Start with simple questions and tasks and let the user build confidence. Once they answer / complete a few tasks, their ego will push them to take on the next few tasks!
For example:
"2 + 1 = ?": you are expected to answer with a number to complete the equation. The question mark is a good cue -- you know to answer with a number and not, for example, "addition".
"5 x 7 = ?": the same formula as before. You now safely assume that math -- especially math with a "?" after an equals sign -- should be answered with a number.
After that, it degenerates into vague riddles where you're left to guess at the intentions of the author, quite literally. The first place this becomes immediately frustrating is: "2,3,5,7,11,13,17 What are they called?"
Early versions of this game omitted the second line, leading people to guess "19" as a continuation of the prime sequence.
Other popular answers are "prime" or "primes", but the ONLY acceptable answer is "prime numbers." There is no clue leading to the acceptable answer and the only possible way to get it is trial and error or sheer luck.
Primed by that, we face the next problem: "3.14159265..."
Given past history, we can guess that the author either wants us to continue the sequence (unlikely, given the lack of _ or ? cues) or name the sequence. This would be much clearer with the addition of the words "What is" at the beginning -- this is frustrating BY DESIGN, not clever BECAUSE OF design.
A great example of this is the VERY NEXT problem: "1+2+3+...+999+1000"
Given the previous cues (? or _ to show a numerical answer is required, ... to show a sequence to be named), the expectation is that this should be named... however, the author really wants the sum of integers 1-1000. Again, this could have been solved easily with a simple, CONSISTENT cue.
Imagine trying to onboard users and you just say "a date box" without any indication of what they're supposed to use it for or what the consequences of it are.
This game does not build any confidence and should not be likened to good UX design at all.
Also my point was that the goal of the first few tasks was not to test your intelligence or resourcefulness, but just to get you accustomed to the environment, the screen etc. I would compare it to "small talk" you would do with someone you meet the first time, so that you get familiar with the mannerisms of that person. It is merely polite.
Good. Snapchat didn't explain anything, people figured it out. It adds to their curiosity. It doesn't need to be consistent to be successful, and it's not just survivorship bias.
All that matters is that the software solves a problem or taps into a user's need. In this case, this game taps into the need to satisfy curiosity.
EDIT: Ah right, people just want to complain. Go for it! Let's ignore the success that Snapchat represents with its inconsistent UI, or the fact that Facebook has ~2B users and an inconsistent decision tree, or that Git is the most popular version control software by far yet has an inconsistent interface. Ditto for Unix.
Ultimately, when we complain about things that are proven not to matter, it becomes hard to take the complaints seriously. If you can reach 2B users without a consistent decision tree, I think it's pretty safe to say that consistency simply is not an issue, no matter how much our cat-like desire for order and perfection believes otherwise.
The thread we're in is both a counterexample and rebuttal to every one of your points. The game is near the top of HN. Everyone loves it. I get that you're trying to argue against the idea that it represents good UX, and yet UX is an integral component of gameplay. This game was successful because it's fun and it had a good experience, just like HN.
I think the game has terrible UX but I also think it has compelling points (competition, bragging ["never completed"], brain teasers, a need to overcome frustration, etc.) that override the terrible UX.
I don't think I ever mentioned that the game would or should fail due to UX -- my primary point is that it doesn't represent a good user onboarding experience. The very things that make it a compelling game would make it a terrible onboarding experience.
And it turns out want your phone number.
It will truly be a never completed game.