Reverse-Engineering Assessments with AI2026-04-10

I've been thinking: if I want to recreate an assessment product like SBTI, do I still need to manually map out and code the questionnaire logic, scoring rules, and result algorithms?

My hypothesis: Since the principles of these tests are widely available online, could I just prompt an AI Agent to autonomously crawl wikis, papers, and discussions, and use LLM reasoning to reverse-engineer the entire scoring matrix and decision tree?

Taking it a step further—not just inferring the logic, but having the AI directly generate the frontend questionnaire and backend calculation code based on those inferred rules. In this workflow, the human only sets the goal ('clone this test for me'), while AI handles the full loop of 'knowledge retrieval -> logic reverse-engineering -> code implementation'. If this works, it would be a fantastic use case for AI replacing traditional business logic mapping.