Congratulations on Your Breakup. The Algorithm Will Now Spend Six Months Rubbing It In.

You did the work. You really did.

You unfollowed them on Instagram. You muted their stories. You blocked them on Twitter, then unblocked them to check, then re-blocked them and felt bad about it for a week. You deleted the thread. You archived the photos — or you deleted those too, depending on what kind of night it was. You did everything the modern breakup playbook says you're supposed to do, and then you sat back and waited for the digital world to reflect your fresh start back at you.

The algorithm had other plans.

Three days later, Spotify served you the playlist you built together on a road trip to Asheville. A week after that, Instagram surfaced them in the background of a mutual friend's birthday story — laughing, looking fine, suspiciously fine. Then Google Maps sent a notification reminding you that you'd rated that Thai place on Divisadero four stars and asking if you'd like to write a review. You had gone there for your six-month anniversary. You did not write the review.

Nobody warned you that breaking up with a person and breaking up with the internet's memory of that person are two completely different divorces.

The System Was Never Built to Handle This

Here's the thing that makes this so uniquely maddening: the algorithm isn't doing this to hurt you. It isn't doing anything to you at all. That's almost the whole problem.

Recommendation systems are built on behavioral data — what you clicked, what you lingered on, what you listened to at 11 PM on a Tuesday in November. They don't know that the person behind those clicks was in love. They don't know that the love ended. They have no architecture for processing human loss, because human loss was never a product requirement. Engagement was the product requirement. Retention was the product requirement. Your heartbreak is, from a systems design perspective, not a relevant variable.

So the machine keeps doing what it was optimized to do: surface content that previously generated positive signals from you. And for a lot of people, the person they're trying to forget generated a lot of positive signals.

You watched their Instagram Lives. You streamed their favorite band. You Googled restaurants in their neighborhood. You searched for gifts. You clicked on their tagged photos. You built, without realizing it, a behavioral fingerprint of your relationship — and that fingerprint doesn't disappear when the relationship does. It just keeps getting read.

The Ghost in the Recommendation Engine

There's a specific kind of horror in being haunted by something that doesn't know it's haunting you. A ghost that rattles chains is at least trying. The algorithm isn't trying anything. It's just pattern-matching its way through your grief like a Roomba bumping against the walls of a room it will never understand.

Spotify's Discover Weekly doesn't know you broke up. It knows you used to listen to Phoebe Bridgers together and that you have historically responded well to artists in that sonic neighborhood. So it serves you more. It will keep serving you more until you actively, manually, deliberately train it otherwise — and even then, the collaborative filtering that links your taste profile to theirs through mutual listening patterns is still quietly doing its math somewhere in a server farm in Virginia.

Amazon still thinks you're shopping for a household of two. Netflix still has that rom-com in your Continue Watching queue — the one you paused twenty minutes from the end on the last good night, before everything went sideways. Google Photos is out here generating Memories slideshows with the energy of a well-meaning relative who doesn't know they shouldn't bring up the ex at Thanksgiving.

Each of these systems is, individually, just doing its job. Together, they form a kind of distributed haunting that no single company is responsible for and no single setting can fix.

The Mutual Friends Problem Is Actually a Mutual Data Problem

The cruelest vector isn't even direct. It's the second-order stuff — the ways they bleed into your feed through proximity rather than connection.

You can block someone. You cannot block every person who has ever been photographed standing next to them. You cannot tell Instagram's object recognition to stop noticing their face in the backgrounds of other people's stories. You cannot explain to Facebook's event recommendation engine that the reason you're tagged in photos from that bar is because you used to go there together, and that you would prefer not to receive targeted promotions from that bar for the foreseeable future, thank you.

The social graph doesn't respect the emotional graph. Your network of connections was woven through with theirs, and when you pull your thread out, all those intersecting threads are still there — still generating data, still creating pathways the algorithm can use to walk you right back to a face you were trying to put some distance between yourself and.

Nobody Is Going to Fix This

It would be easy to frame this as a design failure — and in some sense it is. But the more uncomfortable truth is that building grief-awareness into recommendation systems isn't actually on anyone's roadmap, because grief doesn't drive engagement. Nostalgia does. Longing does. The slight emotional ache of being reminded of something you lost? That keeps you scrolling. That keeps you on the app.

There's no version of this where the incentives align in your favor. The algorithm surfacing your ex is, from a pure engagement standpoint, working exactly as intended. You stopped. You looked. You felt something. You kept scrolling. The system logged all of that as a win.

You are not a person going through something difficult. You are a session that has not yet timed out.

The Slow, Manual Labor of Digital Erasure

If you want to actually move on in the digital sense, the work is tedious and mostly invisible. You retrain Spotify by skipping songs aggressively and seeding new artists. You bury old Google searches with new ones. You manually remove tagged photos. You go through shared playlists and either delete them or let them sit there like a box in a closet you keep meaning to deal with.

None of this is romantic. None of it is cathartic. It's data hygiene, and it takes way longer than the breakup conversation did.

And even when you've done all of it — even when you've been diligent and deliberate and borderline obsessive about scrubbing the digital record — there's still some server somewhere holding a behavioral pattern that links you to them. Some model that remembers, even if you've managed to forget.

The relationship ended. The data didn't.

Welcome to the part nobody puts in the breakup advice thread. The algorithm doesn't know you're hurting. It doesn't know anything. It just keeps serving, and serving, and serving — completely indifferent to the fact that every recommendation is a small, stupid knife.

There's no closure notification. There's no setting that says I am no longer this person's person. You just have to outlast the memory of a machine that was never paying attention to begin with.