The stacks.

111 pods on 17 shelves — browse a shelf sideways, or type to cut straight through all of them.

Generative AI Fundamentals

3 pods

AI vs ML vs DL vs GenAI

Stage 0 · Ground — the maths and the map

Recurrent Neural Networks from Scratch

Stage 0 · Ground — the maths and the map

What is a VAE? The Smooth Bridge Between Math and Generative AI

Stage 0 · Ground — the maths and the map

Foundations of Machine Learning

11 pods

Vector, Basis and Linear Transformation

Stage 0 · Ground — the maths and the map

Importance Eigen Values and Eigen Vectors in Machine Learning

Stage 0 · Ground — the maths and the map

Probability Distribution in Machine Learning

Stage 0 · Ground — the maths and the map

All About LSTM

Stage 0 · Ground — the maths and the map

Statistics in Machine Learning

Stage 0 · Ground — the maths and the map

Gradient Descent and Why Learning Rates Misbehave

Stage 0 · Ground — the maths and the map

Backpropagation from Scratch: Computing Every Gradient by Hand

Stage 0 · Ground — the maths and the map

The Bias-Variance Tradeoff, Diagnosed on Real Data

Stage 0 · Ground — the maths and the map

Convolution from Scratch: How a CNN Actually Sees an Image

Stage 0 · Ground — the maths and the map

Regularization from Scratch: L2, L1 and Dropout — How to Cure an Overfit Model

Stage 0 · Ground — the maths and the map

Softmax and Cross-Entropy from Scratch: Why the Gradient Is Just Prediction minus Truth

Stage 0 · Ground — the maths and the map

Build LLM from Scratch

14 pods

interview answer

Foundations of Language Modeling

Stage 1 · Language — how a model reads and writes

interview answer

Self-Attention from First Principles

Stage 1 · Language — how a model reads and writes

interview answer

Building a GPT-style Model from Scratch

Stage 1 · Language — how a model reads and writes

Understanding BERT from Scratch

Stage 1 · Language — how a model reads and writes

Training Pipeline Engineering

Stage 1 · Language — how a model reads and writes

interview answer

Inference & Scaling

Stage 1 · Language — how a model reads and writes

Beyond the Residual Stream: Building Hyper-Connections and mHC

Stage 1 · Language — how a model reads and writes

Tokenization from Scratch: Building Byte-Pair Encoding by Hand

Stage 1 · Language — how a model reads and writes

Mixture of Experts (MoE) from Scratch

Stage 2 · Modern architecture — what changed since GPT

Sparse Mixture-of-Experts from Scratch

Stage 2 · Modern architecture — what changed since GPT

interview answer

Understanding Multi-Head Latent Attention (MLA) from Scratch

Stage 2 · Modern architecture — what changed since GPT

Linear Attention and the Delta Rule: Building Kimi Delta Attention

Stage 2 · Modern architecture — what changed since GPT

Sparse Attention for Million-Token Context

Stage 2 · Modern architecture — what changed since GPT

Build Qwen3.5 from Scratch

Stage 2 · Modern architecture — what changed since GPT

AI in Production

7 pods

interview answer

KV Cache: The necessary evil of LLM Inference

Stage 1 · Language — how a model reads and writes

Vectorless, Reasoning-based RAG

Stage 5 · Working with the model — context and agents

Data Engineering & Feature Stores

Stage 6 · Shipping it — production

MLOps & CI/CD for ML

Stage 6 · Shipping it — production

Model Deployment & Serving

Stage 6 · Shipping it — production

Monitoring & Observability

Stage 6 · Shipping it — production

Model Governance & Reliability

Stage 6 · Shipping it — production

GPU Parallelism for Large Language Models

10 pods

Intro to GPUs and GPU Parallelism for LLMs

Stage 3 · Making it run — GPUs and efficiency

GPU Memory — The Four Major Components

Stage 3 · Making it run — GPUs and efficiency

Activation Recomputation, Gradient Accumulation, and Data Parallelism

Stage 3 · Making it run — GPUs and efficiency

Ring-AllReduce, Choosing Batch Size, and TensorBoard GPU Profiling

Stage 3 · Making it run — GPUs and efficiency

ZeRO-1, 2, 3 — Zero Redundancy Optimizer

Stage 3 · Making it run — GPUs and efficiency

Tensor Parallelism

Stage 3 · Making it run — GPUs and efficiency

Sequence Parallelism

Stage 3 · Making it run — GPUs and efficiency

Context Parallelism

Stage 3 · Making it run — GPUs and efficiency

Pipeline Parallelism

Stage 3 · Making it run — GPUs and efficiency

Expert Parallelism

Stage 3 · Making it run — GPUs and efficiency

Quantization from Scratch

2 pods

TurboQuant: How Google Compressed AI Memory by 6x Without Losing a Single Answer

Stage 3 · Making it run — GPUs and efficiency

MXFP4 Quantization-Aware Training from Scratch

Stage 3 · Making it run — GPUs and efficiency

RL From Scratch

12 pods

Basics of Reinforcement Learning

Stage 4 · Teaching it to decide — reinforcement learning

Value Functions and Q-Learning

Stage 4 · Teaching it to decide — reinforcement learning

Building DQN Atari Agents: When Q-Learning Learned to See

Stage 4 · Teaching it to decide — reinforcement learning

Policy Gradient Methods: Teaching Your Agent to Climb Mountains

Stage 4 · Teaching it to decide — reinforcement learning

RLHF Theory and Implementation: Teaching Machines to Learn from Human Preferences

Stage 4 · Teaching it to decide — reinforcement learning

Group-Relative Policy Optimization (GRPO) -- From Scratch

Stage 4 · Teaching it to decide — reinforcement learning

Building a Reasoning Model from Scratch

Stage 4 · Teaching it to decide — reinforcement learning

Variance-Aware Advantages for Reasoning RL

Stage 4 · Teaching it to decide — reinforcement learning

Reinforcement Learning with Language Feedback

Stage 4 · Teaching it to decide — reinforcement learning

Mini-SWE-RL: Teaching a Small Language Model to Fix Bugs with Reinforcement Learning

Stage 4 · Teaching it to decide — reinforcement learning

OpenClaw-RL: Personalizing AI Agents from Conversation Feedback

Stage 4 · Teaching it to decide — reinforcement learning

Gradient-Aligned Rewards: A Dense Reward that Beats GRPO's Right/Wrong Signal (Build GAR from Scratch)

Stage 4 · Teaching it to decide — reinforcement learning

Context Engineering from Scratch

6 pods

interview answer

Context Engineering for LLMs

Stage 5 · Working with the model — context and agents

Prompt Design Principles

Stage 5 · Working with the model — context and agents

interview answer

Retrieval-Augmented Generation (RAG) Systems

Stage 5 · Working with the model — context and agents

Memory Architectures for LLM Applications

Stage 5 · Working with the model — context and agents

Context Optimization & Evaluation

Stage 5 · Working with the model — context and agents

Build Your Own LLM Wiki from Scratch

Stage 5 · Working with the model — context and agents

AI Agents from Scratch

11 pods

interview answer

What Is an AI Agent?

Stage 5 · Working with the model — context and agents

The ReAct Policy: Reasoning + Acting

Stage 5 · Working with the model — context and agents

interview answer

Tool Use & Function Calling

Stage 5 · Working with the model — context and agents

Memory & State

Stage 5 · Working with the model — context and agents

Planning & Self-Reflection

Stage 5 · Working with the model — context and agents

Agentic RAG

Stage 5 · Working with the model — context and agents

Multi-Agent Systems

Stage 5 · Working with the model — context and agents

Evaluation, Observability & Guardrails

Stage 5 · Working with the model — context and agents

LangGraph vs Claude Agent SDK

Stage 5 · Working with the model — context and agents

Capstone: Build & Deploy Your Own Agent

Stage 5 · Working with the model — context and agents

Learned Model Orchestration from Scratch: Building a Router that Beats Bigger Models

Stage 5 · Working with the model — context and agents

VLMs from Scratch

6 pods

Vision Encoders: How Machines Learned to See — From Convolutions to Vision Transformers

Stage 7 · Seeing — vision and multimodality

Vision Transformers from Scratch: How Treating Images as Sentences Changed Computer Vision

Stage 7 · Seeing — vision and multimodality

Contrastive Pretraining (CLIP-style): Teaching Machines to See and Read at the Same Time

Stage 7 · Seeing — vision and multimodality

Cross-Attention & Token Alignment: How Vision-Language Models Learn to See and Speak

Stage 7 · Seeing — vision and multimodality

Multimodal Fusion Architectures: How AI Learns to See, Read, and Listen — All at Once

Stage 7 · Seeing — vision and multimodality

Multimodal Instruction Tuning: Teaching Language Models to See and Think

Stage 7 · Seeing — vision and multimodality

Principles of Diffusion Models

6 pods

Variational Autoencoders From Scratch

Stage 8 · Generating — diffusion

Energy-Based Models and the Score Function

Stage 8 · Generating — diffusion

Denoising Score Matching

Stage 8 · Generating — diffusion

Noise Conditioned Score Networks (NCSN)

Stage 8 · Generating — diffusion

Denoising Diffusion Probabilistic Models (DDPM)

Stage 8 · Generating — diffusion

From Still to Motion: How Diffusion Models Learned to Generate Videos

Stage 8 · Generating — diffusion

Build Diffusion LLM from Scratch

1 pod

Diffusion LLMs from Scratch

Stage 8 · Generating — diffusion

Build a World Model from Scratch

6 pods

What Is a World Model, Really?

Stage 9 · Acting in the world — robotics

The World Modeler's Toolkit

Stage 9 · Acting in the world — robotics

Your First World Model — MiniPong

Stage 9 · Acting in the world — robotics

Dreams That Last — the RSSM on a Real Robot

Stage 9 · Acting in the world — robotics

A Vector or a Vocabulary? — IRIS and Discrete World Models

Stage 9 · Acting in the world — robotics

The Journey of a Token Through a World Model

Stage 9 · Acting in the world — robotics

Modern Robot Learning

8 pods

Understanding World Models from Scratch

Stage 9 · Acting in the world — robotics

Understanding JEPA from Scratch

Stage 9 · Acting in the world — robotics

What are Vision Language Action Models?

Stage 9 · Acting in the world — robotics

Action Chunking Transfomers

Stage 9 · Acting in the world — robotics

Diffusion Policy for Robotics

Stage 9 · Acting in the world — robotics

Pi0 - Our first Vision Language Action Model

Stage 9 · Acting in the world — robotics

SmolVLA

Stage 9 · Acting in the world — robotics

World Models Inside a Robot Policy: VLAs that Dream

Stage 9 · Acting in the world — robotics

VLAs for Autonomous Driving

1 pod

Vision-Language-Action Models for Autonomous Driving: From Pixels and Words to Steering Wheels

Stage 9 · Acting in the world — robotics

Tiny Recursive Models

1 pod

Tiny Recursive Models: Less is More for AI Reasoning

Stage 10 · Frontier

Claude Certified Architect Prep

6 pods

Agentic Architecture & Orchestration

Side track — certification

Certification Practice Exam

Side track — certification

Claude Code Configuration & Workflows

Side track — certification

Context Management & Reliability

Side track — certification

Prompt Engineering & Structured Output

Side track — certification

Tool Design & MCP Integration

Side track — certification