It Ain't Broke: Why Software Fundamentals Matter More Than Ever — Matt Pocock, AI Hero @mattpocockuk

Matt Pocock argues AI coding tools make software fundamentals more critical, not obsolete, and shares a concrete process for shipping quality code with AI agents.

• Matt Pocock’s ‘grill me’ skill (13,000 GitHub stars) forces AI to ask 40–100 questions before writing any code, establishing a shared design concept.
• ‘Specs to code’ movement — writing a spec and repeatedly recompiling without reading code — produces progressively worse output each iteration.
• Bad code is the most expensive it has ever been: AI performs well in clean codebases and degrades sharply in tangled ones, amplifying existing quality gaps.
• Ubiquitous language (from Domain-Driven Design) stored as a markdown file cuts AI verbosity and aligns implementation with planned terminology.
• Deep modules (few large units with simple interfaces, per John Ousterhout) make codebases easier for AI to navigate and easier to test at boundaries.
• TDD forces AI to take small steps; the LLM’s default is to generate large code dumps and only then type-check, which Pragmatic Programmer calls ‘outrunning your headlights.’
• Pocock’s recommended mental model: treat AI as a tactical sergeant executing changes; the engineer remains the strategic designer investing in system design daily (Kent Beck).

2026-04-23 · Watch on YouTube