kaizen

Apply the YAGNI principle: implement only current requirements with the simplest working solution. Add complexity only when a real requirement demands it, and wait for three or more similar cases before creating an abstraction.

Use discriminated union types so each state carries only its valid associated data, such as an order that can only have a tracking number when shipped. Branded types and non-empty array types also push validation to compile time.

Refactor incrementally whenever the existing code works: fix one issue at a time, keep tests passing, and commit after each improvement. Big-bang rewrites are error-prone and hard to verify compared to small compounding changes.

Validate once at the system boundary, such as request handlers or configuration loading, before any value is used. After boundary validation, pass validated or branded types inward so inner functions can trust their inputs.

Optimization is premature when done without profiling evidence of a bottleneck. Ship the straightforward implementation first, measure performance, and optimize only the measured hot paths while documenting why the change was needed.