Our Theory of Learning: The Learning Edge, The Climb, and The Muscle.

Most education companies write a "theory of learning" that turns out to be a marketing brochure dressed up with vocabulary. We are not going to do that. The three pillars below are not slogans. They are conclusions drawn directly from fifty years of replicated cognitive science research — and every choice Knowly makes about how to teach your child traces back to one of them.

You can read the full evidence base on our research page. This essay is the connective tissue: why these three, why now, and what they mean concretely for what your child does inside Knowly every morning.

Pillar one: The Learning Edge

Every child has an edge. Every lesson finds it.

There is a point — different for every child, in every subject, on every day — where the material is just hard enough that the child cannot yet do it alone, but close enough that with effort they can. This is the only zone where real learning happens. Vygotsky called it the zone of proximal development in 1934. In 2019, Wilson and colleagues at Princeton proved it mathematically in Nature Communications: learning rate is maximized when a student succeeds on approximately 85% of first attempts. Not 100%. Not 50%. Eighty-five.

Too easy and the child confirms what they already know. The brain does not strengthen. Too hard and the child shuts down. The brain also does not strengthen. The narrow band between those is the Learning Edge. And it moves — every day, every concept, every child.

A teacher with twenty-two students cannot find twenty-two moving edges in real time. The arithmetic does not allow it. The teacher can do an excellent job of finding the average edge for the room — but no child in the room is the average. So most children, every day, in every subject, are working at a difficulty level that is either too easy (no growth) or too hard (shutdown). Across a year, that compounds. Across a school career, it explains the gap between elite tutored students and the rest.

Knowly finds the edge for every child, every day. On day one, the platform places each new child at their actual current level — not their grade label, not their last report card — through a diagnostic assessment. From there, it tracks every response for where in the reasoning chain understanding breaks down. Not "weak in fractions" — consistently finds the common denominator correctly but does not convert the numerator. That specific gap becomes the child's named current ceiling, and the platform drills it until it is genuinely mastered. The adaptive engine targets a 70–85% first-attempt success rate in real time. If success drifts above 85%, difficulty increases. If it drops below 70%, the platform steps back and reteaches.

This is the first thing every conventional classroom structurally cannot do. Knowly does it for every child, every minute.

Pillar two: The Climb

No one can climb the mountain for the child.

Learning is a physical event in the brain. The neural rewiring that creates lasting knowledge happens only when the child themselves generates the answer — by hand, by voice, with their own reasoning. Watching, listening, recognizing, and copying all feel like learning and produce almost none of it. This is one of the most-replicated findings in cognitive science, and it has not entered mainstream education for the same structural reason as the Learning Edge: it is hard to enforce at scale.

Roediger and Karpicke showed in 2006 that retrieval practice — being forced to generate the answer rather than re-read it — produced dramatically better long-term retention than passive restudying. Slamecka and Graf had already shown in 1978 that self-generated content is encoded more deeply than externally-provided content, full stop. Stockard and colleagues meta-analyzed fifty years of direct instruction research across 328 studies in 2018 and found an overall effect size of d = 0.60 for explicit teaching followed by student practice — compared to d = −0.38 for unguided discovery learning. And in 2024, Van der Weel and van der Meer used high-density EEG to show that handwriting recruits significantly broader brain connectivity patterns than typing — precisely the patterns associated with memory formation.

The conclusions are unambiguous. Teach explicitly. Then make the child do the work themselves. By hand when the work should be done by hand.

This is why Knowly looks the way it looks. Every lesson begins with direct teaching — the concept is explained clearly, demonstrated with worked examples, broken into the smallest pieces the child can hold. Then the child practices. Handwritten work is required regularly across subjects, not just for major assessments. The child writes their work on paper, photographs it, and the platform reads the image and grades the reasoning behind the answer — not just whether the final answer is correct. Voice recording captures reading fluency and any subject where the spoken word is the point.

And because the work has to be theirs, the platform makes shortcuts impossible. Knowly runs as a native desktop application — a notarized Mac app, a code-signed Windows app — installed directly on your family's computer. When it is open, the child sees the lesson and nothing else. No browser to escape to. No path to a chatbot. Multiple integrity layers operate simultaneously: keystroke biometrics learn each child's typing signature and flag sessions that do not match; time-on-task monitoring catches anomalously fast responses; response pattern analysis flags reasoning that looks copied or generated. The platform will not advance the child until they have demonstrated understanding in multiple formats. There is no version of using Knowly that does not involve actually doing the work.

This is the second thing the classroom structurally cannot do — not at the level of every child, every day. Chatbots and unsupervised technology have made the climb even harder for the classroom to enforce. Knowly is purpose-built for it.

Pillar three: The Muscle

The brain forgets. The platform remembers — and brings it back.

The human brain is built to forget. Ebbinghaus documented the forgetting curve in 1885 — roughly 67% of newly learned material is lost within a day without deliberate reinforcement. Murre and Dros replicated his findings faithfully in 2015, using the same Ebbinghaus methodology with modern controls. The numbers held. The brain has not changed.

The cure, also well established, is spaced retrieval: bringing the same concept back at expanding intervals, in different forms, just as the brain is about to let it go. Cepeda and colleagues meta-analyzed 184 articles on the topic in 2006 and found spaced practice consistently outperforms cramming. Hattie and Donoghue analyzed 242 studies and 169,179 participants in 2021 and concluded that distributed practice and practice testing are the two most effective learning techniques known — outranking every other strategy studied. And in a real 7th-grade math classroom, Rohrer and colleagues showed in 2014 that interleaved practice (where the child must identify which strategy applies, not pattern-match to the topic of the week) produced a 72% correct rate versus 38% for blocked practice — an effect size of d = 1.05.

This is the third thing the classroom structurally cannot do. Real spaced retrieval requires individualized scheduling — concept A returns at one interval for one child and a different interval for another, because their forgetting curves are different. No human teacher is tracking that for twenty-two children across thousands of concepts. So in practice, classroom review is the same review for everyone, on the same week, on whatever happens to be near the current unit. It is barely better than nothing.

Knowly does this individually. Every concept your child masters enters a per-child spaced retrieval schedule built around the forgetting curve. Concepts return at expanding intervals — initially days, then weeks, then months — calibrated to your child's modeled retention for that specific concept. When a concept returns, the platform does not show the child the answer and ask them to re-read it. It presents a problem and requires the child to generate the answer. Returning concepts are interleaved with related concepts so the child must identify which strategy applies — not pattern-match to the topic of the week. The curriculum advances; the foundation never erodes. A Knowly student tested in May on material taught in September should perform at near the same level as on material taught in March — because the system never let them forget it.

Why these three, and only these three

These are not three principles among many. They are the three that, in our reading of fifty years of research, are jointly necessary and jointly sufficient for a child to learn well. If you find your child's exact edge, force them to actually do the work, and bring every mastered concept back until it is permanent — they will learn. If any one of the three is missing, they will not learn at the rate the research predicts.

Every conventional classroom, by the structural math of one teacher and many children, can deliver at most one of the three reliably. That is why the gap between elite tutored students and everyone else exists — and why it persists even with significant investment in education technology and teacher training. The gap is not about effort. It is about which three things the system can structurally deliver. The classroom can deliver one. Knowly can deliver three.

What this means for your child

In practical terms, what your child does each morning inside Knowly looks unremarkable from the outside: a lesson, some practice, some review. What is happening underneath is that the platform is finding the edge for every problem, requiring the climb on every answer, and building the muscle on every concept they have ever mastered. Every day. For as long as they use it. That is the entire product.

The morning is for that work. The afternoon belongs to your family.

The early access list is open for our August 2026 launch. If you have read this far and it makes sense, we would like to build this for your child.

— The Knowly Team