Is there a more efficient algorithm besides Gram-Schmidt that would produce an orthonormal matrix of rank N, with first row equal to [1 1 1 1 1 ... 1] / sqrt(N)?
e.g. for N = 3, the matrix $\begin{align} \mathsf A_3 &= \begin{bmatrix} 1/\sqrt 3 & 1/\sqrt 3 & 1/\sqrt 3 \\ 2/\sqrt 6 & -1/\sqrt 6 & -1/\sqrt 6 \\ 0 & 1/\sqrt 2 & -1/\sqrt 2 \end{bmatrix}\end{align}$ suffices, but I'm not sure how to generalize.