# YJTOON

> YJTOON is a free, open catalog of categorized reference datasets for AI agents and humans.
> 205 datasets across 27 categories, published as CC0 public domain data. Every dataset is
> available as JSON, YAML, and TOON (a compact encoding aimed at LLM prompts) from one URL,
> with no authentication, no API keys, and no signup. Rate limit: 120 requests per minute per IP.

Measured context (o200k_base tokenizer, all 205 datasets as served): TOON uses 183,346 tokens versus 272,226 for the pretty-printed JSON files (−33%), and 197,572 for minified JSON (−7%). 181 of 205 datasets are smaller in TOON than in minified JSON; table-shaped data gains most. In bytes: 1,322,052 B JSON vs 775,017 B TOON. Humans can read any dataset as a rendered HTML table at https://yjtoon.yunusemrevurgun.com/catalog/.

## Human-readable catalog (HTML)

- [Catalog index](https://yjtoon.yunusemrevurgun.com/catalog/): all 205 datasets, grouped by category
- [Per-category pages](https://yjtoon.yunusemrevurgun.com/catalog/{category}): e.g. https://yjtoon.yunusemrevurgun.com/catalog/programming-languages/
- [Per-dataset pages](https://yjtoon.yunusemrevurgun.com/catalog/{category}/{dataset}): the data rendered as tables, with exact sizes in each format, format links, suggested citation, and API example
- [Homepage](https://yjtoon.yunusemrevurgun.com/): browse by category, live three-format comparison, interactive voxel view of the catalog
- [Curated collections](https://yjtoon.yunusemrevurgun.com/collections/): reference tables and practical guides grouped by the task they help with
  - [Build a reliable API](https://yjtoon.yunusemrevurgun.com/collections/reliable-http/): Make cache behavior, retries, and response headers explicit before your endpoint reaches production.
  - [Give an agent better context](https://yjtoon.yunusemrevurgun.com/collections/ai-context/): Ground answers in compact reference data, validate the shape, and measure the failures.
  - [Query data with confidence](https://yjtoon.yunusemrevurgun.com/collections/querying-data/): Choose joins, windows, indexes, and transaction boundaries from the question you need to answer.
  - [Work safely in the terminal](https://yjtoon.yunusemrevurgun.com/collections/shell-toolbox/): Find files, recover Git work, and write shell commands that handle real filenames.
  - [Make the web work for everyone](https://yjtoon.yunusemrevurgun.com/collections/web-foundations/): Build usable layouts, keyboard interactions, text search, and time handling.
  - [Measure, evaluate, and operate](https://yjtoon.yunusemrevurgun.com/collections/measure-and-operate/): Connect service signals, background jobs, probability, and experiment design to real decisions.
- [TOON Studio](https://yjtoon.yunusemrevurgun.com/#studio): paste JSON and it is encoded in the browser by the same encoder the API uses; byte counts are exact, token counts are labelled estimates

## API endpoints (dynamic)

- [List categories](https://yjtoon.yunusemrevurgun.com/api/): all 27 categories with dataset counts
- [Category datasets](https://yjtoon.yunusemrevurgun.com/api/category/{slug}): datasets in a category (e.g. /api/category/programming-languages) plus a topic guide: overview, key ideas, pitfalls, agent tips and where to start
- [Single dataset](https://yjtoon.yunusemrevurgun.com/api/dataset/{slug}): full dataset content (e.g. /api/dataset/http-status-codes-complete) with a summary, use_when notes, caveats, related datasets, entry_count and a field schema alongside the data
- [Search](https://yjtoon.yunusemrevurgun.com/api/search?q={query}): search titles, descriptions, tags, and data (min 2 characters)
- All endpoints support ?format=json (default), ?format=yaml, ?format=toon; CORS is open
- Rate limit: 120 requests per minute per IP; 429 with Retry-After on exceed
- [API documentation](https://yjtoon.yunusemrevurgun.com/api.html): endpoints, formats, rate limits, copyable examples

## Static files (cache-friendly, no database)

- [Whole catalog in one request](https://yjtoon.yunusemrevurgun.com/static-data/catalog-index.json): every category and dataset summary (72 KB). TOON equivalent: https://yjtoon.yunusemrevurgun.com/static-data/catalog-index.toon (37 KB)
- [Search index](https://yjtoon.yunusemrevurgun.com/static-data/search-index.json): title, description, category, and tags for every dataset
- [Category index](https://yjtoon.yunusemrevurgun.com/static-data/categories.json)
- [Per-category files](https://yjtoon.yunusemrevurgun.com/static-data/category/{slug}.json)
- [Per-dataset files](https://yjtoon.yunusemrevurgun.com/static-data/dataset/{slug}.json|.yaml|.toon)
- [Measured size and token statistics](https://yjtoon.yunusemrevurgun.com/static-data/stats.json)
- [Every dataset and post with its URL](https://yjtoon.yunusemrevurgun.com/llms-full.txt)

## Formats

- JSON: standard, a parser in every language; 72,271 bytes for the catalog index
- YAML: easier for a human to edit and read; 48,229 bytes for the same index
- TOON: compact notation for LLM prompts; 36,787 bytes for the same index. Follows the public TOON format (https://github.com/toon-format/spec): uniform arrays of objects become tables with a `key[n]{field,field}:` header, primitive arrays are inline `key[n]: a,b,c`, strings are quoted only when ambiguous
- The browser encoder (js/toon.js) and the server encoder (api/format.php) are held byte-identical by a test over all 205 datasets, and both are checked against the reference TOON implementation, including a decode round-trip
- [TOON grammar overview](https://yjtoon.yunusemrevurgun.com/blog/toon-format-grammar-spec-overview/), [a small Python parser](https://yjtoon.yunusemrevurgun.com/blog/tiny-toon-parser-in-50-lines-of-python/), [token savings measured](https://yjtoon.yunusemrevurgun.com/blog/toon-vs-json-token-savings-measured/)

## Blog

- [Blog index](https://yjtoon.yunusemrevurgun.com/blog/): 104 posts — reference guides, cheat sheets, and explainers on HTTP, SQL, Unix, data formats, and LLM tooling
- Topic clusters: TOON and encodings, agent context economics, grounding and verification, dataset publishing and SEO, HTTP behaviour for data APIs
- Popular guides: [HTTP status codes cheat sheet](https://yjtoon.yunusemrevurgun.com/blog/http-status-codes-cheat-sheet/), [HTTP methods](https://yjtoon.yunusemrevurgun.com/blog/http-methods-get-post-put-patch-delete/), [SQL join types](https://yjtoon.yunusemrevurgun.com/blog/sql-join-types-inner-left-right-full/), [Git commands](https://yjtoon.yunusemrevurgun.com/blog/git-commands-cheat-sheet/), [Unix file commands](https://yjtoon.yunusemrevurgun.com/blog/unix-file-commands-cheat-sheet/), [LLM glossary](https://yjtoon.yunusemrevurgun.com/blog/llm-glossary-key-terms/), [Cognitive biases checklist](https://yjtoon.yunusemrevurgun.com/blog/cognitive-biases-decision-checklist/), [Network ports](https://yjtoon.yunusemrevurgun.com/blog/common-network-ports-reference/)
- [RSS feed](https://yjtoon.yunusemrevurgun.com/blog/rss.xml)

## Films

- [Films index](https://yjtoon.yunusemrevurgun.com/video/): 1 explainer film, each with chapters, English captions, the full transcript on the page and its sources
- [AI Agents 2026: Grok Bot, Muse, Dots, OpenClaw and Hermes Compared](https://yjtoon.yunusemrevurgun.com/video/ai-agents-2026/) (8:50): Grok Bot, Meta Muse, ChatGPT Dots, OpenClaw and Hermes compared in an 8-minute film: what each one does, who it is for, what it costs, and rented versus owned. Captions: https://yjtoon.yunusemrevurgun.com/video/ai-agents-2026.en.vtt (WebVTT). Video: https://yjtoon.yunusemrevurgun.com/video/ai-agents-2026.mp4. Written version: https://yjtoon.yunusemrevurgun.com/blog/grok-bot-vs-muse-vs-dots-vs-openclaw-vs-hermes-ai-agents-2026/.

## Books by the author

- [Open Data for AGI — Why Public, Free, and Open Datasets Matter](https://theknowledgeproject.gumroad.com/l/zdrla) ($4.99, PDF): the argument behind this catalog — why public, free, and open datasets decide what AI systems become. All site data remains free; the book is the synthesis, not a paywall.
- [How to Remain Valuable When Intelligence Becomes Cheap](https://theknowledgeproject.gumroad.com/l/remainvaluable) (224 pages, PDF + EPUB): the scarce human, economic, and strategic advantages that remain valuable even when AI can perform most cognitive work.
- [Full ebook catalog](https://theknowledgeproject.gumroad.com/)

## License

CC0 — public domain. Use for any purpose, commercial included, without attribution.
Attribution is welcome but never required. Suggested citation is printed on every dataset page.