Choose a Roadmap

Start with a roadmap, not with the archive.

The archive is useful for search and preservation, but the roadmap decides what to learn next and what to review later.

Transformer Systems

Choose Transformer Systems if you want to understand the machinery behind modern LLMs.

This track is best when your questions look like:

• How does tokenization become vectors?
• What are Q, K, and V projections doing?
• Why do masks, RoPE, normalization, and FFNs appear in every model?
• How does decoding work after prefill?
• Why does serving an LLM become a memory, batching, and cache-management problem?

The track starts with the model internals and moves toward inference and serving.

AI Agent Frameworks

Choose AI Agent Frameworks if you want to understand how coding agents and tool-using frameworks are built.

This track is best when your questions look like:

• What is the agent loop?
• How do actions and observations move through the system?
• What does the controller own?
• Where does runtime execution happen?
• How are tool calls, memory, microagents, and session state coordinated?

The track starts with OpenHands-style core ideas and moves toward framework architecture.

If You Are Unsure

Use this rule:

• Pick Transformer Systems if you want to understand the model.
• Pick AI Agent Frameworks if you want to understand the system around the model.

Most serious AI engineering eventually needs both. Transformer Systems explains the behavior and constraints of the model. AI Agent Frameworks explains how an application organizes the model into tools, state, actions, observations, and execution.

Recommended First Session

1. Open the dashboard.
2. Choose one of the two roadmap buttons.
3. Read the first lesson in the roadmap.
4. Answer the attached flashcards.
5. Return to the dashboard the next day and review the selected cards.

The first session should be small. The product is designed around repeated contact with the material, not one long binge-reading session.