Evidence and prediction
12 lessons · Starts with “Meet AI: Patterns, Predictions, and Decisions”
BuildA tiny predictor you can test
Open level AI From Zero to BuilderCurriculum atlas · L0–L15 · English v1
From first pattern to production-ready AI systems.
Each level makes one more part of the machine visible. Learn the idea, inspect the evidence, change something, explain what happened, then build the next artifact or system.
Start at Level 0 →- Levels
- 16
- Lessons
- 221
- Route
- L0 → L15
Build the foundations
Learn evidence, machine learning, neural networks, and reusable representations before language models enter the picture.
Trustworthy classical ML
18 lessons · Starts with “What Machine Learning Is Really Optimizing”
BuildA classifier you can audit
Open level Neural networks from first principles
15 lessons · Starts with “Why Neural Networks”
BuildA neural network from scratch
Open level Representations and transfer
14 lessons · Starts with “Representations: What Networks Learn”
BuildA transfer-learning experiment
Open level Build the language model
Turn text into tokens, attention, Transformers, and finally a tiny autoregressive language model you can inspect.
Text and tokenization
14 lessons · Starts with “Text Is Not Numbers Yet”
BuildA tokenizer you can inspect
Open level Attention and transformers
16 lessons · Starts with “Why Attention”
BuildAn attention probe you can explain
Open level Tiny language models
14 lessons · Starts with “Decoder-Only Language Models”
BuildA tiny language model
Open level Build with LLMs
Use modern LLMs deliberately through prompting, model adaptation, and evidence-backed retrieval.
Modern LLM behavior and prompting
13 lessons · Starts with “How Modern LLMs Behave”
BuildA reliable LLM workflow
Open level Fine-tuning and model adaptation
14 lessons · Starts with “When to Fine-Tune”
BuildA model-adaptation experiment with evidence
Open level Retrieval-augmented generation
14 lessons · Starts with “Why Retrieval-Augmented Generation”
BuildAn evidence-backed RAG system
Open level Build tools and agents
Move from tool calls to bounded agents and the runtime structures that make agent behavior testable and reliable.
Tool Use, Structured Workflows, and Multimodality
13 lessons · Starts with “LLMs That Use Tools”
BuildA tool-using assistant with validated actions
Open level Agents and bounded loops
13 lessons · Starts with “What Makes an Agent”
BuildA bounded agent with explicit stopping and approval
Open level Agent harnesses and runtime engineering
13 lessons · Starts with “Agent Harnesses and Runtime Structure”
BuildA durable agent harness
Open level Build AI systems
Connect interoperable agents, production serving, evaluation, safety, security, and release evidence.
Interoperability, MCP/A2A, and Multi-Agent Systems
12 lessons · Starts with “Why Interoperability Matters”
BuildAn interoperable agent system
Open level Production serving and operations
13 lessons · Starts with “From Notebook to Service”
BuildA production AI service you can operate
Open level Evaluation, safety, and security
13 lessons · Starts with “Evaluation as a System”
BuildAn AI product you can evaluate and defend
Open level