Thinking
Essays and courses on technology economics, AI operating models and engineering organisations.
Executive Thinking
Technology economics, AI operating models, and engineering organisations: how I think about running technology, not just how to build it.
How Private Equity Actually Works
Private equity firms don't pick stocks, they become the company for five to eight years. A plain-English walk through LPs, buyouts, the fund lifecycle, and how the fees actually work.
Read article →
Technical Debt Isn't a Feeling. It's a Carrying Cost.
Technical debt isn't unmeasurable, it's unmonetized. A framework that turns velocity tax, incident cost, and hiring drag into one carrying cost a CFO can fund.
Read article →
The Vendor Contract Red Flags Most Technical Leaders Miss
A fair price at signing doesn't make a fair contract. Auto-renewal windows, price-lock expiration, data portability, and change-of-control clauses decide what it actually costs.
Read article →Also: Technical Notes
Implementation-level engineering work: continuing technical depth, not the executive-thinking layer above.
Reading a Job Description Like a Due Diligence Document
A job posting is a pitch deck, not audited information. A requirement-by-requirement evidence framework, borrowed from M&A due diligence, for evaluating roles honestly.
Read article →
Claude Code Skills vs Subagents vs MCP Servers: When to Use Which
Skills, subagents, and MCP servers aren't tiers of the same capability - they solve three different context-cost problems. A primary-sourced decision framework.
Read article →
Postgres Just Became a Graph Database. Nobody Needs a Fourth One.
pgGraph adds bounded graph traversal to plain PostgreSQL tables, no migration, no new database required. Here is what changes, and what still does not.
Read article →Also: Technical Courses
Hands-on technical teaching on retrieval, agents, orchestration and evaluation, built from real production work (27 courses).
Adaptive RAG: Routing Between the Series' Five Strategies
A linear, one-concept-per-lesson path through Adaptive RAG, classifying each question and routing it to whichever of naive, hybrid, graph, corrective, or agentic retrieval actually fits, from a single-tier router to a complete FastAPI service wiring in the real implementations from courses 1-5.
View course →
Corrective RAG
A linear, one-concept-per-lesson path through Corrective RAG (CRAG), grading retrieved chunks for relevance and correcting a bad retrieval before it ever reaches generation, built from scratch with no framework, from your first relevance grade to a complete FastAPI corrective RAG service backed by ChromaDB.
View course →
Multimodal RAG
A linear, one-concept-per-lesson path through Multimodal RAG, retrieval-augmented generation extended to cover images as well as text, built from scratch with no framework, from your first multimodal generate_content call to a complete FastAPI service that returns cited answers grounded in both text and images.
View course →