanie dwóch paradygmatów wykorzystania LLM do pracy z dokumentami: Retrieval-Augmented Generation (RAG) oraz persistent LLM Wiki.
|
||||
Porównanie dwóch paradygmatów wykorzystania LLM do pracy z dokumentami: Retrieval-Augmented Generation (RAG) oraz persistent LLM Knowledge Base.
|
||||
|
||||
## How It Works
|
||||
- **RAG**: Działa w trybie "bezstanowym". Na każde pytanie wyszukuje fragmenty w surowych plikach i składa z nich odpowiedź "ad hoc".
|
||||
- **Wiki**: Działa w trybie "stanowym". Wiedza jest wstępnie przetworzona (skompilowana) do struktury Wiki, a odpowiedzi są generowane na podstawie tej syntezy.
|
||||
- **Knowledge Base**: Działa w trybie "stanowym". Wiedza jest wstępnie przetworzona (skompilowana) do struktury Knowledge Base, a odpowiedzi są generowane na podstawie tej syntezy.
|
||||
|
||||
## Comparison Table
|
||||
|
||||
| Cecha | RAG | LLM Wiki |
|
||||
| Cecha | RAG | LLM Knowledge Base |
|
||||
| :--- | :--- | :--- |
|
||||
| **Dane** | Surowe dokumenty | Skompilowane strony wiki |
|
||||
| **Dane** | Surowe dokumenty | Skompilowane strony knowledge |
|
||||
| **Stanowość** | Bezstanowy (każde zapytanie od zera) | Stanowy (wiedza kumuluje się) |
|
||||
| **Skala** | Miliony dokumentów | 100-500 wybranych źródeł |
|
||||
| **Koszt** | Tani ingest, drogie zapytania (tokeny) | Drogi ingest, tanie zapytania |
|
||||
@@ -29,11 +29,12 @@ Porównanie dwóch paradygmatów wykorzystania LLM do pracy z dokumentami: Retri
|
||||
|
||||
## When To Use
|
||||
- **Wybierz RAG**: Gdy masz ogromne, dynamicznie zmieniające się zbiory danych (np. dokumentacja techniczna firmy, akta prawne).
|
||||
- **Wybierz Wiki**: Gdy pracujesz nad konkretnym tematem badawczym, piszesz książkę lub budujesz osobistą bazę wiedzy (Second Brain).
|
||||
- **Wybierz Knowledge Base**: Gdy pracujesz nad konkretnym tematem badawczym, piszesz książkę lub budujesz osobistą bazę wiedzy (Second Brain).
|
||||
|
||||
## Related Concepts
|
||||
- [[concepts/knowledge-compilation]]
|
||||
- [[concepts/compounding-knowledge]]
|
||||
|
||||
## Sources
|
||||
- [[summaries/karpathy-llm-wiki-breakdown]]
|
||||
- [[summaries/karpathy-llm-knowledge-breakdown]]
|
||||
|
||||
@@ -4,7 +4,7 @@ type: "concept"
|
||||
tags: [AI, programowanie, machine-learning, karpathy]
|
||||
created: 2026-05-14
|
||||
updated: 2026-05-14
|
||||
sources: ["[[summaries/software-2-0]]", "[[summaries/karpathy-llm-wiki-breakdown]]", "[[summaries/how-to-deploy-autoresearch]]"]
|
||||
sources: ["[[summaries/software-2-0]]", "[[summaries/karpathy-llm-knowledge-breakdown]]", "[[summaries/how-to-deploy-autoresearch]]"]
|
||||
confidence: high
|
||||
---
|
||||
|
||||
@@ -41,4 +41,5 @@ Zamiast budować algorytm, programista 2.0:
|
||||
|
||||
## Sources
|
||||
- [[summaries/software-2-0]]
|
||||
- [[summaries/karpathy-llm-wiki-breakdown]]
|
||||
- [[summaries/karpathy-llm-knowledge-breakdown]]
|
||||
|
||||
|
||||
@@ -7,7 +7,7 @@ Poniższa lista zawiera kluczowe pytania i obszary zainteresowań, które są "t
|
||||
- [[summaries/jak-zbudowalem-agenta-newslettera-n8n]]: Automatyzacja powtarzalnych zadań pozwala zaoszczędzić godziny pracy tygodniowo.
|
||||
- [[summaries/one-file-diary]]: System jednego pliku tekstowego minimalizuje narzut narzędziowy.
|
||||
- [[summaries/google-nanobanana-workflow]]: Skrócenie czasu tworzenia zasobów wizualnych o 90% dzięki [[concepts/ai-pipelines]].
|
||||
- [[summaries/karpathy-llm-wiki-breakdown]]: Wykorzystanie LLM do "księgowości wiedzy" (bookkeeping).
|
||||
- [[summaries/karpathy-llm-knowledge-breakdown]]: Wykorzystanie LLM do "księgowości wiedzy" (bookkeeping).
|
||||
- [[summaries/how-to-deploy-autoresearch]]: Automatyzacja pętli badawczej – agent pracuje, gdy człowiek śpi.
|
||||
|
||||
## 2. Zarządzanie bazami danych
|
||||
@@ -30,8 +30,9 @@ Poniższa lista zawiera kluczowe pytania i obszary zainteresowań, które są "t
|
||||
- [[summaries/jak-zbudowalem-agenta-newslettera-n8n]]: Workflow n8n jako praktyczny agent redakcyjny.
|
||||
- [[summaries/ai-impact-on-dba]]: Kierunek ku autonomicznym bazom danych.
|
||||
- [[summaries/google-nanobanana-workflow]]: [[concepts/ai-pipelines]] integrujące wiele modeli.
|
||||
- [[summaries/karpathy-llm-wiki-breakdown]]: Model LLM jako "programista" i "maintainer" bazy wiedzy.
|
||||
- [[summaries/karpathy-llm-knowledge-breakdown]]: Model LLM jako "programista" i "maintainer" bazy wiedzy.
|
||||
- [[summaries/how-to-deploy-autoresearch]]: [[concepts/autoresearch]] – najwyższy stopień autonomii agenta AI w procesie badawczym.
|
||||
|
||||
---
|
||||
*Miejsce na kolejne problemy (5-12)...*
|
||||
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
# Wiki Index
|
||||
# Knowledge Base Index
|
||||
|
||||
Główny katalog wiedzy, podzielony na obszary tematyczne dla łatwiejszej nawigacji.
|
||||
|
||||
@@ -25,6 +25,7 @@ Dla osób szukających konkretnego formatu pliku.
|
||||
---
|
||||
|
||||
## Log i Administracja
|
||||
- [[log]] — Historia zmian w Wiki.
|
||||
- [[log]] — Historia zmian w Knowledge Base.
|
||||
- [[feynman_problems]] — Centralne śledzenie problemów.
|
||||
- [[journal/index]] — Dziennik badawczy.
|
||||
|
||||
|
||||
@@ -3,7 +3,7 @@
|
||||
Zbiór zasobów dotyczących nowej paradygmatu programowania, systemów opartych na LLM oraz autonomicznego researchu.
|
||||
|
||||
## Podsumowania
|
||||
- [[summaries/karpathy-llm-wiki-breakdown]] — Analiza LLM Wiki Karpathy'ego.
|
||||
- [[summaries/karpathy-llm-knowledge-breakdown]] — Analiza LLM Knowledge Base Karpathy'ego.
|
||||
- [[summaries/how-to-deploy-autoresearch]] — Przewodnik po Autoresearch.
|
||||
- [[summaries/software-2-0]] — Manifest Software 2.0.
|
||||
|
||||
@@ -13,9 +13,10 @@ Zbiór zasobów dotyczących nowej paradygmatu programowania, systemów opartych
|
||||
- [[concepts/autoresearch]] — Autonomiczne badania nad AI prowadzone przez agentów.
|
||||
- [[concepts/ai-pipelines]] — Zautomatyzowane rurociągi przetwarzania AI.
|
||||
- [[concepts/ai-hallucinations]] — Ryzyko błędów w modelach AI.
|
||||
- [[concepts/rag-vs-wiki]] — Porównanie podejścia stanowego i bezstanowego.
|
||||
- [[concepts/rag-vs-knowledge]] — Porównanie podejścia stanowego i bezstanowego.
|
||||
- [[concepts/human-in-the-loop]] — Model współpracy człowieka z AI.
|
||||
- [[concepts/val-bpb]] — Metryka jakości modelu (Bits Per Byte).
|
||||
|
||||
## Podmioty
|
||||
- [[entities/tavily]] — Silnik researchu dla AI.
|
||||
|
||||
|
||||
@@ -13,4 +13,5 @@ Narzędzia i strategie automatyzacji powtarzalnych zadań oraz łączenia różn
|
||||
- [[entities/n8n]] — Narzędzie do automatyzacji workflow.
|
||||
|
||||
## Syntezy
|
||||
- [[syntheses/porownanie-strategii-automatyzacji-ai]] — Zestawienie metodologii (n8n, Nanobanana, LLM Wiki).
|
||||
- [[syntheses/porownanie-strategii-automatyzacji-ai]] — Zestawienie metodologii (n8n, Nanobanana, LLM Knowledge Base).
|
||||
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
# Wszystkie Koncepcje
|
||||
|
||||
Słownik terminów i modeli mentalnych zgromadzonych w wiki.
|
||||
Słownik terminów i modeli mentalnych zgromadzonych w knowledge.
|
||||
|
||||
- [[concepts/human-in-the-loop]]
|
||||
- [[concepts/accidental-dba]]
|
||||
@@ -11,10 +11,11 @@ Słownik terminów i modeli mentalnych zgromadzonych w wiki.
|
||||
- [[concepts/ai-pipelines]]
|
||||
- [[concepts/knowledge-compilation]]
|
||||
- [[concepts/compounding-knowledge]]
|
||||
- [[concepts/rag-vs-wiki]]
|
||||
- [[concepts/rag-vs-knowledge]]
|
||||
- [[concepts/autoresearch]]
|
||||
- [[concepts/val-bpb]]
|
||||
- [[concepts/software-1-0]]
|
||||
- [[concepts/software-2-0]]
|
||||
- [[concepts/idea-collision]]
|
||||
- [[concepts/metacognition-as-a-service]]
|
||||
|
||||
|
||||
@@ -14,7 +14,8 @@ Metodyki gromadzenia wiedzy, budowania własnego PKM (Personal Knowledge Managem
|
||||
## Koncepcje
|
||||
- [[concepts/plain-text-productivity]] — Produktywność oparta na plikach tekstowych.
|
||||
- [[concepts/work-journaling]] — Metodyka prowadzenia dziennika pracy.
|
||||
- [[concepts/knowledge-compilation]] — "Kompilowanie" surowej wiedzy do Wiki.
|
||||
- [[concepts/knowledge-compilation]] — "Kompilowanie" surowej wiedzy do Knowledge Base.
|
||||
- [[concepts/compounding-knowledge]] — Wiedza, która staje się gęstsza z czasem.
|
||||
- [[concepts/idea-collision]] — Zderzanie pomysłów w celu generowania innowacji.
|
||||
- [[concepts/metacognition-as-a-service]] — AI jako wsparcie analizy własnych procesów myślowych.
|
||||
|
||||
|
||||
@@ -7,8 +7,9 @@ Pełna lista streszczeń materiałów źródłowych.
|
||||
- [[summaries/accidental-dba-intro]]
|
||||
- [[summaries/ai-impact-on-dba]]
|
||||
- [[summaries/google-nanobanana-workflow]]
|
||||
- [[summaries/karpathy-llm-wiki-breakdown]]
|
||||
- [[summaries/karpathy-llm-knowledge-breakdown]]
|
||||
- [[summaries/how-to-deploy-autoresearch]]
|
||||
- [[summaries/software-2-0]]
|
||||
- [[summaries/loom-thinking-tool]]
|
||||
- [[summaries/free-blog-seo-strategy]]
|
||||
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
# Wiki Log
|
||||
# Knowledge Base Log
|
||||
|
||||
Chronologiczny zapis operacji na wiki.
|
||||
Chronologiczny zapis operacji na knowledge.
|
||||
|
||||
## [2026-05-14] Ingest | Advanced AI & Knowledge Management
|
||||
- Przetworzono `Jak wdrożyć autoresearch_.md`: dodano koncepcję `[[autoresearch]]` i metrykę `[[val-bpb]]`.
|
||||
@@ -13,8 +13,8 @@ Chronologiczny zapis operacji na wiki.
|
||||
## [2026-05-14] Synteza | Strategie Automatyzacji AI
|
||||
- Utworzono stronę syntezy: `[[syntheses/porownanie-strategii-automatyzacji-ai]]`.
|
||||
|
||||
## [2026-05-14] Ingest | Karpathy LLM Wiki Breakdown
|
||||
- Przetworzono artykuł o LLM Wiki.
|
||||
## [2026-05-14] Ingest | Karpathy LLM Knowledge Base Breakdown
|
||||
- Przetworzono artykuł o LLM Knowledge Base.
|
||||
|
||||
## [2026-05-14] Ingest | AI impact & Nanobanana Workflow
|
||||
- Przetworzono artykuły o wpływie AI na DBA i workflow Nanobanana.
|
||||
@@ -26,4 +26,5 @@ Chronologiczny zapis operacji na wiki.
|
||||
- Zaktualizowano `GEMINI.md` i strukturę katalogów.
|
||||
|
||||
## [2026-05-14] Inicjalizacja
|
||||
- Utworzono fundamenty Wiki LLM.
|
||||
- Utworzono fundamenty Knowledge Base LLM.
|
||||
|
||||
|
||||
@@ -0,0 +1,32 @@
|
||||
---
|
||||
title: "Andrej Karpathy’s LLM Knowledge Base: Create your own knowledge base"
|
||||
type: "summary"
|
||||
tags: [karpathy, llm-knowledge, knowledge-management, AI, obsidian, RAG]
|
||||
created: 2026-05-14
|
||||
updated: 2026-05-14
|
||||
sources: ["raw/articles/Andrej Karpathy’s LLM Knowledge Base_ Create your own knowledge base.md"]
|
||||
confidence: high
|
||||
---
|
||||
|
||||
# Andrej Karpathy’s LLM Knowledge Base: Create your own knowledge base
|
||||
|
||||
Artykuł Urvila Joshiego szczegółowo analizuje koncepcję "LLM Knowledge Base" zaproponowaną przez Andreja Karpathy'ego. Tekst wyjaśnia, dlaczego tradycyjne podejście RAG (Retrieval-Augmented Generation) jest niewystarczające do głębokiej syntezy wiedzy i jak budowa trwałej, "skompilowanej" bazy wiedzy rozwiązuje ten problem.
|
||||
|
||||
## Key Points
|
||||
- **Kompilacja wiedzy**: Surowe źródła to "kod źródłowy", a Knowledge Base to "plik binarny" – zoptymalizowany pod kątem szybkości i gęstości informacji.
|
||||
- **Trwałość i kumulacja**: Knowledge Base to artefakt, który staje się bogatszy z każdym nowym źródłem. LLM aktualizuje istniejące strony, zamiast tworzyć izolowane fragmenty.
|
||||
- **Architektura 3-warstwowa**: Raw Sources (niezmienne), Knowledge Base (zarządzane przez LLM), Schema (zasady postępowania).
|
||||
- **RAG vs Knowledge Base**: RAG jest lepszy dla milionów dokumentów i wyszukiwania faktów; Knowledge Base jest lepsza dla mniejszych, wyselekcjonowanych zbiorów (~100-500 źródeł), gdzie kluczowa jest synteza i powiązania.
|
||||
- **Księgowość wiedzy**: LLM przejmuje nudne zadania (linkowanie, aktualizacja indeksów, sprawdzanie spójności), co zapobiega porzucaniu bazy przez ludzi.
|
||||
|
||||
## Relevant Concepts
|
||||
- [[concepts/knowledge-compilation]] - Koncepcja "kompilowania" źródeł do formy syntetycznej.
|
||||
- [[concepts/compounding-knowledge]] - Budowanie wiedzy, która staje się gęstsza z czasem.
|
||||
- [[concepts/rag-vs-knowledge]] - Porównanie metodologii zarządzania wiedzą z AI.
|
||||
|
||||
## Source Metadata
|
||||
- **Type**: Artykuł / Analiza
|
||||
- **Author**: Urvil Joshi (na podstawie wpisów Andreja Karpathy'ego)
|
||||
- **Date**: 2026-04-20
|
||||
- **URL**: https://medium.com/@urvvil08/andrej-karpathys-llm-knowledge-create-your-own-knowledge-base-8779014accd5
|
||||
|
||||
@@ -1,31 +0,0 @@
|
||||
---
|
||||
title: "Andrej Karpathy’s LLM Wiki: Create your own knowledge base"
|
||||
type: "summary"
|
||||
tags: [karpathy, llm-wiki, knowledge-management, AI, obsidian, RAG]
|
||||
created: 2026-05-14
|
||||
updated: 2026-05-14
|
||||
sources: ["raw/articles/Andrej Karpathy’s LLM Wiki_ Create your own knowledge base.md"]
|
||||
confidence: high
|
||||
---
|
||||
|
||||
# Andrej Karpathy’s LLM Wiki: Create your own knowledge base
|
||||
|
||||
Artykuł Urvila Joshiego szczegółowo analizuje koncepcję "LLM Wiki" zaproponowaną przez Andreja Karpathy'ego. Tekst wyjaśnia, dlaczego tradycyjne podejście RAG (Retrieval-Augmented Generation) jest niewystarczające do głębokiej syntezy wiedzy i jak budowa trwałej, "skompilowanej" bazy wiedzy rozwiązuje ten problem.
|
||||
|
||||
## Key Points
|
||||
- **Kompilacja wiedzy**: Surowe źródła to "kod źródłowy", a Wiki to "plik binarny" – zoptymalizowany pod kątem szybkości i gęstości informacji.
|
||||
- **Trwałość i kumulacja**: Wiki to artefakt, który staje się bogatszy z każdym nowym źródłem. LLM aktualizuje istniejące strony, zamiast tworzyć izolowane fragmenty.
|
||||
- **Architektura 3-warstwowa**: Raw Sources (niezmienne), Wiki (zarządzane przez LLM), Schema (zasady postępowania).
|
||||
- **RAG vs Wiki**: RAG jest lepszy dla milionów dokumentów i wyszukiwania faktów; Wiki jest lepsza dla mniejszych, wyselekcjonowanych zbiorów (~100-500 źródeł), gdzie kluczowa jest synteza i powiązania.
|
||||
- **Księgowość wiedzy**: LLM przejmuje nudne zadania (linkowanie, aktualizacja indeksów, sprawdzanie spójności), co zapobiega porzucaniu bazy przez ludzi.
|
||||
|
||||
## Relevant Concepts
|
||||
- [[concepts/knowledge-compilation]] - Koncepcja "kompilowania" źródeł do formy syntetycznej.
|
||||
- [[concepts/compounding-knowledge]] - Budowanie wiedzy, która staje się gęstsza z czasem.
|
||||
- [[concepts/rag-vs-wiki]] - Porównanie metodologii zarządzania wiedzą z AI.
|
||||
|
||||
## Source Metadata
|
||||
- **Type**: Artykuł / Analiza
|
||||
- **Author**: Urvil Joshi (na podstawie wpisów Andreja Karpathy'ego)
|
||||
- **Date**: 2026-04-20
|
||||
- **URL**: https://medium.com/@urvvil08/andrej-karpathys-llm-wiki-create-your-own-knowledge-base-8779014accd5
|
||||
@@ -1,10 +1,10 @@
|
||||
---
|
||||
title: "Porównanie strategii automatyzacji researchu i tworzenia treści"
|
||||
type: "synthesis"
|
||||
tags: [automatyzacja, AI, n8n, newsletter, content-creation, llm-wiki]
|
||||
tags: [automatyzacja, AI, n8n, newsletter, content-creation, llm-knowledge]
|
||||
created: 2026-05-14
|
||||
updated: 2026-05-14
|
||||
sources: ["[[summaries/jak-zbudowalem-agenta-newslettera-n8n]]", "[[summaries/google-nanobanana-workflow]]", "[[summaries/karpathy-llm-wiki-breakdown]]"]
|
||||
sources: ["[[summaries/jak-zbudowalem-agenta-newslettera-n8n]]", "[[summaries/google-nanobanana-workflow]]", "[[summaries/karpathy-llm-knowledge-breakdown]]"]
|
||||
confidence: high
|
||||
---
|
||||
|
||||
@@ -14,7 +14,7 @@ Niniejsza synteza zestawia trzy kluczowe podejścia do wykorzystania agentów AI
|
||||
|
||||
## Comparison
|
||||
|
||||
| Cecha | Agent Newslettera (n8n) | Workflow Nanobanana | LLM Wiki Pattern |
|
||||
| Cecha | Agent Newslettera (n8n) | Workflow Nanobanana | LLM Knowledge Base Pattern |
|
||||
| :--- | :--- | :--- | :--- |
|
||||
| **Główny cel** | Produkcja cyklicznego tekstu (newsletter) | Produkcja masowa grafik/zasobów wizualnych | Budowa trwałej bazy wiedzy (Second Brain) |
|
||||
| **Kluczowe narzędzia** | n8n, Tavily, OpenRouter, Gmail | n8n, Gemini 2.5 Flash, Telegram | LLM (np. Gemini/Claude), Obsidian |
|
||||
@@ -28,7 +28,7 @@ Niniejsza synteza zestawia trzy kluczowe podejścia do wykorzystania agentów AI
|
||||
Analiza wykazuje ewolucję od prostych agentów do złożonych rurociągów:
|
||||
- **Metoda n8n** skupia się na linearnym procesie: od researchu do tekstu. Jest to klasyczny przykład "asystenta pisarza".
|
||||
- **Workflow Nanobanana** wprowadza koncepcję "rurociągu" ([[concepts/ai-pipelines]]), gdzie jeden impuls (seed) generuje wielokrotne plony w różnych formatach. Skupia się na szybkości i niskim koszcie (model Flash).
|
||||
- **LLM Wiki** to podejście najbardziej zaawansowane pod kątem struktury. Nie generuje "produktu" na zewnątrz, ale buduje wewnętrzny "system operacyjny wiedzy".
|
||||
- **LLM Knowledge Base** to podejście najbardziej zaawansowane pod kątem struktury. Nie generuje "produktu" na zewnątrz, ale buduje wewnętrzny "system operacyjny wiedzy".
|
||||
|
||||
### 2. Wspólny mianownik: Rola człowieka
|
||||
We wszystkich trzech strategiach rola człowieka przesuwa się z **wykonawcy** na **architekta i kuratora**:
|
||||
@@ -40,12 +40,13 @@ We wszystkich trzech strategiach rola człowieka przesuwa się z **wykonawcy** n
|
||||
### Kiedy stosować daną strategię?
|
||||
- **Wybierz Agenta n8n (Newsletter):** Jeśli Twoim celem jest regularna publikacja ekspercka i potrzebujesz bazy do pisania (tryb "asystent").
|
||||
- **Wybierz Nanobanana Workflow:** Jeśli budujesz markę osobistą w wielu kanałach social media i potrzebujesz masowej produkcji wizualnej przy zachowaniu spójności stylu.
|
||||
- **Wybierz LLM Wiki Pattern:** Jeśli Twoim celem jest długoterminowe zrozumienie skomplikowanych tematów (np. badania nad AI, bazy danych) i chcesz, aby Twoja wiedza "kumulowała się" ([[concepts/compounding-knowledge]]), a nie znikała w archiwach.
|
||||
- **Wybierz LLM Knowledge Base Pattern:** Jeśli Twoim celem jest długoterminowe zrozumienie skomplikowanych tematów (np. badania nad AI, bazy danych) i chcesz, aby Twoja wiedza "kumulowała się" ([[concepts/compounding-knowledge]]), a nie znikała w archiwach.
|
||||
|
||||
## Pages Compared
|
||||
- [[summaries/jak-zbudowalem-agenta-newslettera-n8n]]
|
||||
- [[summaries/google-nanobanana-workflow]]
|
||||
- [[summaries/karpathy-llm-wiki-breakdown]]
|
||||
- [[summaries/karpathy-llm-knowledge-breakdown]]
|
||||
- [[concepts/ai-pipelines]]
|
||||
- [[concepts/human-in-the-loop]]
|
||||
- [[concepts/knowledge-compilation]]
|
||||
|
||||
|
||||
Reference in New Issue
Block a user