MasteryMath is the point where an abstract learning model had to survive contact with an actual interface. The project asks what MasteryGraph looks like when a learner is holding an iPad and needs the next problem, not a theory of the next problem.
A SwiftUI iPad MVP with deterministic problem generation and checking, local progress storage, learner practice flows, and a parent-facing dashboard.
The aim. Test whether the MasteryGraph methodology can produce a calmer, more accountable learning tool.
01The problem behind the project
The project turns an abstract mastery model into a testable artifact and avoids using an LLM where exact mathematical checking is more reliable.
Mathematics also provides a useful boundary for AI. A model may help design the system or explain a concept, but exact problem generation and answer checking can remain deterministic, local, and inspectable.
Learners and parents are the intended users. Any future use with children requires careful privacy, accessibility, and pedagogical review.
02How it took shape
SwiftUI, SwiftData, deterministic math generation and validation, local-first storage, a structured mastery model, and smoke-tested app flows.
The SwiftUI and SwiftData MVP includes deterministic practice, validation, local progress storage, learner flows, and a parent dashboard. It has built and passed initial simulator smoke testing, making it a working prototype rather than a presentation-only concept.
Josiah defined the learning philosophy, feature boundary, local-first posture, and acceptance criteria while directing agent-assisted implementation.
A working build and core smoke tests exist. There is not yet evidence of sustained learner use or improved outcomes.
03What the project means now
What remains unproven is educational effect. Curriculum breadth, diagnostic validity, interface quality, and sustained learner use require real testing. The prototype's value is that it makes those questions concrete and exposes which parts of the broader methodology have actually been implemented.
The curriculum, interface, assessment validity, and content breadth need refinement before external use.
AI can help build the learning system while deterministic code remains the right authority for exact answers.
Refine the artifact, document the implemented methodology, and conduct a small consented usability test.