AI Deep Dive
Practitioner-level AI — architecture, training, agent design, and the frontier for those who want the full picture
Read it in the library →The AI Story
The technical and historical arc of AI — paradigm shifts, scaling laws, the compute stack, and the economics that make the current era structurally different from all previous ones.
Generative AI & LLMs
Transformer internals, MoE architectures, training and post-training pipelines, tokenisation, inference optimisation, hallucination mechanics, RAG engineering, and the adaptation decision framework.
- The Transformer Architecture
- Beyond Vanilla Transformers: MoE and Friends
- Pre-training & Fine-tuning Pipeline
- Tokenisation In Depth
- Inference & Latency Optimisation
- Hallucination: Mechanisms & Mitigations
- RAG Architecture In Depth
- Fine-Tuning vs RAG vs Prompting
- LLM Governance & Safety — When Each Layer Applies
Agentic AI
The architectural overview of agentic systems: architecture patterns, orchestration frameworks, MCP and A2A protocols, the attack surface, production reliability, context engineering, and evals. This module orients you across the whole shape of the problem. Three courses on this site — Agent Engineering, Securing AI Systems, and Does Your AI Actually Work? — take these same topics to production depth, and each lesson below says which one picks up where it stops.
AI Frontiers
The current edge of the field — RL-trained reasoning, multimodal understanding and generation, alignment as an engineering discipline, the regulatory landscape as it actually stands, and where the next few years are heading.
Every lesson is free, with no sign-up. Reading happens in the library, where your progress is saved on your device.