I was listening to Kate Compton on the Approximately Correct podcast from AMII. She’s a generative artist, researcher, and self-described weird futurist, probably best known for creating Tracery and generating the planets in Spore. And she said something that cracked something open for me.
She’s not interested in the debate over whether AI can be creative, because AI lacks what she thinks is the most important part of art: social connection. Kate said this: don’t pick a weird thing to make. Pick a person. Figure out what you’d use to send them the strangest love letter.
She talked about mix tapes.
That sent me straight back to the 80s.
I used to make mix tapes.
Not playlists. Tapes. Ninety minutes, two sides, cassette in a Walkman the size of a paperback novel. Saturday afternoons American Top-40 played and you sat in front of a stereo after the Vulcan nerve pinch of Pause-Play-Record. Then you hovered your finger over the pause button, waiting for the song to start on the radio so you could catch it before Casey Kasum talked over the lead in. Or you carefully queued up all your favourite songs in your tape and CD collection you got from Columbia House. I still owe the Columbians money. You planned the order in your head for days. You rewound to check timing. You hand-wrote the track list on the little card that came with the blank tape, in your best printing, because it mattered.
The music wasn’t mine. I didn’t write “The Immigrant Song.” I didn’t play a single note of “Sledgehammer.” I didn’t compose anything.
But the mix tape was mine.
Every choice was mine. The opener that set the temperature of the whole thing. The sequencing. The way a furious song bled into something aching and slow, and that contrast created a feeling that existed in neither track alone. The side break, which was itself a choice. The closer, which had to land right or the whole thing felt unfinished. The intentional misfit. The song that didn’t belong by any obvious logic but fit in a way only you and maybe the recipient would understand.
Metallica followed but the Crash Test Dummies was a choice. I stand by it.
Compton’s core argument is this: creativity is a verb, not a noun. It’s something you do, not something you are. And the mix tape proves it. You didn’t need a music degree. You didn’t need a studio or an instrument or a record deal or anyone’s permission. You needed a blank tape, a stereo, and someone worth making it for. The mix tape democratized creative expression before anyone had a word for that. It said: you don’t have to be a professional creative to make something that matters.
The Creativity Question
I’ve made the tool argument myself. AI generates output, humans evaluate it, therefore humans are still creative. Tool as prosthetic, human intent as animating force.
It holds until it doesn’t. When the tool starts generating the intent too, what’s left that’s yours?
The mix tape metaphor cuts through the noise because it separates two things we keep conflating: composition and curation. I didn’t compose the songs. But I curated the experience, and that curation was a creative act with real stakes, real judgment calls, real risk of failure. The tape could be bad. My taste could be wrong. The sequence could land flat. The misfit track could just be a misfit. Those were real creative risks I was taking.
The creativity lives in the choosing and the ordering. The technical skill of the recording is almost beside the point. What matters is the human intent behind it.
The question for the AI age isn’t “did a human make this.” It’s: where did the consequential choices get made, and by whom.
What Made a Great Mix Tape
It wasn’t taste alone. Taste is necessary but not sufficient.
What made a great mix tape was terrain knowledge. You knew the recipient. You knew what they needed to hear at 2am on a Tuesday. You knew which song would undo them and which one would put them back together, and you knew to sequence them in that order, and you knew to put the one they didn’t know they needed between those two.
The mix tape is a strange love letter. You’re not picking songs. You’re picking for someone. With all the history, the unsaid things, the specific texture of that relationship written in every track. That knowledge isn’t transferable and it isn’t generatable. It lives in you, accumulated through years of paying attention to a specific human.
You knew their history. You knew what was unsayable between you. Sometimes the tape said it anyway. Sometimes the tape said something you couldn’t say out loud, and you handed it over and watched their face, and they understood.
Spotify can identify that you listened to a lot of sad bastard music in November and serve you more sad bastard music. That’s pattern matching. It’s not the same thing as knowing you well enough to choose badly on purpose. Most mix tapes weren’t made from deep knowing anyway. Plenty were aspirational, or projective, or just an excuse to show off taste. That’s fine. Even thin terrain knowledge beats no terrain knowledge. The algorithm has none.
The intentional misfit is everything.
The song that shouldn’t be there that breaks the mood, that’s too fast or too slow, that’s from a band they think they hate, that’s so on-the-nose it hurts. That is the creative signature. It’s where the maker shows their hand. It’s where intent exceeds taste. The algorithm cannot put that song there. The algorithm optimizes for coherence. The maker optimizes for truth.
The mix tape didn’t have to be the greatest creative act of all time to matter. It was participation. It was showing up in a shared culture of making and giving. The bar wasn’t genius. The bar was thinking about you. That’s low-stakes enough for anyone to clear. High-stakes enough that clearing it actually means something.
The Constraint Was Generative
The modern playlist doesn’t have constraint. Mix tapes did.
Ninety minutes. Two sides. Songs you had to record in real time. The blank tape cost you three dollars. Redoing a side because you botched the timing cost you another hour and sometimes another tape. Every choice had weight because every choice had cost.
We talk about creativity like abundance enables it. More options, more tools, more output, more better. But the mix tape suggests the opposite. The constraint wasn’t a bug. The constraint forced decisions. Every song that made it onto that tape displaced something else. You couldn’t just add it all. You had to choose, and choosing meant committing, and committing meant you were accountable to the thing you made.
Infinite AI output at zero friction is the enemy of that. Not because AI is bad, but because the friction was doing something important. The resistance was where the judgment happened. Where you got better at knowing what you actually thought.
Dave Snowden frames it as the Apprentice Model: automate past the point where humans can do the work manually and you lose the ability to supervise the thing doing it for you. The draft skills atrophy, and atrophied skills make you stupid to when the AI is wrong. The mix tape taught you what a good sequence felt like in your bones. You learned to recognize a bad one.
Even the cover mattered. You drew on it. You picked the one of the four colors on your fat Bic pen for the track list. You added a caption or an inside joke. Ownership through marking, nothing really, technically speaking, but enough for the maker to feel genuine creative agency over the thing. The doodle on the J-card wasn’t decoration. It was a signature. It said: I made this. I made it for you.
You can’t learn terrain from maps. You can study maps until you can recite every contour line. Then you go out there and misjudge a root underfoot, and you learn painfully something the map cannot tell you.
The HITL That Actually Matters
Human-in-the-loop has become a governance term. A safety layer. Someone checks the AI output before it ships. Important. Not unimportant. But fundamentally defensive.
The mix tape reveals something different: the human isn’t in the loop to catch errors. They’re in the loop because the loop is where meaning gets made.
The consequential choices in a mix tape happen inside the maker. Listening, comparing, second-guessing, feeling something shift when the order works, abandoning an idea you were attached to because the tape tells you it’s wrong. That inner process is the loop. The tape is just the output.
The tighter the loop, the more you learn, the more your intent sharpens. Automate it and your intent thins out. Fully close the loop: AI selects, AI sequences, you approve– and the “thinking about you” connection evaporates. The recipient stops feeling thought-about. They feel processed.
AI is inductive: it finds patterns in past data. Human creativity is abductive: it leaps to a hypothesis the data doesn’t contain.
The mix tape was an abductive act. You weren’t extrapolating from the recipient’s listening history. You were making a leap: this song, right here, even though nothing predicted it.
You cannot automate abduction. You can only automate induction.
Confusing one for the other is a slow way to hollow out your soul.
The loop has to be real. A human rubber-stamping AI output at the end of a conveyor belt isn’t a human in the loop. It’s theater at best. You have to have done enough of the manual work to know what good feels like before you hand it off.
If you said “make me a playlist for a breakup” and published what came back, that’s not a mix tape. That’s Spotify with extra steps. The difference isn’t whether AI was involved. The difference is where the judgment lived.
What Creativity Is For
If it’s for producing novel shiny output: AI wins. Unambiguously, at scale, faster, cheaper, tireless. If maximizing the quantity of coherent, useful, well-organized creative product is the goal, AI solves that problem and the humans can go do something else.
But that’s not what creativity is for.
The making does something to the maker. Transformation is the point, not the byproduct.
You’re different after you’ve made a great mix tape than you were before. You know something about the recipient you didn’t know when you started. You know something about yourself. What you were reaching for, what you were afraid to say, what you finally said anyway. The process revealed something the tape alone can’t fully express.
The long slow education of learning to pay attention to specific humans, specific moments, specific things that are true and hard to say. You can’t shortcut it. You can’t compress it. Every mix tape you made, good or bad, taught you something the next one could use. The terrain accumulates. The draft skills build. And when the algorithm fails, you’ll know. Because you’ve been in there, doing the actual thing, long enough to feel it.
When you offload the consequential choices. When the AI selects the songs and determines the sequence and you approve it. You get the tape but not the education. The output looks the same. The maker is unchanged.
You’re a Maker if You Make Choices
I didn’t compose a single song on any mix tape I ever made. Never played an instrument, never wrote a lyric. Still made things that mattered to people. Still made things that said what I couldn’t say out loud. Still took creative risks that could fail. Still learned something every time, good tape or bad.
The creativity wasn’t in the composition. It was in the caring act of curation. The choosing, the sequencing, the marking, the giving. The judgment. The terrain knowledge applied under constraint with something at stake. None of that required a credential. It required showing up and giving a shit long enough to think about someone else.
That’s still available to you. Right now. With whatever tools you have.
The blank tape became a CD-RW became a streaming platform became an AI. What hasn’t changed: somewhere in the process, a bona fide meatspace person has to be doing the creative action. Making consequential choices from a position of genuine knowing. Knowing the terrain, knowing the recipient, knowing what it means when something doesn’t fit but you put it there anyway.
When that’s present, the human soul made the thing, regardless of what tools were involved.
When that’s absent, no tool in the world fills the gap.
The algorithm will never put the wrong song in on purpose. That’s your job. That’s always been yours.
This post was sparked by Kate Compton’s appearance on the Approximately Correct podcast from the Alberta Machine Intelligence Institute (AMII), published March 2026. Compton is a generative artist, researcher, and Assistant Professor at Northwestern University. The Snowden threads draw on his 2025–2026 AI writing series — “Algorithmic Induction,” “Anthropomorphising Idiot Savants,” “A New Animism,” and others — collected at thecynefin.co. Both worth your time.
