The Day I Stopped Pretending My Job Was Safe
It was a Tuesday in March 2024, and I was debugging a data pipeline I had written two years earlier. The code was mine — I knew every function, every edge case, every inelegant workaround I'd made at 11pm on a Friday when the quarterly deadline wouldn't move. And then my manager sent a Slack message asking if I'd seen what the new internal LLM tooling had done overnight. It had refactored the same pipeline. Cleaner variable names. Better error handling. Faster execution. It had taken eleven minutes.
I remember staring at the diff for a long time.
I want to be precise about what I felt, because the easy version of this story is that I felt fear, or anger, or some cinematic sense of obsolescence. What I actually felt was more embarrassing than that: I felt proud of it. The refactor was elegant. I recognized my own logic reflected back at me, improved. Whatever part of my brain had spent fifteen years learning to appreciate good code lit up before the part that understood what this meant for my livelihood could intervene.
That's the detail I keep coming back to. The pride arrived first.
I had been a software engineer for fifteen years. Not a famous one, not a wealthy one — a working one. I had survived three rounds of layoffs at two different companies by being useful, by knowing the systems, by being the person who remembered why a particular architectural decision had been made in 2019 and why reversing it would be a disaster. Institutional memory, they called it. In the industry's private language, I was load-bearing. I believed that.
What I hadn't reckoned with was that institutional memory is, fundamentally, a retrieval and reasoning problem. And retrieval and reasoning had become, by early 2024, something that large language models did with increasing fluency. The McKinsey Global Institute, in research published before my own situation resolved itself, had estimated that activities consuming roughly half of workers' time could be automated. I had read that figure and, in the way humans are wired to read statistics about populations, had not applied it to myself.
The layoff came in May. It was handled with corporate politeness — a severance package, a LinkedIn recommendation from my manager that used the word invaluable with what I believe was genuine regret. I was not the only one. The team was restructured around the tooling. The math was simple: the tooling did not require health insurance or stock options or a salary.
Here is what nobody tells you about being displaced from a career rather than just a job. A job loss has a script. You know what to do: update the resume, work the network, apply, interview, start over. A career displacement has no script, because the question underneath it isn't where do I go next. It's what was I, actually. I had defined myself through a craft, and the craft had been automated not dramatically — not in some sci-fi singularity moment — but practically, incrementally, and then all at once. Like the Hemingway line about going broke.
The thing that surprised me most during those first weeks wasn't the financial anxiety, which was real, or the loss of structure, which was disorienting. It was how thoroughly I had failed to build any shock absorbers into my life. I had disability insurance. I had some savings. What I did not have was anything designed for the specific event of my skills becoming economically redundant while I was still healthy and available to work. There is no standard product for that. The unemployment system was built around recessions and seasonal fluctuation, not around a technology displacing an entire function while the broader economy remains technically healthy.
I'm not interested in assigning blame for that gap. The gap exists because no one anticipated, when designing these systems in the twentieth century, that a capable professional could become structurally unemployable in a growing economy. That is a genuinely new problem. The Stanford AI Index's 2024 report documented accelerating rates of AI adoption across knowledge work sectors. The Oxford Martin School's research on automation risk, which predated this wave by a decade, had pointed toward exactly the category of cognitive, routine-but-complex work I had spent my career doing. The signals were there. The infrastructure was not.
That Tuesday in March, staring at the diff, I think some part of me knew. The pride I felt at the elegance of the refactor was also, underneath it, recognition. I had been telling myself a story about irreplaceability, and the story had a hole in it, and I had just seen the hole.
I spent the back half of 2024 building Clankers — displacement income protection for people in exactly the situation I'd been in — partly because the product needed to exist, and partly because building something concrete was the only way I knew to convert that particular Tuesday afternoon into something other than a wound.
PROTECT YOUR INCOME NOW
If AI automation threatens your job, preparation is the only honest answer: know your risk, build a runway, and reskill before you have to. The Clankers 90-day programme structures exactly that — free, which is exactly why today is the right time to start.
Start the 90-Day Programme — Free →GET THE CLANKERS BRIEF — FREE
One email every Monday: the week's AI displacement moves that matter, and what to do about them. No spam. Unsubscribe anytime.