Because they had an absurd take home assignment that was basically recreating their entire mobile app.
That's the difference between startup culture and big company culture. Big companies get away with accurate estimates, story points, burndown charts. Work is rounded down to what's accomplishable in a week. Startups just have an infinite blob of things to do and need those who can deal with that.
Which is a BS rationale, of course. If you (at lease claim to) respect the candidate's time, then you don't give out slippery-slope assignments with fake / unknowable goalposts, period. There is no benefit to this tactic that can possibly outweigh the basic insult and the egregious waste of their time.
Startups just have an infinite blob of things to do and need those who can deal with that.
Which, face it now -- you will not be able to measure the capacity for in the interview process. And the fantasy that you are somehow able to is no justification for willfully and knowingly burning through their time like this.
A standard question might be "Build a chess game in 4 hours". Now that's clearly impossible. So you shouldn't take offense at not being able to do it.
But part of the job is being able to break down a complex task and propose a solution. There will be people who jump into architecture and stuff on how to build it in 4 weeks. No, you've got 4 hours, think about it.
So you have to scope down the question, and see what you can tackle in these few hours.
A good applicant will ask what the goal is. Is it entertainment? How about a checkers game instead? Is it for practice? Maybe it can be a "mate in 1" game with some hardcoded moves.
How would you even render the pieces? Would you reverse engineer lichess? Do you build a plaintext prototype? Do you TDD possible moves? How would you do user input and selection?
You get points for coming up with possible solutions. There's going to be some whiz who probably made some battle chess prototype for Ludum Dare. That's fine. They're smart, they get points. But most of the points will come from the learning process to the results.
It is designed to filter out the people who walk out when they see the problem.
That doesn't sound impossible at all. I am sure I could do it in a weekend, and pretty sure I could do it in a day, if strongly motivated and well rested. So, maybe a more skilled person than me could get it done in 4 hours, although it would definitely be impressive and I would expect anyone who can do it, already has done bits of it before (eg they already know their graphics library well, and have implemented alpha-beta search or something for the AI part before -- those are the only two hairy parts, other than that chess is pretty simple)
I think my point is that the question is mostly selecting for experience-- someone who has already implemented parts of it before will be able to do it so much faster. If you don't have a graphics library you already are comfortable with, for example, forget it, that's gonna eat up days on its own.
If they politely (or perhaps not-so-politely) decline your request - that doesn't mean (as you seem to presume) they have no imagination, cannot think outside the box, or are perennial quitters.
More likely it's because when you ask questions like that, you're coming off as incredibly cocky and pretentious. And because they're serious professionals with lots of pertinent stuff to talk about - and really don't see a need to join you in what amounts to an unsolicited 4-hour intellectual masturbation session.