Built on the specification
“I understood it in the lesson. Then I sat down with the homework at nine at night and had no idea where to start.”GCSE Maths student
The problem
Momentum stops, and the topic quietly gets avoided.
The solution
Works through the method with them and explains every wrong answer, rather than just marking it red.
The problem
Confidence in the room, then a shock in the exam hall.
The solution
Genuine AQA and Edexcel papers, marked against the board's own published mark scheme.
The problem
The same method error is rehearsed until it's a habit.
The solution
Handwritten answers are read and assessed step by step, so method marks are picked up as an examiner would award them.
The problem
Revision goes to whatever feels easiest, not what's weakest.
The solution
Every mock maps to a 9–1 or A*–U estimate, with mastery tracked per subtopic so weak areas are named, not guessed.
Exam-spec alignment
The fastest way to lose a student's trust is a practice question that would never appear on their paper, or a mark scheme that's simply wrong. Both are engineering problems, and we treated them that way.
Every question template is verified with SymPy, a symbolic mathematics engine, before it reaches a student. If the template's answer doesn't hold symbolically, it never ships — so a student is never marked wrong by a mark scheme that was wrong to begin with.
Mock exams use genuine AQA and Edexcel past papers. Each question is marked against the board's own published mark scheme — extracted once, cached, and applied per question, with a grade estimate on the 9–1 and A*–U scales.
100+ papers · 2018–2024Questions mirror real exam phrasing, command words and mark allocation, including multi-part structures. Multiple-choice distractors are built from the misconceptions students actually hold, not from arbitrary wrong numbers.
Content is structured subject → chapter → subtopic → template, each tagged by skill and learning objective. That's what makes mastery tracking meaningful: progress is measured against the specification, not a vague subject average.
AQA · Edexcel · OCRWhy it's different
A question bank can't explain itself. A chatbot can't mark a real exam paper. A past-paper site can't tell a tutor who's fallen behind. Each tool solves its slice and hands the problem on. PeddyLoop closes the circuit.
An AI tutor works through the problem with them — explaining the method, not revealing the answer — and adapts difficulty as they improve.
Handwritten working is photographed and marked against the real mark scheme, per question, the way an examiner would award the marks.
Tutors and parents get the same live picture — mastery by subtopic, session history, and reports written for them, not extracted by them.
Each stage feeds the next: what the student struggles with shapes what they're taught tomorrow, and what the tutor sees on Monday morning.