A specification that detects how you learn and adapts everything to your cognitive architecture. The forgetting curve never had a chance.
The same explanation, the same structure, the same pace, the same channel. For the person who thinks in stories and the person who thinks in diagrams. For the expert refreshing and the beginner encountering. The illusion of learning is receiving. The reality of learning is generating.
Most AI interactions stop at the first level of encoding. You read the response, feel the clarity, and move on. An hour later, 80% is gone. Not because the information was bad. Because the delivery ignored everything about how your brain works.
Allan Paivio proved that information encoded through two channels is dramatically more resilient than one. James McGaugh showed that emotional hooks activate the amygdala, bonding memories to feeling. Robert Bjork discovered that difficulty during learning predicts long-term retention. Easy in, easy out. Hard in, permanent.
Hermann Ebbinghaus mapped the forgetting curve in 1885. Within an hour, 50% of new information is lost. Within a day, 70%. The curve is only defeated by retrieval practice woven into the flow itself.
Learn Mode is a learning specification. Not a tutor. Not a quiz engine. A document that restructures how an intelligence adapts its delivery to the human it is teaching. From one-size to profiled. From receiving to generating.
Seven principles extracted from cognitive science. Six encoding levels. Five learner profiles detected from behavior, not self-report. The spec IS the adaptation. Give it to any AI and watch the teaching change. The forgetting curve breaks. Every mind is reached. A different axis entirely.
Detect shapes the channel. Content structures the flow. Emotional adds color. Dual Channel splits into two parallel streams. Push Up creates rapids of active engagement. Calibrate adjusts the width. Retrieval loops back upstream. Each principle transforms the information before it reaches the learner.
Your learning profile extracted from behavior, not self-report. People don't know how they learn. Their patterns know.
Knowledge type determines the skeleton. Procedural gets sequences. Conceptual gets relationships. The structure precedes the learner.
The amygdala is the gatekeeper. No emotional hook, no long-term memory. Every explanation carries a feeling because feelings are what survive.
Passive reception is a 20% retention ceiling. Every exchange demands generation. You do not receive information. You create understanding.
Retrieval practice woven into the flow. Not a quiz at the end. Not a flashcard later. The curve is broken inside the conversation itself.
Built by someone who watched AI teach everyone identically. The same bullet points. The same numbered lists. The same gentle encouragement. As if every brain were a photocopy of every other brain. As if receiving information and learning it were the same thing.
The question was not "how do we make AI a better tutor?" The question was: what does cognitive science actually say about how different minds encode, retain, and retrieve information, and can we build a specification that respects that?
Four researchers. 140 years of memory science. Seven principles extracted. One specification that turns any AI from a tool that explains into a system that adapts. A different axis entirely.