Rejection Is a Door, Not a Score

The largest study ever done on hiring says it is not you

In May 2026, researchers at Stanford, Chapman, and Northeastern published "Algorithmic Monocultures in Hiring." They looked at more than four million job applications from three million people across 156 employers, all screened by one vendor's algorithms. They found something that should change how you read your own rejection pile. A small number of models, 42 of them, were shared across all those employers. When you take the assessment, your score is stored and reused for up to 330 days, and you usually cannot retake it inside that window. So if a dozen companies use the same vendor, you are not getting a dozen evaluations. You are getting one score, played back a dozen times.

The researchers have a name for what happens next. They call it an algorithmic blackball. One model decides you are not a fit, and then that single judgment quietly follows you across every company wired to it, for the better part of a year. The study found people who applied to ten roles and got recommended for rejection from all ten, at a rate far higher than chance. It also found the outcomes fell harder on Black and Asian applicants, by margins large enough to trigger federal discrimination scrutiny.

Read that again, because it matters for how you feel on application four thousand and one. You were probably not rejected by fifty companies. You were rejected by one model, fifty times. That is not a measurement of your capability. It is a measurement of a system that mistook a single narrow signal for the whole of a person.

What hiring stopped measuring

Here is what that system stopped measuring along the way. It used to be that your work spoke, and people who had worked with you spoke for you. Then it became a keyword match against a resume. Then it became a score from a timed test you take alone. And now it is increasingly a video you record of yourself, performing enthusiasm into a webcam for a panel that may never watch it live. Each step moved further from the work and closer to the performance. Each step rewarded the person who is good at being evaluated over the person who is good at the job.

If you are the heads-down type, the one who would rather ship the thing than narrate it, this is a hard era to be measured in. I am that type. I built things and let them speak. I said "we" when I meant the team, not "me," because that is how the good work actually gets made. For a long stretch none of that registered. What registered was whether I could pass a test on someone else's clock, and once a test decided I was not qualified, that was the door.

I built my own door

So I built my own door.

I will not oversell this, because the work should carry it. In the time since, we have built a customer-hosted cloud security agent that runs inside a company's own account, reasons about who can reach what, and can safely apply fixes behind hard, fail-closed gates. We have shipped a tool that is live in production for a paying client. We run an operating layer that supervises other AI systems doing real work. None of it would have shown up on a timed assessment, because the thing a real system requires, the judgment to design it and the stubbornness to steward it, is exactly the thing those tests cannot see.

That is the whole point, and it is the part I want you to take with you. The test was not wrong about your worth. It was just never able to measure it. A score is a snapshot of one narrow moment against one narrow rubric. It is not a verdict on what you can build, learn, or become.

Rejection is a door, not a score

Rejection is a door, not a score. Sometimes it is a door someone else closed for reasons that have nothing to do with you. And sometimes a closed door is the thing that finally makes you build your own, on your own terms, where no shared model gets a vote.

If you are in the silence right now, here is the most useful thing I know. Make something. Small is fine. Put one real piece of your own work into the world, where it can be seen by a human instead of scored by a monoculture. It is the one move the algorithm cannot blackball, and it is how the next door tends to open.

The silence was never the measure of you. It was the measure of a system that forgot how to look.