I'm trying to prove that an inefficient double-lengthening PRG exists, i.e. construct a PRG $G: \{0,1\}^n \rightarrow \{0,1\}^{2n}$
My current approach is to bound the number of poly-time non-uniform algorithms as the number of boolean circuits, which is $2^{n^2}$, then we can say there $2^{n^2}$ functions, we will call them $F_i, 1 \leq i \leq 2^{n^2}$. Then we will have $2^{2n}$ variables, each $t_{i}, 1 \leq i \leq 2^{2n}$, and we will want to bound the $min ( {max_{1 \leq j \leq 2^{n^2}}} (\sum_{1\leq x \leq2^{2n}} t_x F_j(x)) - \sum_{1\leq x \leq2^{2n}} F_j(x)))$
with the constraints that: $\sum_{1 \leq x \leq 2^{2n}} t_x = 2^n$ and $\forall 1 \leq x \leq 2^{2n}, 0 \leq t_x$
My aim is to show that this minimum is surely negligible, but I don't have a lot of background on constraint equations and I'm not sure if perhaps it's the right approach.
