Built on one idea:
you don't understand something until you can teach it.
Traditional tests check recall. Axiom Flow checks understanding, by having students teach a concept to an AI student, then scoring how that AI performs on an exam. It's live in real university courses today, not a concept.
Why we built
Axiom Flow
Higher education hit a wall. Students can memorize answers, and generative AI can produce a correct-looking response without anyone understanding anything. Recall-based testing can't tell the difference anymore. We needed a way to test something AI can't fake: explaining a concept well enough to teach it.
“If a student can teach it clearly, they understand it. If they can't, no quiz score will tell you that, but teaching will.”
The Flaw in Checking-Based Systems
Multiple-choice and recall-based tests measure recognition, not comprehension. Generative AI made this worse, it's now trivial to produce a correct-looking answer without understanding it. Axiom Flow flips the dynamic: instead of answering questions, students have to teach the concept to someone who knows nothing.
The Rigor of the Feynman Technique
The premise is simple: you don't truly understand a concept until you can teach it clearly. When a student has to correct a flawed AI student's misunderstanding, the gaps in their own knowledge surface immediately, gaps that a multiple-choice exam would never catch.
A Dedicated Assessment Layer, Not a Teaching Tool
Axiom Flow doesn't deliver content or replace your LMS. It plugs into your existing Moodle courses through LTI and does one job: measure whether students actually understand what they were taught.
The Dual-Agent Assessment Architecture
Two separate AI agents, each with a narrow, specific job: one designs the assessment, the other is the one being taught.
Atlas
EXAM DESIGNER & GRADER
Atlas has full access to the learning material. It analyzes the content, identifies the key concepts, and builds a set of realistic misconceptions mapped directly to exam questions. After the teaching session ends, Atlas also grades Sam's exam performance and reports the resulting score.
Full Material Access
Exam Design & Grading
Sam
AI STUDENT
Sam starts with zero outside knowledge: no access to the source material, no way to look anything up. He's initialized with the misconceptions Atlas created, and the only way his understanding changes is through what the student teaches him. He then sits the exam using only that taught understanding.
Strictly Context-Bounded
Learns Only From the Student
Rajika Chathuranga
A solo software engineer based in Hikkaduwa, Sri Lanka. Axiom Flow started as a question: how do you actually prove someone understands something? and is now running in real university courses with over 100 students using it through Moodle.
Connect on LinkedInRegistered Address
Axiom Flow (PVT) LTD
129/20C, Weralugahawatta, Nalagasdeniya, Hikkaduwa, Sri Lanka
Get in touch
For integrations, partnerships, and institutional licenses, reach out directly.
rajika@axiomflow.app