The journal
Notes on the exam, the method, and the model.
A small, slow blog. Long-form on the Claude Architect exam, opinion pieces on adaptive learning, and the occasional public deep dive into a single knowledge point.
Anatomy of an agentic AI system: a reference architecture for builders
A long-form walk through the nine layers every production agentic system shares, the tools that won at each layer in 2025 and 2026, and a sober look at what is genuinely new versus what is a relabelled prototype.
Read the pieceThe Claude Certified Architect exam: every domain, every task statement, scored by difficulty
A definitive, opinionated walk-through of all five exam domains and 30 task statements, with our internal difficulty rating and the concepts that catch most people out.
Bayesian Knowledge Tracing, explained without the maths
Why a forty-year-old algorithm is still the right way to track what you actually know, and how we use it to retire concepts at 0.90 mastery probability instead of 'you saw it three times'.
Why Socratic tutoring beats flashcards for technical certifications
Recall is cheap, transfer is expensive. A short essay, with one paper to back it up, on why an AI tutor that refuses to give you the answer is the right tool for an exam built on scenarios.
Prompt hierarchy: what `system` actually does (KP-042 explained)
A public sample of a knowledge-point deep dive. Why system prompts are not just 'instructions that come first', and why getting this wrong wastes most of your context window.
More posts arrive as the question bank grows. One per knowledge point is the long-term target.
5 of ~175