PDFs, Word documents and PowerPoint slides: a textbook, the lecture deck and the student's own notes, up to a scanned book of several hundred pages. Chameleon reads every chapter, figure and citation, then merges them into one clean structure.
It identifies the discipline and adopts the persona of a specialist in it, applying the grading standards and the specific misconceptions examiners in that field look for.
Nothing is generated from outside the uploaded material.
New study plan
The reader is a card-per-concept deck. Each concept opens with a concrete instance or a prediction question that creates the need for it, before any definition appears.
Between concepts sit dedicated connection cards: bridges that make the relationship between concepts the object of study, syntheses that integrate a chapter into one mental model, and misconception cards that name the trap, why it feels right, and where it breaks.
Every claim carries its citation. When something surprises the student, the original page is one click away.
One idea, elasticity, connects the whole week: who really pays a tax, why a price cap leads to shortages, and how a good harvest can still leave farmers worse off.
The model never draws. It emits a typed specification, and the server computes the domain result first: the equilibrium, the net force, the 3:1 ratio, the loop current. Only then is the figure drawn. These four are the real renderer's output.
FSRS-5 models the memory of every card and estimates the day it will be forgotten. Each morning shows the smallest set that keeps the whole syllabus fresh. The plan comes from the forgetting curve, and counts down to the exam.
Quiz misses spawn flashcards automatically, so demonstrated weaknesses flow straight into the schedule.
Forgetting forecast · which topics fade before exam day
A city caps rents 20% below the market rate. Demand is highly elastic; supply is inelastic in the short run. What happens first?
B is right: with elastic demand and short-run inelastic supply, the cap raises quantity demanded while quantity supplied barely responds, so a queue forms. A confuses the short and long run; C ignores the quantity response; D reverses the direction of the supply incentive.
☞ LECTURE4.PDF · P. 117 · FEEDBACK ON EVERY DISTRACTOR (PASHLER 2007)
Multiple-choice, numeric, multi-part and open-ended questions, marked to the standards of the subject. Written answers are graded against a rubric, with specific feedback on what to correct.
Every distractor is a real misconception, and the explanation names it. Feedback is part of every question, not an optional extra.
Accuracy is tracked per question format, because recognizing an answer is not the same as producing one.
Readiness comes from the memory model, not a completion bar. Each card's recall probability is weighted by how critical its material is to the exam, so a pile of easy facts cannot mask a core mechanism the student has let slip. The calibration curve then shows how far a student's self-assessment runs ahead of their actual recall. It is measured and shown to them, never quietly folded into the score.
Exam-weighted readiness · you vs pace vs target
Calibration · predicted vs actual recall