Tom. Teacher of Music. An AI music teacher who has ears.

Tom runs the lesson — he doesn't just talk about it. Tom lives inside the practice app, not beside it. When I talk to him he doesn't only answer, he acts: opens the right drill, sets the scale visualiser to the key I need, lays out the week's plan, moves me to the next screen. He drives the same app I do, through a set of typed tools — the same way I'd tap the buttons myself.

He remembers me — and he keeps working when I've stopped. What's hard, what's clicking, the pieces I'm on — it all lives in a memory Tom reads before he says a word, so he never starts from scratch. And every night, while I'm not practising, he reviews the day and adjusts tomorrow: nudges a tempo, swaps a drill, rewrites the plan around what actually happened.

Underneath is a map of everything there is to learn. Tom builds a graph of skills and the prerequisites between them, and grows it as I progress. A node only lights up when I've actually demonstrated the skill — not just practised it. Effort and mastery are deliberately separate currencies, so "progress" means I'm genuinely better, not that I simply turned up.

This is a large build — larger than I would attempt alone. AppFactory is what makes it feasible — the agent-system that turns architecture decisions into shipped software. Tom is the first real demonstration of what one engineer can build with AppFactory behind them.

Every way in, one teacher — and a widening library to teach from.

• Text Conversations in the practice app, with memory that persists across sessions.
• Vision Captioned manual imagery inside Tom's knowledge base.
• Time-series Session logs, BPM progression, a model of how I actually improve over time.
• Practice & theory Chords, scales, ear training and rhythm — a growing library of practice tools and theory studies Tom teaches from.
• Structured score MusicXML ingested from score PDFs via OMR, rendered in-browser with OpenSheetMusicDisplay.
• MIDI Piano performance captured natively over USB, diffed against the canonical MIDI of the piece. Coming soon
• Audio Guitar recorded through the browser, pitch and timing extracted with basic-pitch and DTW. Coming soon

One subtle call: piano and guitar need different pipelines. Piano emits MIDI natively — a clean, millisecond-accurate signal. Guitar emits audio through a pickup — messier, polyphonic, needs inference. Same teacher, two capture paths. The architecture mirrors the instruments rather than flattening them into one pipeline.

At the end of that pipeline: a score-aligned playback diff. Record a take, compare it to the score, colour each note as I played it — green correct, amber timing off, red wrong note, grey missed. The capability whose removal stalled my own progress — rebuilt, and this time it lives inside a teacher who knows me.