A debugging life that starts with std::sort

Jigu-i trusts std::sort completely. Writing a fast sort by hand takes more than $20$ lines, but std::sort is one line. Using it on an array of structs needs extra work such as defining comparison operators, but that drawback is nothing next to writing a sort from scratch.

So Jigu-i codes while shouting S.M.T. every day. One day the program came back Wrong Answer for no clear reason. The task is simple.

There are $N$ points in the region $x \ge 0$, $y \ge 0$ of the plane. Sort them so that for every $i < j$, the $i$-th point $(x_i, y_i)$ and the $j$-th point $(x_j, y_j)$ satisfy $x_i y_j \le x_j y_i$.

Jigu-i treated $l = (l_x, l_y)$ as coming before $r = (r_x, r_y)$ when $l_x r_y < l_y r_x$, then called std::sort. The result was still wrong. The code is only std::sort, and feeding data by hand never produced a counterexample.

Print a counterexample for Jigu-i. The required output is unique. $N$ is $2$, and the two points are $(1, 0)$ then $(0, 1)$.

There is no input.

Print the number of points $N$ on the first line. $N$ must satisfy $2 \le N \le 1000$. In this problem $N$ is $2$.

On the next $N$ lines print an $x$ coordinate and a $y$ coordinate. Each coordinate must satisfy $0 \le x \le 1000$ and $0 \le y \le 1000$. The two points are $(1, 0)$ then $(0, 1)$.

The judge compiled Jigu-i's code with libstdc++.