The Importance of Thin Slicing in Coding
Why did I anticipate the need for thin slicing, planning, coding to behavioral expectations, verifications checks, and continuous improvement loops in agentic coding? Because that’s how adaptive processes have operated for decades.
There is a continuum from human directed to autonomous that look very similar structured on a traditional plan do check act cycle. This has existed since Toyota Production System for industrial design and manufacturing. It contains analyze and plan, build, verify completeness, then learn and adapt to improve next time.
In software, the practice is to break work down into thin vertical slices, develop against behavioral expectations test first, incorporate manual and automated verification and then retrospection to generate continuous delivery and improvement.
The same challenges in communication and shared understanding what is to be built and how best to build it exist but by driving to small work in progress and clear behavioral definitions of success, we can first own the code that we write with agents, THEN as we achieve fluency and the costs and capabilities allow for it, cede more and more autonomy to the agents depending on the value proposition and risk profile of the application.
