The production path. Parse a beat_this .beats file,
build a tempo model with 01-the-math, play it back, and
watch the playhead's musical position (β) follow the audio clock through
the warp map. The pickup-beat offset that the teaching layer skips lives
in grid.js, kept clearly separate from the time shift that
anchors the model at {0,0}.
For each beat in the file: the time we PARSED, the beat we INFER
via secondsToBeats, the time we re-derive via
beatsToSeconds, and the residual. Should be zero to
numerical noise for a piecewise-constant model.
| i | parsed t | β = s→b(t) | t' = b→s(β) | residual |
|---|