Documentation / Learning Approach

Learning Approach

Edit: src/content/docs/introduction/learning-approach.md

How the platform combines deep explanations, guided paths, and spaced repetition.

The learning approach has three layers: explanation, structure, and recall.

Articles explain the topic. Roadmaps decide the order. Flashcards make the learner come back to the important details after the first reading.

Explanation

The source material is technical and detailed. The platform keeps that depth, but gives each lesson a clearer learning role:

  • introduce one mechanism or system boundary
  • connect the mechanism to previous lessons
  • show where the idea appears in transformer or agent implementations
  • extract review cards that can be answered without rereading the whole article

This is why lessons can be long, but cards are short. The article builds understanding. The review deck checks whether the understanding survives.

Structure

The archive contains many useful posts, but an archive is not a curriculum.

The platform turns the archive into ordered structures:

  • roadmaps for broad mastery
  • courses for guided sequences
  • skill paths for narrower goals
  • lesson pages for one concept at a time
  • challenge decks for review and recall

The same lesson can appear in more than one structure. For example, a KV cache lesson can belong to the Transformer Systems roadmap and also to an LLM inference optimization skill path.

Recall

The platform uses Anki-style thinking without requiring the learner to manage a separate deck.

Each lesson can produce flashcards. A card should target one thing:

  • a definition
  • a mechanism
  • a comparison
  • a formula
  • a failure mode
  • a trace through a system
  • a reason why an implementation choice exists

The dashboard then chooses review items from the roadmap and updates a recall score. The score is not meant to be a vanity metric. It is a signal that the learner can explain the topic again after time has passed.

The Basic Loop

  1. Choose a roadmap. Pick Transformer Systems or AI Agent Frameworks depending on the system you want to understand first.
  2. Read the next lesson. Work through the article, diagrams, examples, and equations.
  3. Answer the attached cards. Use the flashcard deck to check whether the key ideas are retrievable.
  4. Return through daily practice. Let the dashboard surface cards again until the concept is stable.

What Counts as Progress

Progress should eventually be tracked at several levels:

  • lesson read
  • card answered
  • card remembered after delay
  • challenge deck completed
  • roadmap node retained
  • roadmap recall score improved

The important distinction is between completion and retention. Completion says “I visited this lesson.” Retention says “I can still explain it.”