Over the weekend, my family and I finally watched the most popular film in Netflix history: KPop Demon Hunters. For context, by the end of 2025 it had been watched more than 500 million times. On paper, the film is flawless: a fantastical, modern mythological story with catchy music, tantalizing visuals and an empowering message about self-acceptance, belonging, connection and… I hated every minute of it. But why? I dug for the words to articulate my intuition.
I started by looking into one of the core elements of the film: the music. I learned that an estimated 80 percent of K-pop songs are written by a small group of professional songwriters in the US and Europe, many of them, oddly, Swedish. I also learned that music labels test “challenge videos” on social media before a song drops to measure traction in an audience. The songs themselves are built around optimized “killing parts”, or hooks of 5-15 seconds, manufactured for short-form video platforms such as TikTok. We are trending toward a world where music (or any other media, for that matter) gets produced only after a critical threshold of the audience has voted for it. This gives new meaning to the term “popular”, not as an output, but as an input.
Then I looked at storytelling, a craft close to my heart as a service designer, the second core element of the film. In 2016, researchers at the University of Vermont ran more than a thousand novels through sentiment analysis and found that the stories readers loved most followed a handful of emotional arcs. In 2020, another team did the same thing for film, tying those arcs to box office returns. And for years, a now-dissolved consultancy called Epagogix sold a neural network that scored scripts for commercial value before a film studio made the decision to finance it, or not. This deconstruction process is limiting the diversity of music and narratives practically available in our culture, and the evidence is uncomfortable to ignore: a 2012 study of nearly half a million recordings found that Western pop music has grown measurably more homogeneous since 1955, with narrower chord progressions and a shrinking palette of sounds. True artistry is becoming the underdog (which gives me hope).
I was surprised to learn that most of this wasn’t new. For example, “Tin Pan Alley” put songwriting on a production line as far back as the 1890s when staff writers wrote sheet music to spec. And the studio system did the same thing for film. There are now around a dozen canonical storytelling frameworks used in commercial production studios. The machinery of mass media is truly impressive, but it’s hollow.
I paused to reflect, and realized something disturbing: the historical direction of the feedback loop (moving from artist to expression to audience and back again) is being inverted by this machinery. The Tin Pan Alley writer took an educated guess, published, and found out after the fact whether the public liked their song. For as long as these art forms have existed, an artist sat in between the audience and the production itself (be it film or music, or any other medium): that artist had taste, conviction, and, importantly, the freedom to choose. Today, however, the loop is running in reverse. Specialized companies measure audience response first (the “challenge video”, adherence to storytelling formulae, script scores, predicted box office performance and so on) and the work is assembled backwards to fit the expectations produced by the algorithm, informed by historical data on “what works”. The audience’s appraisal is no longer just a judgment. The audience has bizarrely become the author of the work itself, like a bad sci-fi movie in which the main character has become their own father and mother.
This inverted feedback loop explains why a “flawless” movie felt so intuitively wrong to me. Human judgment naturally leaves fingerprints, residue, like scenes that run too long, surprising rhythmic shifts, or obscure references that only some nerds will catch. Optimization scrubs that gunk, that soul, because every bit of it may cost a point of engagement. The movie was too perfect, too refined.
Ironically for a story about saving souls, I discovered that KPop Demon Hunters has no “soul” of its own. And by “soul”, I mean the presence of a human’s free judgment: the sense that a person stood behind this thing and could have chosen an infinite number of alternatives. No doubt this movie is a labor of love, but the machinery and tools for modern filmmaking buried that love. Formulae are followed for maximum return and minimum variability or risk: no “soul”.
One of the first scenes in the movie vividly captures my point. Aboard their fancy plane, the HUNTR/X trio confronts their demonic flight attendants who betray their lack of humanity through awkward, programmed-sounding dialogue and incorrect actions. The demons are defeated and are cast back from whence they came, still hungry for the taste of a soul. This movie is that, except its disguise isn’t clumsy: it’s perfectly engineered and assembled from analytics.
Ultimately, this movie amounts to a manufactured simulation of the authentic connection humans are supposed to get by interacting with their kin/community. In other words, the opportunity cost of watching KPop Demon Hunters is having those real world experiences depicted in the movie (vulnerability, grief, betrayal, self-sacrifice, anger, denial, acceptance, and love) for yourself. Why rely on your kin/community if you can get your needs met more efficiently by a movie?
What’s striking is just how perfectly this mirrors the “industrialization of experience” today. Service and experience design began as a way to help organizations see their customers as people. Corporations now use it to map journeys, find the “moments of truth” that move metrics, engineer them, and then repeat at scale. Bigger, bigger, bigger! When a policy comes from a dashboard, the person serving you is performing a “killing part”. And I know firsthand because I have drawn those maps myself.
In a parallel way, corporations have to simulate connection with customers because their feedback loops are (dis)located in dashboards. A local business’s feedback loop, on the other hand, is a human with a face and a story: the owner who welcomes back a regular who has not been around for two weeks does not need a retention curve; they are the retention curve. That judgment, standing in the room where the customer is, cannot be manufactured at any scale, which is exactly why so much money goes into simulating it. The only way to lose it is to hand it over: to buy the industrialized version of your own judgment because somebody sold it as a best practice.
Keep the judgment in the room. And slay a few “experience” demons while you’re at it.