How we built Wynna
Each part of Wynna below follows a finding from learning research. This page sets out what each finding says, what Wynna does with it, and where you can read the study for yourself.
Questions come back before your child forgets
Memory fades fast at first, then more slowly. A review that arrives just before something is forgotten makes it last longer, and each review can come later than the one before.
In Wynna: every question keeps its own review date. After a right answer, your child says how sure they were. A guess comes back the next day, "think so" after two days and "certain" after five, and each right answer after that pushes the next review further away. An answer that needed help comes back within a day. A wrong answer comes back about five minutes later, while the explanation is still fresh.
Read the research
- The first study of how quickly we forget (Ebbinghaus, 1885; English edition 1913)
- A large review showing that spreading practice out works better than cramming (Cepeda and others, 2006)
- Spacing out practice tests works better than doing them all at once (Latimier, Peyre and Ramus, 2020)
- Young children learn science ideas better from spaced lessons (Vlach and Sandhofer, 2012)
- How spaced review can be used in schools (Kang, 2016)
- A recent review of spacing practice out and answering from memory (Carpenter, Pan and Butler, 2022)
Your child answers, not just reads
Pulling an answer out of memory strengthens it far more than reading the same page again. A large review of study techniques rated practice testing as one of only two with high usefulness. The other was spaced review. A 2022 brain-imaging study of secondary school students found the same: words learned by answering from memory were remembered better a week later, and recalling them used more of the brain areas involved in memory.
In Wynna: every question asks your child to produce an answer: type it, fill a gap or pick one of four options. Wynna records whether each answer came from memory or after help, and an answer from memory counts for more.
Read the research
- Testing yourself helps you remember more than rereading (Roediger and Karpicke, 2006)
- Ten common study techniques compared (Dunlosky and others, 2013)
- Quizzes help pupils learn in real classrooms: a review of 222 studies (Yang and others, 2021)
- Answering from memory helps pupils in schools, again and again (Agarwal, Nunes and Blunt, 2021)
- A brain-imaging study of secondary school students learning by answering from memory (Wiklund-Hörnqvist and others, 2022)
- What brain research says about why testing yourself works (van den Broek and others, 2016)
Hints first, the answer last
Children learn most from support pitched just beyond what they can do alone, which is then taken away as they improve. Researchers call this scaffolding. Recent studies of AI point the same way. When AI gives students the answers, homework scores go up but test results fall. In a 2025 trial with nearly 1,000 secondary school maths students, an AI tutor that gave hints instead of answers avoided most of that harm.
In Wynna: help comes one step at a time, starting with a hint and moving to a stronger clue (on a multiple-choice question, fewer options to choose from). The AI helper is told never to give a direct answer. It asks a guiding question first, then offers a clue or sentence starter, and only then names the mix-up and asks your child to apply the idea. After three tries Wynna shows the answer, and the question returns later for another try.
Read the research
- The study that first described helping a child step by step as "scaffolding" (Wood, Bruner and Ross, 1976)
- A review of 144 studies of software that guides students step by step (Belland and others, 2017)
- Online maths homework with hints and feedback raised test scores in 43 schools (Roschelle and others, 2016)
- AI that gives answers can harm learning, and AI that gives hints avoids most of the harm (Bastani and others, 2025)
- Secondary pupils who handed homework to AI did worse in exams (Strömberg, Lei and Wu, 2026, not yet peer-reviewed)
Wynna spots the mix-up behind a wrong answer
Children often hold a wrong idea that makes sense to them, such as "heavier things fall faster". Hearing the right answer seldom shifts it. The wrong idea has to be brought out, shown not to work and replaced.
In Wynna: when Wynna builds a revision plan, it lists the mix-ups children commonly make on each topic. When a written answer is marked, the marker names the wrong idea behind it if there is one, and Wynna counts each one. If the same misunderstanding keeps coming back, your dashboard tells you and says how many times it has happened.
Read the research
Half-right answers still count
Feedback works when it shows the gap between what a child wrote and what was needed, and how to close it.
In Wynna: written answers are marked correct, partial or incorrect. Partial means the right idea is there but something is missing. Wynna records what your child got right and what was left out, the question comes back sooner than a confident right answer would, and your dashboard shows what keeps getting left out. If Wynna is not confident how to mark an answer, it asks for a review instead of marking it wrong.
Read the research
- How feedback while learning improves results (Black and Wiliam, 1998)
- Feedback that shows the gap between an answer and what was needed (Sadler, 1989)
- A review of 435 studies on how feedback helps pupils learn (Wisniewski, Zierer and Hattie, 2020)
- Mistakes help learning when they are corrected (Metcalfe, 2017)
Topics are mixed, like in a real test
Practising one kind of question many times in a row feels productive. Mixing topics tends to work better over time, because your child has to work out what kind of question each one is, as in a test.
In Wynna: Topic training works on one topic, which helps when something is new. Mixed practice shuffles questions from across your child's topics. Practise next draws on the three topics with the most wrong answers.
Read the research
- Mixing up maths problems helps pupils learn (Rohrer and Taylor, 2007)
- Seeing examples mixed together helps people tell things apart (Kornell and Bjork, 2008)
- A trial in 54 maths classes: mixed practice raised test scores a month later (Rohrer and others, 2020)
- When mixing helps most, and when it does not: a review of 59 studies (Brunmair and Richter, 2019)
No topic is ticked off too soon
Moving on before a topic is understood leaves gaps that grow. Mastery learning keeps the standard fixed and lets the time vary, and a large review of mastery programmes found they raised results.
In Wynna: a topic is only called Secure after at least four marked answers with a score of 80% or more, where an answer from memory counts twice as much as one given with help or one that was nearly right. When Wynna suggests what to do next, it favours overdue reviews, recent mistakes, weaker topics and topics on a test coming up soon.
Read the research
Help with writing longer answers
A longer answer is hard to write without a plan for the paragraph. In a major review of writing instruction, teaching a clear writing strategy step by step was among the approaches with the largest effects.
In Wynna: each writing task comes with a structure that suits the subject: PEE (Point, Evidence, Explain) for humanities, CER (Claim, Evidence, Reasoning) for science, SOLVE for maths and DESCRIBE for creative writing. The coach praises one thing your child did well, finds the weakest or missing part of the structure and suggests a sentence starter for it. It is told never to write the sentence for your child.
Read the research
Made from your child's own schoolwork
Practice helps most with a test when it covers what that test will ask. A review of 50 studies of tutoring software found much bigger gains on tests that matched what pupils had practised than on general tests.
In Wynna: you upload notes, worksheets, past papers or photos of textbook pages. Wynna builds topics only from what that material covers and never adds one it doesn't. When a worksheet asks questions without giving the answers, Wynna writes the standard answers. You check the topics and questions, and nothing reaches your child until you approve it.
Read the research
Characters that make your child want to come back
Rewards can strengthen motivation or weaken it. They help when they support a child's sense of choice, of getting better and of belonging. Game features work best when they are part of the learning itself, not a prize added on at the end.
In Wynna: your child picks a world, Heroes of the Forge, Mermaids of the Sea or the Haunted House, and rescues new characters by practising. Characters level up as your child answers, fastest from memory. Each character has its own power, such as taking away one wrong option, and the power gets stronger with each level. No power gives away the answer. There are no timers, no adverts and no public profiles.
Read the research
- What keeps people motivated: choice, progress and belonging (Deci and Ryan, 2000)
- Rewards given for doing a task can weaken the wish to do it (Deci, Koestner and Ryan, 1999)
- Having a choice increases motivation (Patall, Cooper and Robinson, 2008)
- Children aged 7 to 11 learned more when the maths was the game itself (Habgood and Ainsworth, 2011)
- Games help learning, and better-designed games help more: a review of 57 comparisons (Clark, Tanner-Smith and Killingsworth, 2016)
Your dashboard tells you what to do next
Research on parents and learning separates being involved with school from being engaged with a child's learning at home. It finds that the second makes the bigger difference.
In Wynna: the dashboard works out which topics are weakest and why. If the same misunderstanding keeps appearing, or your child is leaning on hints instead of answering from memory, it tells you. The AI revision coach turns these signals into two or three specific things to try this week, such as a question to ask at dinner.
Read the research
Safety has its own page: how content is checked, what happens to your child's data and what you control. Safety · How it works