Skip to content
All case studies

Personal project · Learning tools

AI Exam Prep

A home-learning system built around one child, genuine question patterns and what worked in practice.

Period
2026
Role
Product, learning design and implementation
Platform
Browser-based learning system
Outcome
22 games + generated FMC practice

Problem

Generic worksheets repeated the same questions and could not focus closely enough on the skills my son needed to practise.

Solution

Twenty-two short games plus fresh FMC-style maths tests, with explanations available only after marking.

Impact

A reusable practice system that could change quickly whenever a question type or interaction was not working.
FMC practiceQuestion 14 of 2021:36

A clock gains 5 minutes every hour. If it is correct at 9:00, what will it show at 3:00?

  1. A3:05
  2. B3:25
  3. C3:30
  4. D3:35
  5. E4:00
No hints during the test45 minutes

The problem

My son Krish needed more practice with the reasoning behind the 11+ and the First Mathematics Challenge. Past papers set the standard, but repeating them would eventually test memory rather than understanding.

The solution

I built 22 short browser games covering verbal, spatial, logic and maths skills, plus an FMC practice lab that generates fresh twenty-question tests with a forty-five-minute timer.

Keep the method; change the question

I studied the 2019–2025 papers and grouped questions by the method they test. Each group defines the maths, sensible number ranges, the answer and believable distractors. Names, values, diagrams and answer positions can change while the tested skill stays consistent.

The evidence

The live test hides hints and working until it has been marked. The final six questions come from a harder pool containing multi-step balances, shared-purchase equations, arrangements, digit problems and Venn reasoning.

Where AI helped

AI sped up pattern analysis, interface variations and question production. I chose the skills, set the rules and checked the answers. When Krish found the final questions too easy or an interaction confusing, I changed the product.

The impact

The result is tailored practice that can change quickly, without handing the learning decisions or quality checks to a model.

The result

Fresh practice without sacrificing question quality.

Start a conversation
Next case studyForgeKit