Personally I would consider it a 3-way tie between Codes, Isogenies, and multivariate schemes. Each family has schemes that seem quite robust, and more aggressive wonky schemes.

Nice blogpost, but I somewhat disagree with "We need to have a better understanding of what the hard problem underlying UOV actually is." The problem is IMO clear and fairly natural: given some quadratic forms sampled uniformly from those vanishing on a secret linear subspace, find that subspace.