How Rabbithole makes its AI lessons

Every Rabbithole lesson is written from an approved list of sources, and every factual claim in it traces back to one of them.

An allowlist, not the open web

A source exists for the lessons only once it is on the list. The list has three kinds:

• Primary sources: official documentation from the companies that make the tools, research papers from the AI labs and universities, and official courses.
• Named researchers speaking: university lectures and talks by researchers with a verifiable affiliation.
• Explainers by researchers and lab engineering blogs, checked against a rubric.

Left out by default: listicles, SEO content, AI news channels, anything without a named author whose credentials can be checked, and summaries of a source we could read in the original.

Every claim traced

No lesson, exercise, hint or answer ships unless every factual claim in it resolves to a passage in an approved source. A claim that can't be traced is dropped, not reworded. When sources disagree, the lesson says so.

The WHY button

After an answer, tap WHY to see the source behind it.

In our own words

Lessons teach in their own words, and every lesson links to the sources it was made from. Short examples name and link their source.

No price lists

Prices change too often to teach. When a lesson needs another company's price, it points you to that company's own pricing page.

The sources today

Today's 672 lessons draw on 384 sources: 332 primary sources, 31 lecture series and talks by named researchers, and 21 researcher-written explainers. The most used:

1. Lil'Log (researcher explainer, 57 lessons)
2. Lovable documentation (primary source, 37 lessons)
3. Hugging Face LLM Course (researcher explainer, 36 lessons)
4. LLMs from Scratch (researcher explainer, 32 lessons)
5. ElevenLabs documentation (primary source, 29 lessons)
6. Gemini API Documentation (primary source, 24 lessons)
7. openai/openai-cookbook (primary source, 24 lessons)
8. Supabase Documentation (primary source, 23 lessons)
9. Claude Cookbooks (primary source, 23 lessons)
10. Machine Learning Crash Course (primary source, 23 lessons)
11. The Illustrated Transformer (and series) (researcher explainer, 21 lessons)
12. Modal Guide (primary source, 20 lessons)
13. Transformer Circuits Thread (primary source, 20 lessons)
14. torchtitan: A PyTorch native platform for training generative AI models (primary source, 20 lessons)
15. MIT 6.S191: Introduction to Deep Learning (named researcher, 20 lessons)
16. OpenAI Codex and ChatGPT desktop documentation (ChatGPT Learn) (primary source, 19 lessons)