MathPilot

The idea

MathPilot diagnoses a learner's level, tracks mastery on a prerequisite graph, recommends the next useful step, and keeps progress on the Mac. Practice, cumulative review, homework help, symbolic checking, and optional Codex coaching live in one desktop workflow.

The problem

Calculus learners are often forced to assemble videos, generic chat, homework tools, and spaced repetition themselves. Those tools rarely share a mastery model or preserve a private, durable learning history.

How it works

I combined a Tauri desktop shell, React interface, SQLite persistence, FSRS review scheduling, MathLive typesetting, bundled Python and SymPy, Vision OCR, and deterministic offline fallbacks. Optional Codex features use the local CLI rather than an in-app API key.

A prerequisite graph connects diagnostic results to practice recommendations. Scheduled review revisits earlier material, while SymPy checks supported symbolic answers locally.

The work

The public repository includes the calculus bank, diagnostic and review flows, local SymPy checking, and automated tests. Optional coaching uses the local Codex CLI.

Calculus 1 and 2 · local symbolic checks · scheduled review

Where it stands

The macOS app connects a calculus problem bank, diagnostic sessions, practice, scheduled review, and local homework analysis. Learning history stays on the Mac; optional AI coaching supplements the offline tools.

Scope and constraints

Private local-first learning history; optional Codex uses local CLI, not an in-app API key.