#include<bits/stdc++.h>
using namespace std; 
char nl = '\n';
using i64 = long long;

vector<bool> prime(1e7 + 1, true);
vector<int> pref(1e7 + 1); 
void sieve() { 
    prime[0] = prime[1] = 0;
    for (i64 p = 2; p * p <= int(1e7); p++) {
        if (prime[p] == true) { 
            for (i64 i = p * p; i <= int(1e7); i += p) prime[i] = false; 
        } 
    }
}

void solve(int t) {
    // cout << "test #" << t << << nl;
    int l, r; cin >> l >> r;
    cout << pref[r] - pref[l - 1] << nl;
} 

int main() {
    sieve();
    for(int i = 2; i <= int(1e7); i++) {
        pref[i] = pref[i - 1];
        if(prime[i] && prime[i - 2]) pref[i]++;
    }
    int tt = 1;
    cin >> tt;
    for(int t = 1; t <= tt; t++) solve(t);
}

The contest feels a bit math-heavy. I do feel like it could have covered a few more algorithmic topics than this, E is a bit unnecessary (I meant, I might be in the dark, but who are the participants of this contest to have a HelloWorld-like problem? If they are freshmen, then sure, it's acceptable), but overall the maths weren't so bad.

Speaking of E, I was kinda confused at first as the statements and the outputs didn't really linked together. The backstory seemed to imply us to try spelling Exactly, while the actual output is a different story.

Problem constraints seemed a bit inconsistent. B could have had $$$n \le 2 \cdot 10^5$$$ without harming the intended solution. And why did D have $$$n \le 10^6$$$? I don't get the idea of what you're trying to block with this constraints, but AFAIK, the most braindead solution for D is a very high coefficient STL-based $$$\mathcal{O}(n \log n)$$$ — and it still worked just fine here. It could have been $$$2 \cdot 10^5$$$ and thus removing the hassle of fastio.

Problem statements could have made use of LaTeX more. I think the writer just learnt it recently, and could only apply it sparsely and yet without a regulation of when to use those. It's alright, it didn't harm the overall readability, but I hope they could do better next time.

The rest is standard educational stuff, so I don't have much to comment.

At any rate, I really appreciated these contents, regardless of how basic they are. After all, newbies tend to need those building blocks the most.

UPD: Edited recommended constraints to $$$2 \cdot 10^5$$$ because sometimes weird pragmas actually pushed $$$\mathcal{O}(n^2)$$$ bruteforce past 1s TL if $$$n \le 10^5$$$ only.

to pay a little bit more attention to making the problem statements clearer. For instance when I first read F, I thought that we are looking for numbers that are sums of three different primes, like 2+3+5 for instance. Also in problem G the third rule should have been "The absolute difference between number of tasks of consecutive two days is one".

Other than that I feel the problem set was varied enough in difficulty and style for the intended audience.