Suppose that you chose a bean of color \(c\). Then, the minimum possible deliciousness is the minimum deliciousness among the beans of color \(c\).

Thus, this problem requires to find for each color the minimum deliciousness of a bean of that color, and choose the maximum value among them.

One can obtain the correct answer by scanning all the \(N\) beans for each color, but there are at most \(N\) distinct colors, so implementing this naively costs \(O(N^{2})\) time, which exceeds the execution time limit.

In this problem, you can use an associative array (std::map or std::unordered_map in C++, or dict or deafaultdict in Python) to manage the minimum deliciousness of each color, so that the program runs in \(O(N\log{N})\) or expected \(O(N)\) time. Specifically, iterate the given beans; if the associative array contains the index of the color of the current bean, update the minimum deliciousness; otherwise, add a new index of the color and let the minimum value the deliciousness of that bean. This way, you can obtain the minimum deliciousness of each color in the end.

For more details, please refer to the sample code.

Sample code (C++):

#include <bits/stdc++.h>
using namespace std;

int main() {
  int n;
  cin >> n;

  map<int, int> minimum_taste; // Minimum deliciousness of each color
  for (int i = 0; i < n; ++i) {
    int a, c;
    cin >> a >> c;
    if (minimum_taste.contains(c)) {
      // Update the minimum value if color `c` already exists
      minimum_taste[c] = min(minimum_taste[c], a);
    } else {
      // Add new color
      minimum_taste[c] = a;
    }
  }

  int ans = -1;
  // Scan all the existing colors
  for (auto [c, val] : minimum_taste){
    ans = max(ans, val);
  }

  cout << ans << endl;
}