.... is having weak pretests on not one, but TWO problems in today's Div2 round...

To all the people I hacked on A (including GM Hamed_Ghaffari), I'm sorry. It's such an unbelievably braindead hack. All I had to do to hack you guys was find a number just below a very very big power of 4 (288230376151711674 worked), and then the log functions output some form of overflowed garbage that made your solutions output a number that was twice as big as it should have been.

On C, somehow we had blatantly O(n^2) solutions getting unpunished, for example this one: 298315499

This works well for random tests (because the alternating 1's and -1's make the range of possible sums quite small, so just doing an unoptimised:

~~~~~ for (auto i : st1) { for (auto j : st2) { ans.insert(i + j + arr[pos]); } }~~~~~

works... but it's very obvious how to make a countertest (just have lots of 1's, then 100000, then lots of -1's, then you have a nested loop that runs about 10^10 times, which will obviously TLE in any language)

It's just so sad when tests that are this obvious don't get added to the pretests.