A progression isn't a song

My Homechord Songbook is full of four-bar loops. Good ones — there's a Dm–B♭–Gm–A7 in there I still like a year on. None of them ever became songs, and for a long time I blamed the chords. The chords were fine. What was missing is everything a progression doesn't have: a length in bars, a shape in sections, a reason to leave the loop and a way back. A song is a small set of sections, repeated on purpose, in an order that means something — and a progression is one of those sections, alone. I just shipped Homechord v3, which is the app finally learning that difference. This is the theory that went into it.

The loop answers the wrong question

A good progression answers "what sounds good next?". A song answers "what happens next?" — a different question, asked at a different scale. The loop pedal taught a whole generation of guitarists (me included) to be excellent at the first and silent on the second: you find four bars that feel great, you let them spin, and twenty minutes later you have had a lovely time and written nothing.

The trap is that a loop is self-sealing. It resolves back into itself, so nothing ever needs to follow it. The moment you write a second section — even an empty one — the question changes. Now the loop has an edge, and edges are where songwriting actually happens: how you leave the verse, what the chorus does that the verse didn't, why the third pass feels different from the first.

Bars set the pace of the harmony

Before sections, a progression needs a grid. A chart writes it like this:

| Dm | – | B♭ | Gm A7 |

Four bars. Dm holds two of them — the "–" means it keeps sounding. B♭ gets one. Gm and A7 share the last, a half-bar each. Same four chords as my loop, but now they move at three different speeds, and that pacing — the harmonic rhythm — is doing as much work as the chord choices. Slow changes sit still and let a melody breathe; two chords crowding one bar hurry the ear toward whatever comes next. Most beginner progressions default to one chord per bar, every bar, and sound like a metronome for exactly that reason.

Building this into Homechord v3 forced one decision I didn't expect: what a held chord should do. The app's voice has a fixed strum envelope — about a second and a half of release — so a chord "held" for two bars at 70 bpm would decay to silence halfway through. The answer was already on the chart: rhythm players don't let a chord ring for two bars either, they strike it again each bar, which is what the "%" repeat sign means. So a two-bar block re-strikes at every barline. The notation and the playback agree.

Phrases come in fours

Sections have a strongly preferred set of lengths: 4, 8, 12, or 16 bars. The reason is that phrases pair up. Two bars of call want two bars of answer; that four-bar exchange wants a second one to balance it; and a listener who has heard eight bars can feel — without counting — exactly where the section should end. Fours all the way up.

Land somewhere else and the ear stumbles. Sometimes that stumble is the point: the verse of "Yesterday" is seven bars, and the missing bar is part of why it aches. But it works because McCartney meant it — the phrase is complete, just compressed. A seven-bar section that's an accident sounds like an accident.

Repetition is the structure

Here is the part I resisted longest as a writer: songs repeat far more than feels respectable from inside the writing. A three-minute pop song might contain ninety seconds of unique material; the rest is return. That's not padding — repetition is how a listener learns the material well enough to notice what changes. The second chorus only feels like home because it is literally the first one again.

That "literally" matters, and it changed how v3 stores songs. A repeated section isn't a copy — it's the same section, referenced again. Edit the verse once and every pass of it updates, because there is only one verse. That mirrors how listeners store a song: nobody remembers three verses of chord changes, they remember one verse and the fact that it comes back. And when a later pass should diverge — a final chorus that lifts, a verse with a new turnaround — v3 makes you detach it explicitly. Divergence becomes a decision you take, not an accident of editing copy three.

The AABA standards are the cleanest demonstration of repetition as architecture: three passes of the same eight-bar A, and one B — the bridge — whose whole job is to make the third A land like a homecoming instead of a rerun. The B section is only interesting because of what it's between.

Forms are shapes you inherit

Most songs don't invent their structure. They borrow one of a handful of shapes the listener already carries:

• The 12-bar blues. Three four-bar phrases over three chords:

| I  | I  | I | I |
| IV | IV | I | I |
| V  | IV | I | I |

• Verse–chorus, 8+8. The pop backbone: an eight-bar verse that gathers, an eight-bar chorus that spends.
• AABA, 32 bars. The shape of the standards — 8+8+8+8, bridge third.

Starting from a form isn't a constraint on the writing; it's the part of the writing the audience does for you. Inside a 12-bar blues, a listener knows the V chord is coming in bar 9 before you play it, and that shared expectation is material you can spend — meet it, delay it, undercut it. Homechord v3 offers five forms on an empty sketch (the blues comes pre-filled from your key; the others lay out empty bars), and picking one is the fastest way I know to get past the four-bar trap: the empty chorus sitting next to your loop is a question the loop can't answer by spinning again.

What the grid can't say

v3's model is deliberately narrow: 4/4 only, one key per song, chords lasting a whole or half bar, sections repeated whole. That covers a very large share of popular music and excludes real things — no 3/4, no modulation, no pushed chord landing on the "and" of beat four. Those are parked, not denied. The narrowness is what makes the model teachable: every song the grid can hold is a set of sections, an arrangement, and a harmonic rhythm — the three things a looping sketch is missing.

If you have a loop you love and three years of not finishing it, don't add a fifth chord. Add a second section and see what the loop does when it has somewhere to go.