The People Who Apply Are the Wrong People
Every selective opportunity has a front door, and the front door filters for the skill of walking through front doors before it filters for anything the opportunity is actually about. A structural argument that the application step taxes a trait orthogonal to merit, and that no form redesign fixes it — because optimisers optimise for whatever you ask. The uncomfortable conclusion: for the things you most want to find, the open door is exactly the wrong instrument.
Every selective opportunity has a front door, and the front door does something nobody designing the opportunity intends: it filters for the skill of walking through front doors, before it filters for anything the opportunity is actually about.
Call it the application paradox. The act of applying is not neutral. It is a test — of a specific, narrow, real ability that has almost nothing to do with the thing being selected for. To apply, a person has to find the opportunity, decode what the selector wants, and present themselves as a credible version of that. Those are genuine skills. They are also a completely different set of skills from being a brilliant scientist, an original artist, a careful thinker, or whatever the prize claims to reward. So every selective process, from the most prestigious fellowship to the most idealistic grant, quietly runs two filters at once: the one it advertises, and the one nobody mentions, which is are you good at applying for things. The second filter runs first. It runs on everyone. And it has no off switch.
The optimiser
The people who are good at applying are optimisers. This isn’t an insult — it’s a precise description of a capability. An optimiser reads a landscape of opportunities, identifies the ones with their name on them, works out the selection criteria, and shapes their self-presentation to fit. They have a CV that anticipates what reviewers want. They know that “passionate about impact” beats “I needed a job.” They’ve done this before and they’ll do it again, because they understand, correctly, that articulated opportunities are won by the people who show up articulate. The optimiser sees a prize and thinks: I can win that. They are already in motion while everyone else is still unaware the thing exists.
The problem is that for a large class of opportunities, the optimiser is exactly who you don’t want. Anything genuinely trying to find originality, independence, or the un-obvious is trying to find people who are defined by not optimising for external legibility. The deeply original researcher is following a question nobody’s rewarding yet. The real artist isn’t tracking which grants are open. The genuinely independent thinker is, almost by definition, not oriented toward winning the approval of selection committees — that orientation is the opposite of the trait. So the more an opportunity is designed to reward non-conformity, the more violently the application paradox bites: you’ve built a front door that admits the conformist-at-finding-doors and turns away the very people whose disinterest in doors is the quality you were hunting.
You can’t fix it by redesigning the form
This is the part that traps everyone who notices the problem and thinks they’re clever enough to engineer around it. Make the application weirder — ask for a strange essay instead of a CV, reward unconventional answers, lower the polish requirement. It doesn’t work, because optimisers optimise for whatever you ask. Tell people you want unconventional, and the optimisers will perform unconventional more convincingly than the genuinely unconventional, who aren’t performing at all and often perform it badly. Any signal you request, the people who are skilled at sending signals will send better. The only people who can’t game your filter are the people your filter is meant to catch. That’s not a flaw in your design. It is what a filter is. A filter rewards the people who are good at being filtered for, and “good at being filtered for” is a trait, and it is rarely the trait you want.
Selection bias wearing a halo
The mechanism underneath is selection bias wearing a halo. We’re used to selection bias as a statistical nuisance — the survey only reaches people who answer surveys. The application paradox is the same effect, but the selected trait is self-promotion, and we don’t notice it because we mistake the optimiser’s polish for the quality we were looking for. The confident applicant seems more promising. The one who found the opportunity and nailed the brief seems more capable. They are more capable — at applying. We read that fluency as a proxy for the underlying excellence, and sometimes it is, and sometimes it’s a person who is mediocre at the actual work and exceptional at being chosen. The two are uncorrelated often enough that any process trusting the application as a signal is partly, sometimes mostly, selecting for the wrong thing.
Where this shows up
This shows up everywhere once you see it. Scholarships go disproportionately to students coached in the genre of scholarship essays, not the students who’d benefit most. Accelerators select for founders who pitch well, and pitching is a skill that can fully decouple from building. Job postings reward people fluent in the dialect of job applications over people who’d do the job better but write a flat cover letter. Awards accrue to the people who know awards exist and how they’re won — which is why the same names recur and why “never even occurred to them to apply” quietly removes whole populations of the deserving. Open calls for “bold, unconventional” anything fill up with people performing boldness, while the actually bold are off doing the bold thing instead of writing about how bold they are. In every case the structure is identical: the front door taxes a skill orthogonal to the merit, collects that tax from everyone, and lets it determine who gets considered at all.
What the selector can actually do
So what do you do, if you’re the one trying to find people? The uncomfortable answer is that for anything where the target trait is anti-correlated with self-promotion, the application is the wrong instrument and a better form won’t save it. The people worth finding frequently won’t come to you, because coming-to-you is the behaviour you’re trying to select against. You have to go and find them — through people who know them, through their work rather than their pitch, through the slow unscalable method of noticing someone before they’ve packaged themselves. That doesn’t scale into a clean process with a deadline and a number of slots, which is exactly why almost nobody does it and almost everybody runs an application instead. The application is easy to administer, defensible, and fair-looking. It just happens to filter, first and hardest, for the one trait you’d have wanted to ignore.
The open door is not neutral ground
The deeper point is that openness and selectivity sit in a tension we rarely admit. We tell ourselves that an open call is the democratic, meritocratic option — anyone can apply. But “anyone can apply” guarantees that the people who apply are the people inclined to apply, and inclination-to-apply is a trait with its own heavy biases of temperament, background, and self-regard. The open door is not neutral ground. It’s a sieve shaped like an invitation. And the more glittering the prize behind it, the more it pulls the field toward the people who chase glittering prizes, and away from the ones who were never looking.
Which leaves the quiet, deflating conclusion that the most important selection is the kind that doesn’t take applications at all. The best things often find their people sideways — a recommendation, an accident, someone paying attention who reaches out first. The instinct to systematise that, to build a fair open process so the right people can find their way in, is decent and well-meant and quietly self-defeating, because the process you build to be fair is the same process that taxes the people you most wanted at the door. The people who apply are the people good at applying. Whether that’s who you wanted is a question worth asking before you open the door — and sometimes the honest answer is to not build the door at all.
What would change my view
Three things, in declining order of weight.
Evidence that the optimiser-skill and the target trait are more correlated than the piece assumes. The piece treats the two as orthogonal or even anti-correlated for high-merit selection. If there were systematic empirical work showing that scientific originality, artistic depth, or founder durability are positively correlated with the disposition to apply for things, the central claim would weaken. The honest reading of the limited available evidence is that the correlation is weak, but I have not seen a meta-analysis that settles it.
A counter-example: a redesigned application that demonstrably caught the people standard applications miss. The piece argues that no form redesign can fix the underlying filter. A clear case — not the selector’s own self-report but a follow-up study finding the redesign caught a measurably different population — would falsify the “you can’t redesign your way out” argument. The MacArthur Fellowship’s no-application model and the Mathematical Olympiad’s domain-test model are partial existence proofs that something other than the standard application can work; whether they actually select for what they claim is a separate empirical question.
A working alternative at scale. The piece’s suggested alternative — finding people sideways through networks — is honest about its unscalability. If someone has built a selection process that is both scalable and applies-the-target-trait directly rather than applying-fluency, that would be the most important existence proof of all, and the piece’s pessimism about the open door would need substantial revision.
The publication’s working position is that the application paradox is a real and substantial source of selection bias in a wide class of opportunities, that it gets worse the more idealistic the opportunity, and that it cannot be designed away. The publication does not consider this position decisively proven.