Also was wondering the same thing. Also, the linked adjoint method tutorial discusses more general class of problems and it is not immediately clear to me how

Also was wondering the same thing. Also, the linked adjoint method tutorial discusses more general class of problems and it is not immediately clear to me how to map them to this linear equation version: here B is the unknown we want to evaluate and the constraint is set on B; in the pdf tutorial one wants to compute gradient_p f, which sounds similar than the formulation in the blog post, because we usually want the gradient (B?) to evaluate d_p f(x,p) at some point p (u?), but the constraints are given for g(x,p) = 0.