How these pages are made, and how they are checked

How a page gets its answer

Each entry in the library stores one spelling. For the C major chord that is C, E and G; for F sharp major it is F sharp, A sharp and C sharp. Everything else the page shows is read back out of that spelling by the site’s theory kernel: the intervals, the scale degrees, the inversions, the fingerings, the keyboard diagram, the notation. Nothing is typed out page by page.

That is why sibling pages agree with each other. A page for one key and a page for the next are the same arithmetic applied to a different root, so a mistake cannot exist on one page and not on its neighbour. It also means a correction lands everywhere at once rather than in the one place somebody happened to notice it.

It also fixes what the arithmetic cannot. Deriving notes from a spelling gets the letters right; it cannot tell you whether the stored spelling was the right one to begin with. A D half diminished chord contains the same sounds whether you write the third tone as A flat or as G sharp, and only one of those is what a musician reading the page would write. That judgement is not arithmetic. It is the part a person does.

What a review covers

When an entry is reviewed, the review is signed against exactly ten fields: the note names, the formula, the intervals, the scale degrees, the root voicing, the first, second and third inversion voicings, and the right and left hand fingerings. Those ten are hashed together and the hash is stored with the review.

The point of storing the hash is that the review can expire on its own. Every time a page renders, the hash is recomputed from the live data and compared to the one the review was signed against. If they differ, the musical data has changed since a person last looked at it, and the page says the entry needs re-checking rather than continuing to show a review date. A stamp that keeps claiming a page was checked, after the content it vouched for has moved, is worse than no stamp at all.

The ten fields are the ones a review is about. The prose, the images, the frequently asked questions and the song references sit outside it deliberately, so that editing a paragraph does not silently invalidate a check on the music, and so that a review never claims more than it looked at. A reviewed entry also refuses further automated edits to those ten fields: a correction pass that would overwrite hand-checked data is rejected by the database and raised as a defect to be looked at, not applied quietly.

What the review count means

1 of 1,530 chord and scale entries have been reviewed by hand (0.1%). That count was taken on the last build, in July 2026.

It is a coverage figure, not an accuracy rate. It says how much of the library a person has signed off, and nothing about whether the rest is right. An entry with no review is an entry nobody has hand-checked yet; it is not an entry known to be wrong. Most of the library has never needed a correction, and the derivation described above is why.

The number is recounted every time the site is built, so it is current as of the last deploy rather than as of whenever somebody last edited a figure by hand. It can only go up by someone reading a page and signing it, which is slow, and it is meant to be.

What “checked most days” means

It means what it says, and it is worth being precise about why it is not a stronger claim. Most days, some part of this library is read, corrected or reviewed. Not every day. There is no schedule, no shift, and no team.

Automated checks do run, but they run when the site is built rather than on every change, so the honest version is that nothing reaches the site without passing them, not that every edit is inspected the moment it is made. Where a page names a cadence, it names the one that is actually true.

The correction log

When musical data changes, the change is recorded: which entry, which field, what it said before, what it says now, and the rule that decided it. The record goes back through the corrections already made, including a pass in July 2026 that re-spelled 118 entries so that each note letter follows its interval degree, and a smaller pass that reordered sixteen extended chords and moved a sharp eleventh into the octave where it actually sounds.

The log is kept privately rather than published. A running feed of every note name this site has ever got wrong would be a strange thing to put on a reference work, and it would invite reading individual corrections as a verdict on the whole library. What it is for is making the claim on this page checkable by the person responsible for it, and making sure a correction is a recorded event rather than a silent overwrite.

Who is accountable

One person: Justin Evans, a pianist of twenty five years who studied jazz with Ted Howe, a former Berklee professor. The credentials, the playing history and the reasons this site exists are on the about page.

If something here is wrong, the correction link at the bottom of every reference page reaches a real person, and corrections are read.

This page describes what is true today, including the parts that are unfinished. If a claim on it stops being accurate, the claim changes rather than the wording softening.