Articles and Lessons
Edit: src/content/docs/learning-materials/articles-and-lessons.mdHow long-form posts become structured learning units.
Articles and lessons are the main explanation layer of ML Learning Lab.
The original posts remain useful because they preserve the source material. Lessons make that material usable in a curriculum.
Articles
Articles are the long-form technical source.
They can contain:
- derivations
- diagrams
- architecture walkthroughs
- implementation details
- comparisons between techniques
- notes from papers or source code
The existing Transformer Series and OpenHands posts are preserved under the old routes while the platform is being reshaped.
Lessons
Lessons are the learning-platform version of the articles.
A lesson should add structure around the source:
- domain and track
- module placement
- difficulty
- estimated time
- prerequisites
- related lessons
- flashcard deck
- roadmap node
The lesson page should answer: “Where does this topic fit, and what should I retain from it?”
Lesson Anatomy
A strong lesson has these parts:
- Context - what problem the concept solves.
- Mechanism - how the concept works.
- System role - where it appears in a transformer or agent framework.
- Failure modes - common misunderstandings or implementation traps.
- Recall targets - the facts and mechanisms worth turning into flashcards.
How Lessons Connect to Cards
Cards should be extracted from the lesson, not invented separately.
For example, a lesson on attention masks can produce cards about:
- padding masks
- causal masks
- decoder masks
- training-time vs inference-time masking
- masked softmax behavior
- bugs caused by shape mismatch or incorrect broadcast
The card deck is therefore a compressed memory model of the lesson.