Path · Go deeper · 6 steps · 30 lessons · 64 min
I want to understand how it works
How models learn, why they answer the way they do, how they are built into products, and the big open questions.
Step 1 · 5 lessons
How models learn
Tokens, attention, training and the overfitting trap.
Step 2 · 5 lessons
Why answers behave as they do
Randomness, memory, reasoning and flattery.
- L012Temperature: safe or creative2 min
- L047Why the same question gets different answers2 min
- L068Context window vs memory2 min
- L033How chatbots "reason" step by step2 min
- L053Sycophancy: when AI agrees too much2 min
Step 3 · 5 lessons
Shaping a model
Feedback, fine-tuning, distillation and experts.
Step 4 · 5 lessons
Building blocks of AI apps
Embeddings, RAG, agents, MCP and APIs.
- L024Embeddings: meaning as a map2 min
- L025RAG: give AI your documents2 min
- L026AI agents: AI that takes action2 min
- L027MCP: a universal plug for AI1 min
- L050What is an API?2 min
Step 5 · 5 lessons
Safety inside the system
Injection, jailbreaks, guardrails and what tests miss.
- L052Prompt injection: hidden orders in the text3 min
- L051Jailbreaks: talking AI past its rules2 min
- L169Guardrails on the way in2 min
- L170Guardrails on the way out2 min
- L066How AI is tested, and what tests miss2 min
Step 6 · 5 lessons
The big picture
Black boxes, AGI, energy and open models.
- L065The black box problem2 min
- L034What AGI means2 min
- L035Is AGI close? Both sides3 min
- L069AI's electricity and water3 min
- L040Open vs closed models2 min
Finish the path
Take the quiz, get the certificate
Score 80% or more on the quiz to unlock a certificate with your name.