The Robot Sang, I Told the Truth, and Nobody Ever Called Me Again

I want to open with a professional humiliation, because I’ve found it puts students at ease.

A while back I pitched a vocal track for a show on a prestige cable network I won’t name. The track was good. People used the word “haunting,” which in my line of work is the equivalent of a tip. Things were moving. Then somebody asked how I’d gotten that vocal, and I did the single dumbest thing a person can do in the entertainment industry, which is tell the truth.

“It’s AI,” I said.

I have not heard from those people since. Not a no. Not a “we went another direction.” Just a silence so total that I’ve started to wonder whether I imagined the entire network.

The robot cost me a gig. I am, in the technical sense, a victim. And I still think you should learn every nook and cranny of this stuff. Here’s why.

Some context, briefly

Both my parents were professional musicians in Chicago. My father played trombone; my mother played French horn. Music has been an essential part of my life since before I was born, which is not a figure of speech — I attended a lot of rehearsals I have no memory of. I’ve spent the eighteen years since then designing music for video games, in rooms with composers, orchestras, and grown adults crying at scoring sessions, a category that includes me more often than I’d like to put in writing.

So when I tell you I have gone all the way in on AI — Suno, Claude, the whole strange menagerie of things that will write you a bossa nova about your dishwasher — understand that I did it the way you’d learn the floor plan of a building that’s on fire. Know the enemy. Find the exits. Figure out, specifically and not vaguely, what it can and cannot do. Because “it’s probably bad” is not a plan. It’s a mood.

What I found in there

Here is the thing nobody tells you: these tools are unbelievably powerful, and they are also, at a fundamental level, lazy in a way that should be enormously comforting to you.

AI can produce something insanely creative. Never-been-done. I’ve heard it. It has made me sit up in a chair. But — and this is the whole article, so if you’re skimming, this is the paragraph — it does that only if you tell it to. It does not wake up wanting to. It doesn’t get the itch. It has never, not once, lain awake at 2 a.m. because a chord change wouldn’t leave it alone. It takes no pleasure in having made a thing. (Yet. I’m told “yet” is doing a lot of heavy lifting in a lot of sentences these days.)

That’s it. That’s your job description for the next forty years. The wanting. The taste. The knowing-which-of-these-eleven-is-actually-good. The ability to say “no, not that, the other thing, the thing that hurts a little.” The machines have become shockingly good at execution and remain, as far as I can tell, entirely without appetite. If you have appetite — and you’re in a music program, so I’m going to assume you didn’t enroll for the parking — you are holding the scarce resource.

Your objections, which are correct

Let me take these head-on, because you’ve thought about them and you deserve better than a shrug.

The training data. Yes. Suno and its cousins learned from an ocean of music that nobody agreed to teach from. That’s real. The lawsuits are real. Do not let anyone talk you out of being angry about it, and especially don’t let me. I’ll only point out that being angry about a tool and being ignorant of it are two different positions, and only one of them helps you in the meeting where somebody decides whether your job still exists.

The environment. Also yes. These things run on electricity at a scale that makes a Marshall stack look like a nightlight. I am not going to stand here and pretend a datacenter is a hemp tote bag. I will say the energy cost of a prompt is falling fast, and the carbon cost of me flying to Los Angeles to be ignored in person was not zero either.

Artists. This is the one that keeps me up. Some work is going to disappear — the “make me a loop that sounds like a tense hallway” tier of work. I did that work. I was paid for that work. It’s going somewhere and I won’t insult you by saying otherwise. What I can tell you is what has happened in every room I’ve been in for eighteen years: the person who understood the tools wrote the brief, and the person who didn’t received the brief. You get to choose which one you’ll be. It’s a real choice. You just have to make it with your eyes open.

The part where it gets good

Here’s what actually changed for me, and it’s the thing I want you to steal.

I have ADHD. At any given moment I have roughly four hundred ideas, of which approximately eleven are good and maybe two deserve to exist. The problem was never having ideas. The problem was that the idea and the prototype were separated by six weeks of work, and by week two I’d had three hundred new ideas and was off somewhere else being brilliant about nothing.

Now the idea becomes a prototype in minutes. Not a good prototype. A prototype. Enough to hold up to the light and ask the only question that matters: does this have legs? Most of the time it doesn’t, and I find that out in an afternoon instead of a season. And the one time in ten it does, I’m standing there with a working thing I can hand to someone — which is, I cannot stress this enough, a very different conversation from “so I have this idea.”

I think that’s the next wave, and I’ll put my name on it: personal apps. Not products. Not startups. Not a pitch deck. Little pieces of software built by one person, for one person, to fix one specific irritation, in roughly the time it takes to make a pour-over.

Proof, or something like it

I’m going to prove it, because the guitar article taught me that you people don’t believe anything without a receipt.

Here’s an irritation. My alarm goes off at 7. Some mornings there’s an email waiting from one of three specific people that means I need to be vertical immediately. Most mornings there isn’t, and I would very much like to know that before my body decides. So: an alarm that checks my email first. If one of the Three has written, wake me up like it means it. If not, tell me “you’re clear till nine” and let me sleep.

I wrote that idea down in one paragraph. It became a working app before I finished my coffee. It’s a single HTML file. It lives here. You can look at it. You can steal it. My friend’s tool, TransistorKit, did the lifting.

Here’s the part I love, though. My original idea said “text me.” Within ten minutes the prototype informed me, gently, that a single web page cannot send a text message — that part of my brilliant idea required a server, a phone company, and a monthly bill. So it sends a notification instead. My idea was about eight percent fantasy, and I found that out before spending a weekend on it.

That is the entire value proposition of these tools in one sentence: they show you which part of your dream was made up.

So

The robot cost me a gig. I still don’t regret learning it.

The wanting is the job. The taste is the job. The tools will do almost anything except care, and caring is the one thing you’ve been practicing since the first time a piece of music made you feel something you didn’t have a word for.

Go learn where the exits are. Then go make something you actually want to exist.

I’ll be over here, checking whether that network is still real.