I’m currently wedged between the Great and the Swell divisions of a mid-century pipe organ, my hands covered in a fine grey silt that smells like 1922. It is a specific kind of dust, a mixture of shaved lead, aged cedar, and the pulverized remains of a thousand moths that died during the Coolidge administration. My job is to find the beat-that tiny, pulsing interference pattern that occurs when two pipes are almost, but not quite, in tune. If the middle C is vibrating at 262 hertz and the tenor C is slightly off, the air itself begins to throb. It’s a physical sensation, a rhythmic pressure against my eardrums that tells me exactly how much tension to apply to the tuning slide.

The Infinite Echo of ‘Considering’

My phone, resting precariously on a wooden windchest, vibrates. I wipe a smudge of graphite onto my trousers-I matched 32 pairs of socks this morning, a feat of domestic organization that has left me feeling strangely invincible-and check the screen. It’s an advertisement for the TOTO Drake II Two-Piece Linear Gravity Flush Toilet. Again. This is the 112th time I have seen this specific toilet in the last 12 days. The algorithm is certain, with the cold, hard conviction of a mathematical theorem, that my life is currently a void that can only be filled by 1.28 gallons of high-efficiency flushing power.

The problem, which the internet seems fundamentally incapable of grasping, is that the toilet is already installed. I bought it 22 days ago. I spent 42 minutes bolting it to the floor of my guest bathroom. I have moved on. My physiological relationship with plumbing has reached a state of equilibrium. Yet, to the vast, shimmering consciousness of the ad-tech stack, I am forever trapped in that liminal space of ‘considering a purchase.’ I am a ghost haunted by my own consumerist past, forced to watch the echoes of decisions I have already finalized.

We are told that we live in the age of big data, an era where silicon valley’s predictive models know us better than we know ourselves. They claim to anticipate our desires before they even reach our conscious minds. But the toilet ad reveals the cracks in the facade. It exposes the fact that most ‘artificial intelligence’ is actually just a very fast, very expensive rear-view mirror. It tracks where we have been with terrifying precision, but it has no conceptual framework for ‘completion.’ It understands the search, the click, and the cart, but it cannot fathom the concept of ‘enough.’

The Historian Algorithm

In my line of work, if I kept tuning a pipe after it reached its pitch, I would eventually collapse the metal. There is a definitive end point to the task. But the digital world operates on a loop of infinite persistence. When I searched for ‘best low-flow toilets’ 32 days ago, I dropped a digital breadcrumb. That breadcrumb was picked up by a pixel-a tiny, invisible 1×1 image hidden in the code of a home improvement blog. That pixel reported back to a server, which tagged my IP address with a specific identifier. Within 2 milliseconds, I was entered into an automated auction. My attention was sold for a fraction of a cent to a plumbing supply conglomerate.

This is retargeting in its crudest form. It’s a system designed to overcome ‘cart abandonment,’ based on the statistical reality that people often need 12 nudges before they pull the trigger. But the system is binary. It knows I looked. It doesn’t always know I bought-especially if I bought it from a different retailer, or in a physical store, or if I used a different device to finalize the transaction. Because the ‘purchase’ pixel didn’t fire on the specific domain that started the chase, the chase continues. I am being hunted for a prize I have already surrendered.

There is a profound irony in the fact that the more data these companies collect, the more ‘stupid’ their interactions feel. If a person followed you around a hardware store for 22 days, whispering ‘Do you want this toilet?’ every time you looked at a watch or read a book, you would call the police. In the digital realm, we call it ‘personalized marketing.’ It’s a failure of context. The machine sees the ‘what’ but is utterly blind to the ‘why’ and the ‘when.’ It lacks the human capacity to understand that a purchase is often an exit, not an entrance.

(LMK.today demonstrates functional relevance).

The Flaw of Infinite Persistence

I find myself thinking about the 82 pipes in the trumpet stop I tuned earlier. Each one has a specific resonance. If I treated them all the same, the organ would sound like a screaming mechanical nightmare. I have to listen to each one’s unique temperament. The internet, by contrast, treats us as aggregates of our recent metadata. It assumes that if I liked a thing once, I want to see that thing for the rest of eternity. It’s a flat, one-dimensional version of identity that ignores the fluid, evolving nature of human existence. I am not the same man who needed a toilet 12 days ago. I am now a man who needs a very specific type of felt for an organ pallet, yet not a single ad has offered me that.

Instead of being helpful, the technology becomes a nuisance, a digital fly buzzing against the windowpane of my attention. It creates a friction that is the exact opposite of what good tech should do. When I use a tool like LMK.today, there is a sense of functional relevance-it’s about providing value in the moment, responding to the actual state of the world rather than screaming about a state that ended weeks ago. The disparity between a tool that waits for your signal and an ad-bot that screams into the void of your past is the difference between a finely tuned instrument and a broken record.

The 72-Hour Gap and Data Silos

There’s a technical reason for this lag, often referred to as the ‘attribution window.’ Advertisers set these windows-sometimes for 32 days, sometimes for 62-during which they want to keep their product top-of-mind. They are terrified of the ‘dark funnel,’ the mysterious period where a consumer disappears before reappearing to buy. To mitigate this fear, they carpet-bomb the user. They would rather annoy 92% of their customers who have already bought the item than miss the 2% who are still undecided. It’s a scorched-earth policy for the attention economy.

Moreover, the siloed nature of data means the left hand rarely knows what the right hand is doing. The ‘Demand Side Platform’ (DSP) that serves me the ad on a news site doesn’t have a real-time link to the ‘Point of Sale’ (POS) system of the brick-and-mortar store where I actually swiped my card. Even within the same company, the marketing department and the sales department might be using databases that only sync every 72 hours. In that 72-hour gap, a thousand redundant ads can be born and died.

Aiden P.K. doesn’t like redundancy. When I match my socks, I don’t leave one stray black sock in the drawer just in case I find its mate in 12 months. I find the mate now, or I discard the set. There is a cleanliness to it. The internet’s junk drawer, however, is overflowing with ‘reminders’ for things we’ve already dealt with. It’s a cluttered, noisy environment that makes the simple act of browsing feel like walking through a minefield of your own history.

The Decibel Hiss

I wonder if this is the inevitable end-point of a system built on surveillance. When you collect everything, you lose the ability to prioritize anything. The ‘signal-to-noise’ ratio gets skewed. In the organ, if a wind leak is loud enough, you can’t hear the pitch of the pipe. You just hear the hiss. Ad-tech has become a 92-decibel hiss that drowns out the actual needs of the consumer. It’s a system that has forgotten how to listen because it’s too busy recording.

Sometimes I contemplate clicking the ads just to see what happens. If I click the toilet ad now, does the algorithm think, ‘Aha! He’s still interested!’ and double down? Or does it finally see the ‘out of stock’ or ‘already purchased’ flag on the destination page? Experience suggests it’s the former. The click is interpreted as ‘engagement,’ which is the holy grail of marketing metrics. Engagement is often just a polite word for ‘accidentally touched the screen while trying to close a pop-up.’

There is a certain comfort, I suppose, in knowing that the machines are still this fallible. We fear a future where AI is an all-knowing god that manipulates our every move, but right now, it’s just a confused toddler pointing at a toilet and shouting. It reminds me that for all the billions of dollars poured into ‘predictive analytics,’ there is still no substitute for the human understanding of a closed loop. A purchase is a period at the end of a sentence. The internet, in its infinite, frantic energy, wants everything to be an ellipsis…

Listening to the Silence

I have 12 more pipes to tune before the sun sets and the church gets too cold for the metal to hold its shape. I’ll put my phone back in my pocket, let the toilet ads play out their lonely, automated drama in the dark, and focus on the beats. There is a clarity here, in the dust and the lead, that the digital world can’t seem to replicate. Out here, when a note is in tune, you stop turning the wrench. You listen to the silence that follows the resolution. It’s a lesson the internet could stand to learn: there is a great deal of power in knowing when to shut up.

I’ll go home tonight, open my laptop to check the weather, and I’m 92% sure I’ll see that TOTO Drake II again. I’ll look at my perfectly matched socks, feel a small surge of internal order, and realize that the algorithm’s stupidity is actually a gift. It’s a reminder that I am still more complex than my search history. I am a person who buys toilets, yes, but I am also a person who tunes organs and matches socks and knows the smell of 1922 dust. The internet only sees the toilet. And as long as it stays that way, I think I’m doing just fine.