My Friend's Guruji Said No. I Shipped Anyway.

I built a musical instrument I can’t play.

A MacBook that thinks it’s a harmonium. You rock the lid to pump the air. You press A through J, and out come the notes Sa to Ni.

Then a master musician heard it and said no.

He was right. I shipped it anyway.

The idea in 20 seconds

• A harmonium is a hand-pumped Indian reed organ. One hand squeezes a bellows for air. The other plays the keys.
• I turned that into a free, open-source Mac app. The laptop’s lid-angle sensor becomes the bellows — you literally rock the screen to pump air.
• I’m not a musician. I love music. I know zero theory.
• So I borrowed an ear I trusted, extracted the notes myself, and handed the numbers to an AI to build the sound.
• My extraction was flawed. The AI built the flaw perfectly. The expert caught it. I couldn’t.

Where a laptop learns to breathe

The whole thing started as wonder.

Someone had already made a project called Hingemonium — same trick, using a laptop’s hinge as a bellows. I saw it and couldn’t let it go.

There’s a sensor in a MacBook lid that knows the exact angle of the screen. Sam Gold did the hard work of reading that sensor. I stood on top of it.

Rock the lid forward, rock it back, and the app hears that motion as breath. Press A, B, C up to J, and you get Sa, Re, Ga up to Ni.

I built almost all of it with an AI — Claude — sitting beside me like a very patient engineer. I described what I wanted. It wrote the code. I got to make a real thing I had no business making.

That part still amazes me.

But what does a harmonium actually sound like?

Here’s where I hit my wall.

A harmonium isn’t just keys. It’s tuning. Each note sits at a specific frequency, and the relationship between those frequencies is the thing your ear calls “in tune.”

I don’t have that knowledge. I have enthusiasm.

So I borrowed an ear I trusted. My friend studies under a teacher — Pandit Natvarlalji. Not my guru. My friend’s. But a real master, the kind of ear that’s spent a lifetime listening.

He gave me the tunes.

Then I did the part that was mine to do, and mine to get wrong. I sat and extracted the notes myself — the frequencies in hertz, the scale, the numbers behind the sound. I wrote them down. I handed them to the AI.

The AI built exactly what I gave it.

What the trained ear caught that mine couldn’t

When the guruji finally heard the finished instrument, he didn’t approve it.

His ear caught something off. My extraction wasn’t clean. Somewhere between his tunes and my numbers, I’d let the tuning drift.

And here’s the humbling part. He was completely right. His standards are far higher than mine, and that’s exactly as it should be. A master should hear what a beginner can’t. That’s what makes him a master.

But I listened to my own instrument, and I was happy. My naive ears couldn’t feel the gap. To me, it sounded like a harmonium.

So I had a choice.

Do you ship when the expert says no?

This is the honest middle of the whole story.

The one trained ear said it’s wrong. The thousand untrained ears — including mine — would never notice.

Who are you building for?

I chose the thousand. I shipped it.

I won’t dress that up as courage. It’s a real trade-off, not a clean win. A free toy that lets people feel the joy of pumping a laptop into song — for that, “close enough” is a gift. A serious instrument for a serious student — for that, the guruji is right and I fell short.

Both things are true at once. That’s the part I had to sit with.

The AI built my mistake perfectly

Here’s the lesson I keep coming back to.

The AI didn’t get this wrong. I did.

I fed it a flawed extraction, and it faithfully built a flawed instrument. It didn’t round off my error or quietly fix my blind spot. It couldn’t. It gave me back exactly what I put in, at full resolution.

That’s the thing nobody warns you about with these tools.

They’ll build whatever you can describe. Including your mistakes. Including the taste you don’t have yet.

The AI supplied the hands. It could not supply the ear.

The honest bit

I still open the app and rock the lid and grin like a kid.

And somewhere, a master could still hear everything wrong with it.

Both of those are true. I decided I could live in that gap — a happy toy that a trained ear would fail.

I’m still not sure I chose right. And I think that’s the most honest thing I can tell you.

Sources and further reading

• Mac Harmonium — the project
• Hingemonium — the inspiration
• Sam Gold’s Lid Angle Sensor work this builds on