← GUMP
Learn Engine
Know what you don't know. Know when you get it.
$29/user/mo
Most education software tracks completion. Watched the video? Check. Passed the quiz? Check. But completion isn't understanding. Learn Engine detects grokking — when a student's answers shift from pattern-matching to solving unseen variations. Hard to fake. Hard to miss when tracked over time.
Student solves 5 problems correctly but fails when the numbers change — that's memorization. After prerequisite review, student solves unseen variations — that moment is grokking. Learn Engine maps every gap, recommends prerequisites in the right order, and marks the transition when it happens.
WHAT YOU GET
Grokking detection: know the moment understanding clicks
Knowledge gap mapping across any curriculum
Prerequisite recommendations in dependency order
Per-student comprehension timeline
Concept clustering by difficulty and relatedness
Plateau detection: when a student is stuck
Curriculum bottleneck analysis
Exportable reports for instructors and students
Neural Engine grokking detection on Apple Silicon
Hardware-signed progress reports via Secure Enclave
WHAT THIS IS / WHAT THIS ISN'T
WHAT THIS IS
An adaptive learning tracker that detects sustained mastery (grokking) through streak analysis and prerequisite mapping.
WHAT THIS ISN'T
A content platform. You bring the curriculum. We track understanding.
YEAH BUT
"Khan Academy is free."
Free doesn't detect understanding. Khan tracks video completion and quiz scores. It never knows if you actually get it or if you memorized the pattern for the test. Learn Engine detects the phase transition. That's a different product.
"Grokking is an AI research term."
It's a phase transition. Same physics as ice melting. A system memorizes, then suddenly generalizes. It happens in neural networks AND in human brains. We detect when the transition fires. The math doesn't care if the learner is silicon or carbon.
"Built on a Mac Mini."
The math runs on anything. Grokking detection is spectral analysis on a knowledge graph — lightweight, local, instant. Hardware scales when you need it. The algorithm already did the hard work.
VS THE COMPETITION
Knewton Alta
$44.95/semester — adaptive, but relies on correctness, not whether the concept generalizes.
Learn Engine: detects actual comprehension. Maps gaps. Recommends prerequisites in the right order.
DreamBox
$12.95/mo — math-only. Adaptive pacing, but no comprehension detection.
Learn Engine: works on any subject. Detects the moment understanding clicks. Clear, exportable reports.
Khan Academy / Khanmigo
$4/mo — tracks video completion and quiz scores. No generalization detection.
Learn Engine: knows exactly what you understand and what you don't. Phase transition detection, not quiz scores.
TRY IT
3 free analyses. Fixed sample data. Your own data requires a license.
THE PHYSICS
Why is understanding so much cheaper than memorization? Landauer's principle: erasing one bit costs kT ln(2) joules. Memorizing 100 samples independently costs ~320,000 bit-erasures. Understanding the pattern (compressing 100 samples into the rule that generates them) costs ~1.43. That's a 224,000× difference. Grokking IS compression. The phase transition from memorization to generalization is the moment the model finds the cheaper encoding. We measure that transition.
TESTING
Unit tests, adversarial input testing (None, wrong types, NaN, empty data, unicode), real user workflow testing, and cross-product integration testing. Every public function handles every input permutation without crashing. 673 quality tests across all products, zero failures. Self-verifying: the product can audit its own output.
By purchasing you agree to our Terms of Service
Get Learn Engine — $29/user/mo
After purchase: setup guide
INSTALL
pip install begump
from gump.learnengine import *