Thank you for the contribution!
Your solution is ok (I suspect the array indexing would look a bit like a magic to a novice, but I do not see much better alternative). Another option is to explain that the stacking functions treat the one dimensional arrays as row vectors, so to obtain the desired result we need

numpy.vstack([A[:,0],A[:,-1]]).T

Unfortunately, I realized transposing has not been introduced in this part yet, so it would require a brief explanation. So if you wish you can add it as an alternative.

I guess array indexing had only been used with slices in the lesson, so I replaced the A[:, [0]] with the equivalent A[:, :1] and likewise for the last column. That shouldn't be introducing anything new into the challenge solution. (Transposing, as you note, would be a new topic.)

I like this solution! Could you just replace the above appearance of A[:,[0]] with A[:,:1], and I will merge.

Thanks!