The Sound That Stays: On Games, Opera, Drums, and What AI Will Never Touch

I need to tell you about a feeling I can't name.

It started with a video game. Which, I know, sounds ridiculous. Bear with me.

I've been gaming since before most people reading this had email addresses. I've got a retro console collection that takes up an entire room and a relationship with Unreal Engine that's lasted longer than some marriages. I've played thousands of games. Some I loved. Some I forgot by the next morning. But Clair Obscure: Expedition 33 did something none of them ever did.

It haunted me.

Not the gameplay, which is brilliant. Not the story, which is genuinely moving. The soundtrack.

Lorien Testard, a French composer who taught guitar for a living before this, spent five years writing 154 tracks for this game. He was joined by vocalist Alice Duport-Percier, about thirty other musicians, and a nine-person choir. The music moves between orchestral opera and heavy metal, between jazz and ambient soundscapes, between heartbreak and defiance. Testard examined the emotional core of every battle, every location, every character. He used leitmotifs the way Wagner did. Classical techniques. Real musicians. Real voices. Real instruments blending with electronic production.

And it hit me in a place I couldn't identify.

I've had the soundtrack on loop for weeks. Not as background noise. As something I keep returning to because it makes me feel things I don't have words for. The game itself is scaffolded on art culture and art history. Its world draws from Impressionism and the Belle Époque. Playing it sent me down rabbit holes about art movements and creativity and what makes something beautiful versus merely competent. I started questioning things I thought I'd settled a long time ago. About what art is for. About why certain sounds make your chest tighten and your eyes sting while other, technically identical sounds, leave you completely cold.

That question turned out to be the most important question I've asked in years.

My son is a drummer.

I want you to sit with that for a second, because there's something most people don't think about when they picture a kid behind a drum kit. They think about noise. They think about the neighbours. They think about, God help them, the school concert.

Here's what they're missing.

Percussion is arguably humanity's oldest form of musical expression. Archaeological evidence puts drums back at least six thousand years, to the cusp of the Neolithic period in China. Hollow logs with animal skins stretched over them. But the instinct, the urge to hit something rhythmically, to create patterns of sound with your body, goes back far further than that. Researchers have found bone instruments dating back 40,000 to 50,000 years. Before agriculture. Before pottery. Before written language. Before civilisation as we understand it, humans were making rhythm.

Think about what that means.

When my son sits down at his kit and locks into a groove, he's doing something his species has been doing since before we built the first wall. The drumbeat is older than the wheel. It's older than religion. It's older than every institution, every technology, every system we've ever created. It is, in the most literal sense, one of the first things that made us human.

Mihaly Csikszentmihalyi spent decades studying what he called "flow state", that condition where a person becomes completely absorbed in an activity, where time dissolves and self-consciousness evaporates and you're just... there. Present. Performing. His research found that musicians experience flow more readily than almost any other group. The specific conditions of musical performance (the balance of challenge and skill, the immediate feedback, the merging of action and awareness) create one of the deepest states of human consciousness available to us.

I've watched my son enter flow state on the drums. His face changes. His body changes. He's not thinking about technique or timing. He IS the rhythm. And there's something ancient and electric about watching it happen. You can feel it in the room. The air shifts.

Every civilisation on Earth, without exception, developed percussion instruments. The djembe in West Africa. The taiko in Japan. The bodhrán in Ireland. The tabla in India. These instruments evolved independently on every continent because the need to create rhythm isn't cultural. It's biological. It's written into whatever we are at the deepest level.

My son isn't just learning an instrument. He's plugged into a current that runs back through tens of thousands of years of human experience. And that current is the thread I want to pull on for the rest of this piece.

Here's something that took me an embarrassingly long time to figure out.

Full disclosure: I'm a classically trained guitarist. Flamenco, theory, the whole curriculum. I got accepted into the Sydney Conservatorium of Music as a young man. My son plans to study there now. But back then, my family preferred I find a "more stable career." So I became a teacher. And I spent the next twenty years burying the classical training under grunge, metal, and rock. Stone Temple Pilots, Soundgarden, Karnivool, Living Colour, 311, Tool. If it didn't have distorted guitars and someone screaming about something, I wasn't interested. Classical music was for old people. Opera was for really old people. The kid who got into the Con became, to put it generously, a musical snob in the wrong direction.

Then I started paying attention to the actual architecture of the music I loved.

Karnivool's "Themata" builds through shifting time signatures and layered polyrhythms that Bach would have recognised as counterpoint. An Australian prog band from Perth doing things with musical structure that mirror 300-year-old compositional techniques. Living Colour's Vernon Reid plays jazz-fusion-metal guitar that carries a direct line from Paganini through Hendrix to something that shouldn't work but absolutely does. 311 blend reggae, funk, and rock in ways that echo the same cross-pollination that created jazz from ragtime and blues.

This isn't coincidence. It's genealogy.

Stone Temple Pilots' Scott Weiland could shift from a whisper to a roar in the same phrase, bending notes the way an opera singer bends vibrato. That's not taught in rock schools. That's instinct shaped by a tradition stretching back centuries. Dream Theater build entire albums using intricate time signatures and thematic development lifted directly from symphonic composition. Deep Purple's Ritchie Blackmore reworked Beethoven's "Ode to Joy" into a rock arrangement. The structures are the same. The amplification changes. The hair changes. The feeling doesn't.

And the deeper you go, the more the through-line reveals itself.

Blues came from African American work songs and spirituals. Rock came from blues. Punk came from rock. Grunge came from punk and metal. Metal itself borrowed its dark, dramatic sensibility from the same emotional tradition as Wagner and Mussorgsky. The tritone, the interval that defines Black Sabbath's heaviest moments, was called "the Devil's Interval" by medieval church musicians who considered it too unsettling for sacred music.

There is an unbroken thread from Puccini to Pearl Jam. From Vivaldi to Van Halen. From a Neolithic drummer striking a hollow log to my son in his bedroom at three in the afternoon, disappearing into the beat.

Classical structures. Blues feeling. Grunge honesty. Opera drama. They're not separate traditions. They're one tradition, evolving, mutating, passing from hand to hand across centuries and continents. The instruments change. The amplification changes. The hair gets longer, then shorter, then longer again. But the thread doesn't break.

And it matters that the thread is human.

Let me tell you about December 30 in Sydney.

I took my extended family to the Opera House. Australian Opera. Full house. Full production. My wife's whole family had come over from China and we brought everyone along. I remember looking around the Concert Hall before it started and feeling unreasonably happy. Families everywhere. Kids fidgeting in their seats. Grandparents settling in. This wasn't my usual scene. My grandmother loved opera. My parents took me to see Phantom of the Opera as a kid. But opera proper? I was, at best, curious. At worst, doing the "family culture" thing where you go because it seems like the right thing to do and secretly check your phone during the slow bits.

Within ten minutes, I forgot I had a phone.

Here's the thing nobody tells you about live opera. It's not just the singing. It's everything around the singing. The Sydney Opera House Concert Hall has famously imperfect acoustics. There's a whole history of renovations and acoustic panels and arguments about sound diffusion that spans decades. Engineers have spent over $200 million trying to fix it. And those imperfections? They're part of what makes it extraordinary.

I could hear footsteps. Tiny creaks. The subtle rustle of an audience collectively holding its breath. Things that would be engineered out of any recording, things that a sound engineer would consider flaws, became part of the experience. They reminded you, at every moment, that this was happening live. Right now. Right here. Humans in a room, together, sharing something that would never happen exactly this way again.

The vocalists were incredible. Not because they were technically perfect, although they were close. Because their voices carried something beyond technique. You could hear lived experience in their phrasing. Joy. Loss. Years of training, sure, but also years of living. Heartbreak and triumph and mundane Tuesday mornings and all of it somehow compressed into how a vocalist bends a note or lets a phrase breathe a half-second longer than expected.

I'm a musician. I've been to hundreds of concerts and festivals. I've seen acts that blew my mind. But I don't cry at concerts. I just don't. I'm not that guy.

Then they performed Nessun Dorma.

"Nessun dorma" is an aria from the final act of Puccini's Turandot. Puccini died before he could finish the opera. It was completed by Franco Alfano. The aria itself is about three minutes long. The prince, Calaf, sings of his certainty that he will win the princess by dawn. "Vincerò" is the final word. "I will win."

But it's not about winning. Not really.

Puccini composed it knowing he was dying. There's a paradox running through all of his work (the distance between what we desire and what we can have) and in Nessun Dorma that paradox becomes almost unbearable. The music builds and builds and builds. The tenor sustains a B4, then drops to an A4, and holds it. The whole room holds it with him. You're not listening anymore. You're inside it.

Luciano Pavarotti made this aria famous beyond the opera world when the BBC used his 1972 recording as the theme for the 1990 FIFA World Cup. Somehow, an aria about a prince trying to win a cold princess's heart became an anthem for football. "Vincerò" stopped meaning "I will win her hand in marriage" and became "I will win three-nil." The Three Tenors performed it at four consecutive World Cup Finals. It reached number two on the UK singles chart. An aria written by a dying Italian composer in the 1920s became the most recognisable piece of classical music on the planet.

But hearing it recorded and hearing it live are different things separated by an ocean.

That night at the Opera House, when the tenor hit the final "Vincerò!", I felt something crack open in my chest. Tears. Actual tears. Not a polite misty eye. The kind of tears that ambush you, that you fight because you're sitting next to your family and you don't want to be the guy crying at the opera but you can't stop it and honestly you don't want to stop it because the feeling is too real and too big and too important to shut down.

My son was sitting next to me. He felt it too. I could tell.

The drive home took forty-five minutes. We spent every second of it trying to explain to each other what had just happened. Trying to break down every detail. The acoustics. The imperfections. The footsteps. The way the tenor's voice didn't just reach us but went through us. The way two thousand people in a room can somehow become one thing, one organism, breathing together, crying together, experiencing something so profoundly shared that it dissolves the boundaries between strangers.

We couldn't explain it. We tried and failed and tried again.

And that failure to find the words was the proof.

Here's what I keep coming back to.

Neuroscientists at Western University in Canada discovered that during live musical performances, audience members' brainwaves synchronise. Not metaphorically. Literally. Their neural oscillations align. A 2024 study published in Scientific American confirmed it: the greater the degree of synchrony between audience members' brains, the more they enjoy the performance. Your brain waves start matching the brain waves of the stranger sitting next to you. You become, at a neurological level, connected to people you've never met through the shared experience of live music.

A 2022 study on "Audience Interbrain Synchrony During Live Music" found that this coupling is shaped by both the number of people sharing the experience and the strength of the emotion. More people feeling it deeply equals stronger synchronisation. The room doesn't just sound different when two thousand people are emotionally engaged. It IS different. The physics of the experience changes.

This is what was happening at the Opera House. This is why it couldn't be replicated by headphones. The experience wasn't just auditory. It was neurological, physical, spatial, social, temporal. It was alive.

And here's where it connects to something I can't stop thinking about.

Mirror neurons fire both when we perform an action and when we observe someone else performing it. Research from Oxford shows that when you listen to a vocalist, your brain activates a motor network similar to the one the singer is using. You're not passively receiving the sound. You're internally simulating the singing. Your brain is, in a very real sense, singing along. And those tiny imperfections (the breath that catches, the note that bends under the weight of emotion, the fraction of a second where the vocalist's lived experience bleeds through the technique) that's where the empathy lives. That's where the mirror neurons go into overdrive.

Perfection doesn't trigger this. Imperfection does.

A flawless, algorithmically generated vocal line doesn't carry the signature of a life lived. It carries the signature of a pattern matched. There's no breath to catch because there are no lungs. There's no emotion bleeding through because there's nothing to bleed.

Let me be honest about AI, because I spend my professional life working with it and I'm not interested in pretending it doesn't do impressive things.

Suno, one of the leading AI music generation platforms, is now valued at $2.45 billion. It raised $250 million in its latest funding round. It settled a $500 million lawsuit with Warner Music Group, gaining the right to train its models on Warner's entire catalogue. Udio, built by ex-Google DeepMind researchers, has generated vocals that in blind tests have fooled professional producers. You can type a sentence describing the kind of song you want and have a complete, produced track in under a minute.

I've used these tools. Some of the output is genuinely good. Catchy melodies. Decent production. Competent.

There's that word again. Competent.

The gap between competent and transcendent is everything. And AI lives on the wrong side of that gap.

AI music generation works by pattern matching across enormous datasets of human-created music. It learns statistical relationships between notes, rhythms, timbres, structures. It recombines them into novel arrangements that sound plausible because they are, statistically speaking, plausible. They conform to the patterns humans have established over centuries.

What AI cannot do is deviate from those patterns with purpose.

When Lorien Testard composed the soundtrack for Clair Obscure, he wasn't just combining notes. He was drawing on a life spent falling in love with Joe Hisaishi's Studio Ghibli scores. He was channelling the French artists he absorbed growing up. He was responding to a director who told him, "You arrive at the beginning of the game after the heroes have already lost," and letting that emotional reality shape every chord. He spent five years with this music. It changed him, and that change is audible.

When the tenor at the Opera House sang Nessun Dorma, he wasn't executing a vocal line. He was singing about winning while carrying every loss he'd ever experienced in the timbre of his voice. The audience didn't synchronise their brainwaves to his technical accuracy. They synchronised to his humanity.

When my son plays drums, he's not reproducing rhythmic patterns. He's channelling something that runs through his species like a river runs through a landscape. Something that was there before language. Before thought. Before anything we'd recognise as culture.

This is what AI cannot do. Not "yet." Cannot. The lived experience that travels from a vocalist's lifetime to an audience's nervous system. The tens of thousands of years of cultural memory encoded in a drumbeat. The tiny imperfections that announce to every mirror neuron in the room: this is real. A human made this. A human is here.

In February 2026, Mrinank Sharma resigned from Anthropic.

If you don't know Anthropic, they built Claude. One of the most advanced AI systems on the planet. Sharma led their safeguards research team. He spent two years working on AI sycophancy, defences against AI-assisted bioterrorism, safety cases. His final project was studying how AI assistants could "make us less human or distort our humanity."

He didn't leave to join another AI company. He didn't leave to start a competitor.

He left to study poetry.

Let that sit for a moment.

The guy responsible for making sure one of the world's most powerful AIs doesn't harm people looked at the technology he helped build and concluded that what's missing is art. His resignation letter said: "I feel called to writing that addresses and engages fully with the place we find ourselves, and that places poetic truth alongside scientific truth as equally valid ways of knowing."

Poetic truth. As valid as scientific truth. From an AI safety researcher with a PhD.

He quoted William Stafford:

There's a thread you follow. It goes among
things that change. But it doesn't change.

That thread. The one running from the Neolithic drummer to my son's bedroom. From Puccini's deathbed to the Opera House on a Tuesday night. From a French composer teaching guitar in his living room to a soundtrack that won Game of the Year and haunts me still. The thread doesn't change. The instruments change. The technology changes. The thread stays.

Sharma saw it. And he walked away from one of the most prestigious positions in AI to follow it.

Here's the part that makes me angry.

In Australia right now, the Job-Ready Graduate scheme has doubled arts fees. A humanities degree costs $17,399. A maths degree costs $4,738. Forty-eight creative arts degrees have been axed since 2018. Year 12 arts enrolments have collapsed by 21%. The government invested $75.6 million in STEM education. The amount invested in arts education? Zero.

Zero.

This is happening while the World Economic Forum's 2025 Future of Jobs Report ranks creative thinking as the number one most important skill for 2025 to 2030. While Harvard identifies creativity, meaning-making, judgement, and emotional intelligence as uniquely human capabilities. While Nvidia's 30,000 engineers tripled their code output using AI tools and then got redirected toward, and I'm quoting here, "challenges requiring human ingenuity."

We are systematically defunding the exact skills that the AI economy values most.

The world is drowning in AI-generated content. Text, images, music, video. The stuff is everywhere. It's technically adequate and emotionally empty. The scarce resource is no longer production. Anyone can produce now. A teenager with Suno can produce a song in a minute. The scarce resource is taste.

Taste. The ability to look at two things that are technically identical and know, in your bones, which one has a soul and which one doesn't. The ability to hear a vocal take and say "that's good" versus "that's the one." The ability to curate, to select, to feel the difference between adequate and transcendent.

Taste is an arts skill. It's developed through exposure to great art, through practice, through conversation, through the kind of education we're pricing people out of.

In the higher education report I published last month, I identified seven capabilities that AI cannot replace: judgement, taste, accountability, emotional intelligence, practical application, creative direction, and ethical reasoning. Every single one of those capabilities is developed more effectively through arts and humanities education than through any other pathway. And we're cutting the arts while pouring money into STEM as if the future belongs to people who can do what machines already do better.

As I wrote in that report: the skills being defunded are the skills AI can't replace. It's like building a car with a bigger engine and removing the steering wheel.

I want to be clear about something, because I know how this reads so far, and I don't want to be misunderstood.

This is not an anti-AI argument.

I use AI every day. I build with AI. I help organisations integrate AI. I've written books about it. AI is extraordinary. It's the most powerful amplification tool humans have ever created.

Clair Obscure: Expedition 33, the game that started this whole train of thought, was built using modern game development tools that include AI assistance. The art direction, inspired by Impressionism and Belle Époque aesthetics, was realised through technology that would have been impossible a decade ago. The game reached millions of players through digital distribution platforms powered by machine learning algorithms. AI helped this piece of art exist.

But it didn't create the art.

Lorien Testard created the art. Alice Duport-Percier created the art. Thirty musicians and a choir created the art. They brought their lives, their training, their taste, their humanity. AI helped them realise their vision at a scale and speed that wasn't previously possible. That's the partnership. That's what "AI-first" should actually mean.

AI can help a composer iterate faster. It can help generate variations, handle production tasks, manage distribution. It can analyse what's working and suggest optimisations. It can help a game studio create stunning visual environments. It can transcribe, translate, enhance, amplify.

But the soul comes from the human. Always.

This is what the collaborative future looks like. Not humans versus AI. Not AI replacing humans. Humans doing the thing only humans can do (feeling, curating, tasting, creating from lived experience) and AI handling the production, the distribution, the iteration, the scale. Together, they create things neither could alone.

But only if we understand what each does best. And only if we invest in developing the human side with the same urgency we invest in developing the machine side.

Right now, we're not even close.

The IBM Global Skills Report found that "willingness to be flexible, agile, and adaptable to change" is the number one most critical skill for the workforce. Not coding. Not data analysis. Adaptability. Deloitte redesigned their entire hiring process to assess it. Goldman Sachs screens for it. The formal research calls it AQ. Adaptability Quotient.

But here's what nobody in the AQ conversation talks about. The most adaptable people I've ever met are artists. Musicians who can read a room and shift their setlist. Actors who can inhabit a different human being every night. Painters who can throw out a canvas and start again because they have the taste to know it's not working and the courage to begin fresh.

The unlearning dimension of adaptability (letting go of what you think you know) is the thing artists practice every single day. It's the hardest capability to develop. And it's the one we're cutting from the curriculum.

I keep thinking about that drive home from the Opera House.

My son and I, in the car, trying to find the words. Two people who are not short on words (me professionally, him by genetics) completely unable to articulate what we'd just experienced.

We tried metaphors. We tried technical analysis. We tried comparing it to other concerts. Nothing worked. Every sentence felt inadequate. Every explanation fell short of the experience.

And at some point, I stopped trying.

I stopped trying because the failure to explain it was the most honest thing either of us said all night. The fact that words couldn't capture it was the proof that it was real. If you can fully describe an emotional experience, the experience probably wasn't that deep. The things that truly move us live in the space between language and feeling. They're felt, not said. They're shared, not explained.

That's what music does. That's what live, human, imperfect, breathtaking art does. It takes you to a place that language can't reach. And it takes you there together, with a room full of strangers whose brainwaves are literally synchronising with yours, whose mirror neurons are firing in sympathy, whose breath is held alongside your breath.

No AI will ever do that. Because to do that, you need to have lived. You need to carry loss and joy and Tuesday mornings and sleepless nights and the memory of your grandmother who loved opera and the afternoons watching your kid disappear into the drums. You need a body that breathes and a voice that breaks and a lifetime of being human packed into every note.

You need the thread.

The one that runs from a hollow log in the Neolithic to a drum kit in a suburban bedroom. From Puccini scribbling notes while his body failed him to a tenor at the Opera House bringing those notes back to life a century later. From Lorien Testard posting tracks on SoundCloud from his living room to a soundtrack that made a grown man with a retro gaming collection and an Unreal Engine habit sit in the dark and feel things he couldn't name.

The thread doesn't change. Everything else changes. The thread stays.

And if we're smart, if we're brave, if we actually care about what comes next, we'll protect it. We'll fund the arts instead of gutting them. We'll send our kids to music lessons and drama classes and not just coding bootcamps. We'll go to the opera, not because it's cultured but because it's the most technologically advanced emotional experience humans have ever created. We'll listen to the AI safety researcher who left to study poetry and hear what he's actually saying: that the thing we've been building all this technology for, the thing that makes it all worth building, is the human experience on the other side.

AI can do incredible things. Humans can do transcendent things. And together, if we get the balance right, we might just create a world worth living in.

But only if the soul stays human.

Only if the thread holds.

My son is probably playing drums right now. Somewhere in the house, there's a beat that started forty thousand years ago, moving through him, connecting him to every human who ever picked up a stick and hit something and felt the universe answer back.

I still can't explain what happened at the Opera House. I don't think I'm supposed to.

The soundtrack is still playing. Nessun dorma. Let no one sleep.

Vincerò.

The Clair Obscure: Expedition 33 soundtrack is available on Spotify, Bandcamp, and wherever you listen to music composed by actual humans. Nessun Dorma is best experienced live. Go to the opera.