AI and Software Testing - Evil Tester meets Tech League

AI and software testing intersect when code generation accelerates dramatically, forcing teams to adapt how they validate behavior, maintain structure, and gather useful information. Alan Richardson, interviewed by Toby Sears and Christian Fisher, describes practical shifts in workflow that arise when AI produces large volumes of code quickly. The discussion centers on keeping architecture and requirements explicit so that both generation and checking remain controllable.

Testing AI-generated code

Richardson works in small increments rather than long agent chains. He prompts for code in limited chunks, then immediately reviews the resulting tests for coverage and level before expanding further. When the AI struggles to produce clear tests, the underlying code usually lacks good structure. The remedy is to spend extra prompt effort on architecture first—small classes, explicit interfaces, clear separation of concerns—so the generated tests become easier to read and maintain. This approach treats testing as an integrated part of programming rather than a separate later activity.

Structuring tests to match development patterns