Any filter good enough to actually work looks like something other than a filter. This is not a flaw in the design. It is the design.
The people most worth screening — the ones who've already built things, who've already been in the room — have seen every filter. They know which answer advances them and which doesn't. If your filter looks like a filter, they will give you what the rubric wants and you will learn nothing.
So you build an experience instead. And you watch what people do in it.
the question problem
A question asks: what is your answer? This works for factual recall. It fails completely when you're trying to measure character.
Ask someone how they handle ambiguity. They'll say well. Ask them to rate their own agency on a scale of one to ten. Every person in every room has always written down seven.
The answers aren't useless. They're measuring someone's theory of themselves, which correlates loosely at best with what they actually do.
A game creates a situation where the thing you want to measure becomes visible — not through self-report but through behavior. The player's moves are the data.
the skip button
The hardest thing to fake is voluntary behavior under genuine optionality.
You can memorize answers. You can perform confidence. You can fill in whatever the rubric seems to want. But you cannot convincingly skip something you actually need.
If someone skips onboarding — not fails it, skips it — the information that conveys is legible in a way no answer could be. The specific flavor of impatience. The willingness to forfeit the scaffolding. The self-assurance that is either genuine or immediately testable.
Design for the skip, and you find out who actually needed to be there.
three outcomes, not two
Most filters have two outcomes: pass and fail. This is too blunt.
What you actually want to know isn't whether someone is ready. It's when. Some people aren't ready and will never be. Some aren't ready and need two months before they come back. These look identical on a binary filter and are completely different signals for what to do next.
The filter that works has an ejection path: not rejection but redirection. Come back. The door is the same. The game is different when you return.
That requires believing you have something worth coming back to. That belief is itself a high-agency move on the builder's side.
why this is a computer future problem
Intelligence is no longer scarce. Everyone has access to it now. The question isn't who has the processing power — it's who has the agency to use what's available. To move fast enough. To care enough. To be the kind of person who pays for the experience rather than waits for the free version.
That is not something you can ask on a form.