Research note 01 · Elevation 1.3
Rating the difficulty of piano music
Why difficulty is multidimensional, how the scale is calibrated, and why the exact arrangement matters more than the song title.
Read note 01Research archive
The difficult parts.
Problem statements, experiments, benchmarks, failures, and product decisions from building a piano practice instrument.
Published notes
Research should leave evidence.
These notes distinguish implemented behavior from hypotheses, public evidence from restricted research material, and useful negative results from claims of progress.
Research note 02 · Ergonomic DP
Generating playable fingerings
Dynamic-programming search, hard physical constraints, confidence-gated output, corpus rights, and the limits of current evidence.
Read note 02Working principles
EvidenceVersion the analyzer and preserve provenance
EvaluationHold out complete works and measure coverage
ClaimsPublish failure cases and rejected models
Full record
The repository contains the machinery.
Corpus manifests, calibration procedures, benchmark commands, result tables, and clean-room boundaries remain beside the implementation so the work can be inspected and repeated.
Browse research documents on GitHub