Stats

Actions

Available In

Help us improve

Share bugs, ideas, or general feedback.

Kata

A self-evolving knowledge system for AI-paired builders. Inherit a kata, adapt it to your project, then transcend the form.

What Kata is (the layered model)

Layer

What it is

Base

Karpathy's LLM-Wiki principle — compiled once, kept current; humans curate, LLMs maintain; everything optional and modular.

Core

A self-evolving knowledge system built on the base: (1) a self-closing loop — ingest → cross-link → filed-query → compounding pages; (2) auto-dreaming — frozen pages resurface when their relevance returns. The wiki doesn't just persist; it grows back toward what's currently mattering.

Phase 1 (current)

AI-paired engineering. Use the core wiki to compile project business semantics — thresholds, lifecycle invariants, domain conventions — so AI agents read project conventions before they write code. v1.4 → v1.11 ship this reach.

Phase 2 (designed, not yet implemented)

Team spec authoring + dispute resolution. A self-closing loop for spec drafts, ratified positions, and the rejected alternatives — so future decisions don't re-litigate the same questions.

Phase 3+

Open. The core extends to new boundaries as we learn what compounds.

The product is the Core + Phase reaches. Phase 1 is the first concrete boundary; it's not the definition of Kata.

The kata mastery curve

Accept — wiki-init lays down a starter kata: default categories, default schema, default rituals. You run the form as-is.

Adapt — you customize schema, add domain dimensions, extend skills for your project's vocabulary. The kata becomes yours.

Transcend — the form fades. Your AI agents brief themselves, maintain cross-links, resurface frozen knowledge when it matters again. The workflow is invisible; only the work remains.

The wiki is compiled once and kept current (not RAG). The human curates sources and asks good questions; the AI does all the bookkeeping — reading, summarizing, cross-referencing, filing, maintaining consistency. You (almost) never write wiki pages yourself.

13 skills span ingest → search → graph → dream → sync — see the skill index further down. The product is the kata; the wiki is what it produces.

Design lineage

This plugin implements Karpathy's original concept and extends it with three systems he left open. The table below shows which ideas come from the original and which are plugin extensions.

From Karpathy's original

Idea

Original quote / reference

Compiled once, kept current (not RAG)

"Unlike RAG... the wiki is compiled once and then kept up to date"

Human curates, LLM maintains

"You curate sources and ask good questions, the LLM does the rest"

raw/ immutable layer

"You don't modify raw source material"

index.md — read index first

"The LLM reads the index first to find relevant pages"

log.md — append-only action log

"A chronological log of all actions taken"

Ingest: 10–15 pages per source

"A single source can touch 10-15 pages"

Query results file back and compound

"Queries compound in the wiki just like ingested sources"

Lint: stale content + data gaps

"Lint also finds data gaps that could be filled with a web search"

Schema co-evolves with the wiki

"The schema... co-evolves along with the wiki over time"

Git repo by default

"The wiki directory is just a git repo"

Obsidian as the IDE

"Obsidian is the IDE. The LLM is the programmer."

Everything optional and modular

"Pick what's useful, ignore what isn't"

qmd for scaling search

"qmd... for search at scale"

Web search to fill gaps

"The LLM is good at suggesting new questions to investigate"

Plugin extensions (this project)

Kata

A self-evolving knowledge system for AI-paired builders. Inherit a kata, adapt it to your project, then transcend the form.

What Kata is (the layered model)

Layer	What it is
Base	Karpathy's LLM-Wiki principle — compiled once, kept current; humans curate, LLMs maintain; everything optional and modular.
Core	A self-evolving knowledge system built on the base: (1) a self-closing loop — ingest → cross-link → filed-query → compounding pages; (2) auto-dreaming — frozen pages resurface when their relevance returns. The wiki doesn't just persist; it grows back toward what's currently mattering.
Phase 1 (current)	AI-paired engineering. Use the core wiki to compile project business semantics — thresholds, lifecycle invariants, domain conventions — so AI agents read project conventions before they write code. v1.4 → v1.11 ship this reach.
Phase 2 (designed, not yet implemented)	Team spec authoring + dispute resolution. A self-closing loop for spec drafts, ratified positions, and the rejected alternatives — so future decisions don't re-litigate the same questions.
Phase 3+	Open. The core extends to new boundaries as we learn what compounds.

The product is the Core + Phase reaches. Phase 1 is the first concrete boundary; it's not the definition of Kata.

The kata mastery curve

Accept — wiki-init lays down a starter kata: default categories, default schema, default rituals. You run the form as-is.
Adapt — you customize schema, add domain dimensions, extend skills for your project's vocabulary. The kata becomes yours.
Transcend — the form fades. Your AI agents brief themselves, maintain cross-links, resurface frozen knowledge when it matters again. The workflow is invisible; only the work remains.

13 skills span ingest → search → graph → dream → sync — see the skill index further down. The product is the kata; the wiki is what it produces.

Design lineage

This plugin implements Karpathy's original concept and extends it with three systems he left open. The table below shows which ideas come from the original and which are plugin extensions.

From Karpathy's original

Idea	Original quote / reference
Compiled once, kept current (not RAG)	"Unlike RAG... the wiki is compiled once and then kept up to date"
Human curates, LLM maintains	"You curate sources and ask good questions, the LLM does the rest"
`raw/` immutable layer	"You don't modify raw source material"
`index.md` — read index first	"The LLM reads the index first to find relevant pages"
`log.md` — append-only action log	"A chronological log of all actions taken"
Ingest: 10–15 pages per source	"A single source can touch 10-15 pages"
Query results file back and compound	"Queries compound in the wiki just like ingested sources"
Lint: stale content + data gaps	"Lint also finds data gaps that could be filled with a web search"
Schema co-evolves with the wiki	"The schema... co-evolves along with the wiki over time"
Git repo by default	"The wiki directory is just a git repo"
Obsidian as the IDE	"Obsidian is the IDE. The LLM is the programmer."
Everything optional and modular	"Pick what's useful, ignore what isn't"
qmd for scaling search	"qmd... for search at scale"
Web search to fill gaps	"The LLM is good at suggesting new questions to investigate"

Help us improve

Find plugins for your project

Help us improve

kata

Popularity

What's Inside

Help us improve

Health & Quality

Confidence

README

Kata

What Kata is (the layered model)

The kata mastery curve

Design lineage

From Karpathy's original

Plugin extensions (this project)

Similar Plugins

fullstack-dev-skills

context7-plugin

c4-architecture

superpowers

everything-claude-code

ui-ux-pro-max

More by surebeli

spec-team

ak-wiki

hopper

Kata

What Kata is (the layered model)

The kata mastery curve

Design lineage

From Karpathy's original

Plugin extensions (this project)