Ninety-seven essays, five technical articles, and four case studies — that's a lot. Here are four paths depending on where you are right now.
"I use Cursor or Claude Code daily and want to get better"
Start with Part 6 — The Developer as User. It's written for developers already using an AI assistant as a daily tool.
Then hit the essentials: essays #13, #52, #54, #65. Follow with the Cursor case study to understand how the tool you use every day was built.
"I'm building my first agentic system"
Start with Part 1 — Working with Agents to install the right mental models. Continue with Part 3 on building.
Then the technical articles: RAG and Evals. End with the WHOOP Coach case study — a mature example of RAG + evaluation architecture in production.
"I'm evaluating whether to invest in AI for my team"
Read the Klarna case study first — the most instructive counter-example, the triumph → cracks → correction arc. It's the case that protects you from the expensive mistakes.
Then essays #7 (Human in the Loop), #27 (Evals), and #53 (You Are Responsible). Finish with Duolingo to see what an AI-first transformation actually implies — content pipeline, workforce, subsumption risk.
"I'm debugging an agent that isn't working"
Essays #4, #22, #25, #34, #36 — the most practical on debugging. Then the Evals article to build a real safety net.
If the problem is architectural rather than tactical, circle back to Parts 3 and 4 of the essays.