Some Things in Music Are True. The Algorithm Is Not One of Them.
On discovered truth, commercial invention, and what music knows that Spotify's engineers don't.
What Was Already There
When a shepherd in Mongolia plucks a gut string and arrives at a five-note scale, and a griot in Mali strikes a kora and arrives at the same five notes, and a folk singer in Appalachia reaches for the same intervals without ever having heard either of them — something is being revealed about the nature of the physical world.
The pentatonic scale was not designed. It was found. It is a consequence of the harmonic series — the physical fact that any vibrating object produces not a single frequency but a cascade of overtones in mathematically precise ratios, 2:1, 3:2, 5:4, that the human auditory system is specifically built to receive. The five notes of the pentatonic scale are the most consonant distillation of those ratios. They minimize what physicists call auditory roughness — the interference patterns that create dissonance on the basilar membrane. They align with the brain’s predictive reward architecture, triggering dopamine release when the pattern completes itself. They are, in the most literal sense, in tune with the physical structure of the universe and the biological structure of human hearing.
This is why they keep being found. Not because cultures copied each other across oceans and millennia, but because the relationship was already there, waiting in the physics of sound, and every culture that spent enough time with vibrating objects eventually stumbled onto the same truth. Leonard Bernstein observed the scale appearing independently in Scotland, China, Africa, Native American cultures, and the East Indian classical tradition — not as a coincidence of taste but as a convergence on physical reality. The discovery was inevitable because the thing being discovered was real.
The algorithm was not found. It was built.
Two Different Kinds of Truth
There is a distinction that philosophy has long maintained between things that are discovered and things that are invented, and it matters enormously here.
Mathematical relationships are discovered — the Pythagorean theorem was true before Pythagoras, would have been true if he had never existed, and will be true after every civilization that currently uses it has collapsed. Calculus was discovered independently by Newton and Leibniz because the relationships it describes existed before either of them. The law of gravity applied before Newton named it. These things were not created by human ingenuity. They were located by it. The finding is not the same as the making.
The pentatonic scale belongs to this category. It is a feature of the acoustic world, not a cultural preference. When the shepherd and the griot and the folk singer all arrive at the same five notes, they are not imitating each other. They are all finding the same thing that was already there — the way mathematicians on different continents independently discover the same theorem, because the theorem is not invented by any culture but exists independently of all of them.
The recommendation algorithm belongs to an entirely different category. It is not a discovery of a pre-existing relationship. It is a set of deliberate design decisions, made by specific engineers at specific companies, in service of specific financial objectives that were chosen and could have been different.
The Spotify recommendation engine was built to maximize time-on-platform and subscription retention. Those objectives were not discovered in the harmonic series. They were decided in meetings. They could have been replaced by different objectives — maximizing neurobiological wellbeing, cultural diversity, the discovery of music from underrepresented communities, the connection between listeners and artists who share their community. They were not replaced because the people making the decisions were not trying to achieve those things.
The algorithm’s goals are not features of the acoustic world. They are features of a quarterly earnings report. And that distinction determines everything about what the algorithm can and cannot hear.
Studying the Circle of Fifths vs. Studying the Algorithm
Studying the circle of fifths and studying the algorithm look identical from the outside. Both require sustained attention, pattern recognition, and the adaptation of creative decisions to what the system rewards. Both produce expertise. Both shape music.
They are structurally opposite.
When a musician learns the circle of fifths, they are paying attention to physical reality — to what resonates, what resolves, what produces the neurochemical response that signals rightness. The circle of fifths does not update its terms of service. The harmonic series has not released a new version that deprecated the perfect fifth. The knowledge is permanent. It compounds, transfers across instruments and genres, and is as valid today as when it was first learned. A musician who mastered the circle of fifths in 1918 handed down something their students could use in 2025 without revision.
When a musician studies the algorithm, they are paying attention to a commercial artifact — to what the system rewards, what produces distribution, what generates streams and saves and playlist placements. The attention is real. The expertise is real. The thing being attended to is an invention with no claim to permanence or truth, controlled by people whose interests are not aligned with the musician’s interests, and subject to revision without notice.
This creates an asymmetry with no parallel in music history.
The musicians who built careers on early SoundCloud’s discovery mechanics found that knowledge irrelevant when the platform’s priorities shifted. The musicians who mastered MySpace’s recommendation logic found it worth nothing when the platform died. The artists who optimized for the 2018 Spotify algorithm found their reach restructured when the algorithm was rebuilt. In each case, the knowledge accumulated through genuine effort and careful attention was rendered contingent by a unilateral decision made by people who owed the musicians no explanation.
The pentatonic scale has never betrayed anyone. The algorithm was designed to.
Not out of malice. Out of commercial logic. The system must remain unpredictable enough that its preferences cannot be reliably gamed, because if its preferences could be reliably gamed, the commercial objectives it was designed to protect would be undermined. Opacity is not an accident of algorithmic complexity. It is a design requirement of algorithmic commerce. The cat-and-mouse dynamic between musicians gaming the system and engineers updating it is not a bug. It is a structural feature of building creative practice on an invented commercial artifact rather than on discovered physical reality.
A musician who mastered the circle of fifths accumulated wisdom about permanent features of the world. A musician who mastered the algorithm accumulated knowledge about a contingent commercial arrangement that can be revised overnight. These are not the same kind of knowledge, and treating them as equivalent is how musicians end up building careers on sand.
What the Gaming Reveals
When musicians study the algorithm closely enough to reverse-engineer its preferences, they sometimes believe they are doing what the shepherds and griots did — paying careful attention to how music actually works. The analogy is seductive. It is wrong.
The shepherd found the pentatonic scale by attending to acoustic truth — to what resonated in the body, what resolved in the ear, what produced the neurochemical response that had been selected for over millions of years. The discovery rewarded attention to the world.
The musician gaming the algorithm finds what the algorithm rewards by attending to the algorithm — to a scoring function built by engineers whose goal was retention, not resonance. What they discover is not how music works. It is how this particular commercial artifact, at this particular moment in its development cycle, interprets behavioral signals in service of its current objective function.
Then the system responds. This is the step that has no parallel anywhere in the history of musical knowledge. When musicians learn to use the circle of fifths effectively, the circle of fifths does not change. The relationship between those tones is not going to be revised in the next quarterly update. The discovery, once made, belongs to everyone who makes it.
When musicians learn to game the algorithm effectively, the algorithm’s owners update it. Not to serve musicians better. To prevent the gaming from undermining the metric the system was designed to protect. The update is not a correction toward truth. It is a correction toward continued commercial control.
This is what it means to say the algorithm is invented rather than discovered. Discovered truths are durable. They do not have owners. They do not protect commercial interests. They do not change when someone learns to use them well. Invented systems do all of these things, because they were designed by people with interests to protect, and those interests do not coincide with the interests of the musicians the system distributes.
The Alternative Was Always Possible
The algorithm could have been built differently. This point must be stated plainly because the current system presents itself as inevitable — as the natural consequence of scale, data, and technological capability — when it is in fact the consequence of specific choices.
Endel, the AI-driven soundscape service, uses algorithmic design to optimize for neurobiological wellbeing rather than retention. The design principles are rooted in psychoacoustics and circadian rhythm research — cortisol reduction, alpha wave synchronization, the entrainment of biological rhythms to musical rhythms. The same class of technology pointed at a different objective function produces measurably different outcomes for the people using it.
Musinique’s Muzack project makes the same argument in practice. Functional music engineered from neurobiological research — specific tempos for sleep onset, specific modal structures for grief processing, specific rhythmic patterns for sustained attention — built not to keep a user on a platform but to change the state of a specific nervous system. The same AI tools that maximize streaming engagement, pointed instead at the question: what does this person’s nervous system actually need?
The technology is not the problem. The objective function is the problem. And objective functions are chosen. They can be changed. They reflect the values and interests of the people who write them — which means that different values and different interests produce different systems and different outcomes for the musicians and listeners those systems serve.
The pentatonic scale did not need to be chosen. It was already there, in the ratios between harmonics, waiting for anyone attentive enough to find it. The algorithm’s objectives were chosen in meetings and written into code and could be rewritten by people with the authority and the will to do so.
That authority currently belongs to the platforms. It does not have to remain there.
What the Body Has Always Known
The pentatonic scale produces measurable neurobiological effects. Consonant intervals reduce cortisol. Rhythmic entrainment synchronizes heart rates. Music with genuine personal emotional salience releases oxytocin in ways that generic algorithmic content cannot replicate. These effects are not cultural preferences. They are biological responses to physical properties of sound, documented across populations, grounded in the same harmonic relationships that the shepherd and the griot and the Appalachian folk singer all independently found.
The algorithm cannot produce these effects by optimizing for skip rates. It cannot produce the oxytocin of music made specifically for a specific person by someone who loves them. It cannot produce the prolactin release of a grief song that resolves properly because someone understood what resolution means for a mourning nervous system. It cannot produce the entrainment of a lullaby in the grandmother’s language because it does not know the grandmother’s language, does not know the grandmother, and was not built to know either.
What the algorithm can produce is content that retains attention long enough to be counted as a play. This is a real thing. It is just not the thing that music has always been for.
The harmonic series is still there. The neurobiological architecture that responds to it is still there. The relationship between a specific sound made by someone who knows a specific person and the nervous system of that specific person remains the most powerful musical technology that has ever existed.
The algorithm did not discover this. It cannot discover anything. It can only optimize for what its engineers decided to optimize for.
The shepherd was not optimizing for retention. The shepherd was paying attention to what was already true.
That truth has not changed. It has only been temporarily buried under sixty thousand tracks a day and a scoring function that cannot hear the difference between a discovery and a product.
If you were writing the objective function — if the design goals were yours to set — what would you optimize for? What does the alternative actually look like? Leave it in the comments.
Tags: pentatonic scale harmonic series physical truth, recommendation algorithm commercial invention music, discovered vs invented music history philosophy, functional music neurobiological wellbeing retention, circle of fifths algorithmic opacity unilateral revision