The Algorithm Knows I’m About to Quit

The Algorithm Knows I’m About to Quit
I didn’t tell anyone I was thinking about leaving.
Not my manager, not my team, not even myself—not in words that could be quoted back to me. I framed it as tiredness. As a phase. As the natural drag of a job that paid well but asked for more hours than it admitted to. I told myself that everyone in tech felt like this eventually, that burnout was a badge you wore quietly until promotion or resignation made it irrelevant.
The algorithm disagreed.
It started with my recommendations.
On Monday morning, my music app stopped suggesting anything upbeat. No more “Focus Flow” or “Monday Motivation.” Instead, it offered sparse piano playlists titled Letting Go and Soft Endings. I laughed and skipped them. Coincidence. Data drift. A weird week.
By Tuesday, my video feed followed suit. Productivity gurus disappeared. In their place came videos about “knowing when it’s time to walk away” and “careers you can start over at 35.” I turned autoplay off. That feature had always felt a little too intimate anyway.
At work, I was fine. Better than fine. I hit my sprint goals early, answered Slack messages promptly, attended meetings with my camera on and my posture upright. If burnout had symptoms, I wasn’t showing them. My performance metrics were green across the board.
The algorithm saw something else.
It began suggesting articles during my lunch break: The Hidden Cost of Staying Too Long, Burnout Isn’t Failure—It’s Information, Why High Performers Leave Without Warning. I didn’t click them. But I noticed. Noticed how precisely they threaded themselves into the cracks of my scrolling.
I work on recommendation systems. That’s the irony that makes this harder to explain without sounding paranoid.
I know how this works—or how it’s supposed to. Models trained on millions of users. Patterns, not people. Correlation, not intent. The system doesn’t know you; it knows what users like you have done before. It predicts behavior, not feeling.
Except predictions shape behavior. We’ve known that for years.
On Wednesday afternoon, my calendar app suggested I block off Friday “for rest.” Not a meeting conflict. Not a travel buffer. Just rest. The suggestion came with a gentle blue highlight, as if it were doing me a favor.
I declined it.
That night, my job board app—which I hadn’t opened in months—sent a notification: New roles you might like. Not promotions. Not lateral moves. Entirely different industries. Smaller companies. Words like meaning, balance, mission sprinkled through the descriptions like bait.
I sat on the edge of my bed and tried to inventory my feelings, the way therapists and productivity coaches advise. Am I unhappy? I asked myself.
I didn’t have a clean answer.
I wasn’t miserable. I wasn’t crying in the bathroom or staring blankly at my screen. I still enjoyed parts of the work—the puzzle of optimization, the quiet satisfaction of reducing error rates by fractions of a percent. I liked my team. I liked my paycheck.
But when I imagined doing this exact job a year from now, something in my chest tightened. Not pain. Not fear. Just resistance. Like a muscle refusing to engage.
The algorithm had a name for that. I didn’t.
On Thursday, my internal dashboard flagged an anomaly.
We had recently deployed a new burnout-risk model, trained on behavioral signals: login times, message sentiment, task-switching frequency, micro-delays in response patterns. It was meant to be aggregate, anonymized, ethical. A way to help organizations understand systemic strain, not individual weakness.
My user ID appeared in the test output.
Burnout Risk: High
Attrition Probability (90 days): 0.72
I stared at the numbers longer than I should have.
This wasn’t my consumer profile. This was internal. My own work, reflecting me back to myself with unsettling confidence.
I wanted to dismiss it. Models are imperfect. False positives happen. But the feature weights told a quieter story: increased hesitation before starting tasks, longer pauses between keystrokes, subtle shifts in language tone. Nothing dramatic. Nothing a manager would ever notice.
Nothing I had noticed.
The system didn’t say I was unhappy. It didn’t say I was failing. It said I was about to change.
That evening, my phone surfaced a memory: a photo from six years ago, my first week in tech, standing in front of a whiteboard filled with ideas. I looked exhausted and alive in equal measure. The caption I’d written then was embarrassingly earnest. Something about building things that matter.
I had built many things since. Efficient things. Profitable things. Things that worked exactly as designed.
Somewhere along the way, the feeling changed shape.
By Friday morning, I didn’t need the algorithm anymore. It had already done its work—not by persuading me, but by noticing what I wouldn’t name. By holding up a pattern I was still pretending was random.
I took the day off.
Not because an app told me to, but because sitting at my desk felt heavier than usual, and walking outside felt necessary in a way I could no longer ignore. I didn’t apply to any jobs. I didn’t draft a resignation letter.
I just admitted, quietly, that something was ending.
The algorithm didn’t make that choice for me. It never could. But it saw the outline of the decision before I let myself feel its weight.
That’s the part no one talks about.
We think of algorithms as cold, predictive machines. But sometimes they act like mirrors angled just enough to catch what we’re avoiding. Not intention. Not desire. Just trajectory.
And once you see where you’re headed, pretending you don’t know becomes the hardest task of all.