This directory is reserved for runnable teaching crates that follow the lesson progression.
It is the executable companion layer for the course, not the canonical teaching layer.
lessons/stays canonical for authored teaching content.code/grows as modules earn real runnable examples.- Some topic directories may still be placeholders, but active topics should be real Cargo crates with tests.
When executable lessons start, each topic directory becomes a real Cargo crate.
Do not use loose standalone .rs files as the long-term structure.
- Active crates must compile, run tests, and have examples that match the lesson module they support.
- Planned crate directories should stay as roadmaps until the matching lesson module is authored.
- Crate READMEs should explain scope limits so learners do not mistake teaching code for production ML infrastructure.
| Topic | Status | Purpose |
|---|---|---|
| neuron | Active crate | Teaches one-neuron training, manual gradients, SGD, datasets, and a tiny next-token bridge. |
| mlp | Planned crate | Will become the small multi-layer network companion crate. |
| attention | Planned crate | Will become the explicit attention-mechanics companion crate. |
| transformer | Active crate | Real tested encoder-path teaching crate for the current advanced preview module. |