There is a folder in the repository that runs this website, and nothing on the site links to it. It holds fourteen PDF files. They are all resumes, they are all mine, and they all describe exactly the same thirty years.
They are not fourteen different careers. They are fourteen answers to the question "which parts of this will the reader tolerate", which is not the same question as "what did you do". Each is a lossy projection of the same object, flattened along whichever axis a reader was thought to care about.
I keep them because the instrument is broken and you still have to hand one over. But I have stopped pretending it is a measurement.
I studied physics and never worked a day in a laboratory. What the training left me with is a question I still ask before trusting any measurement: what is this instrument actually capable of detecting?
Every instrument has a range, a resolution, and a set of things it is structurally blind to. A thermometer in a fire tells you nothing about smoke. That is not a fault in the thermometer. It is a fault in whoever brought a thermometer.
So here is the honest question about a resume. What is it built to detect, and what passes through without registering?
It detects titles, employers, dates and named technologies, and it is quite good at those. It is competent for the narrow band of a career that happens to be nameable.
Everything else falls straight through.
The clearest example is the work that consists of something not happening.
Some of the best work I have done in thirty years was talking an organisation out of a decision. A migration that would have been a catastrophe and instead was a meeting. A budget line that quietly did not get spent. An outage that never occurred, on a Tuesday nobody remembers, because six weeks earlier somebody insisted on the unglamorous design.
There is no field on a resume for the disaster that did not happen. There cannot be, and the reason is fair enough. A counterfactual has no evidence attached. I cannot prove the outage I prevented, because the proof of my having prevented it is that there is nothing to point at. Any format demanding I list my prevented catastrophes would reward whoever invented the most impressive ones.
This is a different complaint from work that gets reattributed to somebody else, which at least leaves a trace. What I am describing never had a place on the form to begin with.
This is not a conspiracy against experienced people. It is a genuine limit. But a genuine limit is still a limit, and the consequence is unavoidable: the longer you work, the more of your value sits in judgement, and judgement is precisely the part the instrument cannot see. You get better at the job in exactly the dimension the measurement is blind to.
The format itself encodes a second argument, one almost nobody notices they are accepting.
Reverse chronological order is not neutral. It asserts that recency is relevance: the most recent thing you did is the most important thing about you. For a career of five years that is roughly true. For a career of thirty it is often exactly backwards.
The format has no way to say "I have done this twice, fifteen years apart, and the second time I understood why it worked." That sentence is the most valuable thing I could tell a hiring manager about myself. There is no line on the page where it goes. It gets compressed into two entries, in different decades, in a list sorted by date, where the point of the pairing is invisible.
Different industries. Different decades. Same shape of problem. The instrument records both as unrelated rows.
I will put real numbers next to this, though not all of them deserve equal weight.
The strongest evidence is a field experiment, not an opinion survey. David Neumark, Ian Burn and Patrick Button submitted over 40,000 job applications and found that, pooled across the occupations they studied, callback rates ran about 35 per cent lower for applicants aged 64 to 66 than for applicants aged 29 to 31 (NBER working paper 21669, later published in the Journal of Political Economy in 2019). In administrative roles alone the fall was steeper: 14.4 per cent of young applicants heard back, 10.3 per cent in the middle group, 7.6 per cent of the oldest.
The authors are careful to say the evidence is robust for older women and weaker for men, and that they studied retail, administrative, janitorial and security roles, so this does not directly establish the same effect in senior professional hiring. It measures behaviour rather than belief, which is why I trust it most.
For mechanism, the Harvard Business School and Accenture study Hidden Workers found 48 per cent of employers screening out candidates whose work history showed a gap of more than six months, on that basis alone (Hidden Workers: Untapped Talent, 2021). Its fieldwork ran in 2020, so treat it as a snapshot. A gap of six months. A sabbatical, an illness, a parent who needed care, a business that did not work out. Thirty years of depth, filtered on a six-month hole.
And then there is the number everybody already knows.
You have heard that recruiters spend six seconds on a resume. It is the most repeated fact in the entire subject.
I went looking for the study, and there is not one. The number comes from a 2012 press release by a job board, describing an eye-tracking exercise on "dozens of professional recruiters" across ten weeks. No stated sample size. No named research partner. No peer review, and no published protocol.
A follow-up in 2018 revised it to 7.4 seconds, on a sample it also declined to state. The report both releases pointed at is no longer available.
Fourteen years, and I cannot find anything better. The best-known measurement of how the instrument gets read is itself unmeasured. That is the state of the evidence.
The fair counterweight is that the volume is real.
Across more than 6,000 companies on one hiring platform, applications per opening roughly doubled between 2022 and 2025, from about 115 to 244, while recruiting teams were cut in half (Greenhouse recruiting benchmarks, figures as published March 2026). That is platform data from one slice of the market, and I would not stretch it further. But nobody reads 244 careers properly. Nobody could.
Under that load, what a resume actually is becomes clear. It is a triage device, designed to discard quickly and cheaply. By that standard it works, because a system built for fast rejection only has to be right about the rejections it is confident in.
The error was never in building a triage tool. It was in mistaking triage for measurement, and then believing the result. We built a sieve, and somewhere along the way we started reading it as a scale.
So I still keep the fourteen files, I still update them, and last month I spent an afternoon getting one onto a single page.
Being right about the instrument does not excuse you from using it. That is the least satisfying sentence here and the truest. Understanding why a measurement is invalid buys no relief from the people who will measure you with it anyway. What it buys you is smaller: you stop taking the reading personally, and stop organising your life around improving a number that was never measuring you.
What it also does, if you sit with it long enough, is change the question. Not "how do I score better on this thing," which is a lifetime of optimising for a sieve. The question is what you are building that never has to pass through it.
Thirty years of my work exists. Almost none of it is legible on that page.
The next time you update yours, and find yourself trying to fit the strongest decade of your life into eleven words and a bullet point, notice what you are actually doing. Are you describing your work, or writing an entry for a machine that will only ever read you long enough to decide you are not worth reading?