
Old Books, Leading Words, and DDD

AI Skills with Matt Pocock
Topic 9 of 20
Published Sep 17, 2026 · Clip 50:52–1:01:08 (10:16)
Matt recounts expecting to throw out old practices with AI, but spec-driven development gave worse results than hand-coding because agents churned out garbage without good feedback loops. Opening The Pragmatic Programmer, he found software entropy, tracer bullets, and vertical slices that described his exact problems — and discovered 'leading words': repeating concepts from the agent's training prior (like tracer bullets) changes agent behavior. He mined books like Ousterhout's A Philosophy of Software Design (deep modules) and turned to Eric Evans' DDD and ubiquitous language, creating a domain modeling skill where agent-coined terms like 'materialization cascade' make communication with the agent dramatically more efficient. He concludes that clean codebases matter more than ever because agents start fresh every session like the protagonist of Memento.