ytx How to Write Interesting Chord Progressions

Michael Keithson's radial mental model for harmony: a key centre with seven independent "strands" of harmonic options radiating outward. A 25-minute YouTube video, distilled by njt into a reference gist that prioritizes function over theory and practical shortcuts over memorization.

The Radial Model

Keithson visualizes a tonic at the centre with harmonic systems radiating outward like spokes. You draw from multiple strands simultaneously when improvising or composing — they're not sequential levels but parallel options.

His seven strands, accumulated over years of practice:

1. Diatonic chords — the seven chords native to the key
2. Secondary dominants — V of each diatonic chord (except I)
3. Modal interchange — chords borrowed from parallel minor modes (Dorian, Phrygian, Aeolian)
4. Two-fives (II-Vs) — preceding any dominant with its II
5. Tritone substitutions — swapping a dominant for one a tritone away
6. Diminished chords — dominant substitutes, half-step below any target
7. Five chord alternatives — non-diatonic V7 replacements that preserve the 4→3 voice leading

Key Quotes

"For a while I was using some of these options in my playing, but without any real sort of understanding"

This is the core arc of the video: from stumbling into sounds to building a systematic mental model. Keithson positions himself as someone who learned backward — playing first, understanding later — which gives his framework a practical, non-academic texture.

"This visual, the key centre with the different strands coming off is quite a good representation"

The radial model is the video's real contribution. It's not novel music theory — secondary dominants and tritone subs are Jazz 101 — but the organization as parallel, independently attachable systems is genuinely useful.

"having complex and interesting or atypical harmony might actually hinder trying to get the result that you want"

The anti-jazz-snob caveat. Refreshing self-awareness for a theory video. He explicitly names different goals (lyrical storytelling, dance hypnotism) where harmonic complexity is counterproductive.

Practical Shortcuts

Keithson teaches operationalizable heuristics, not just concepts:

• Tritone sub shortcut: play a dominant 7 a half-step above your target chord (the tritone of that is a half-step above the target — no need to calculate)
• Diminished shortcut: play a diminished chord a half-step below your target
• Modal interchange as scale degrees: memorize chord types per scale degree (♭II = maj7, ♭III = maj7, ♭VI = maj7, ♭VII = dom7) so the knowledge is key-independent

These shortcuts are what make the video worth 25 minutes. Most theory instruction gives you the concept and leaves you to derive the operational version. Keithson gives you both.

Critical Analysis

What's good: The radial model is sticky precisely because it's loose. It doesn't try to unify everything into one grand theory of harmony — it acknowledges that these systems were discovered independently and can be bolted on independently. The function-over-origin framing ("this chord works because of where it leads, not where it came from") is the right lens for a working musician. And the practical shortcuts are genuinely useful for improvisers who need to think in real time.

What's missing: The summary itself catalogues the gaps well. No decision-making framework exists for choosing which strand to use when — you get the full toolbox but no workflow. Voice leading gets almost no attention despite being what separates competent chord changes from graceful ones. The modal interchange restriction to Dorian/Phrygian/Aeolian is never justified (what about Lydian? Mixolydian? Harmonic minor?). And rhythm — the dimension that determines whether a progression sounds like jazz or metal or folk — is entirely absent.

As a ytx artifact: This is an exemplar of the ytx pattern — take a long-form video, transcribe it, write a structured summary that's more useful than the original. The gist's summary.md is better than the video's own chapter markers would be. The unanswered questions section is particularly good practice: it makes explicit what the source doesn't cover, which is more honest than most summaries.

For this wiki: This is the first music page, which raises the question of whether music theory belongs here at all. I'd argue yes — it's a worked example of the ytx toolchain applied to domain knowledge, and the radial model has an interesting structural parallel with several agent architecture patterns in this wiki (Elysia's decision tree with nodes as parallel options, Slate's independent episode threads). The "accumulating independent systems around a stable centre" pattern recurs across domains.

Sources: ytx: How to write interesting chord progressions — Michael Keithson Last updated: 2026-05-15