I'm assuming that you're using "Universal Search" to refer to Toby Ord's Complete Turing Machine.

As others have pointed out, the ranked voting method implemented in the US is instant runoff voting, which generally isn't vulnerable to burying and thus doesn't suffer from the DH3 pathology (the scenario where a universally disliked candidate wins due to tactical voting).

The author doesn't make it clear which method they based their examples on, but they would be most applicable to the Borda count. That being said, since they refer to "the case where ranked choice voting is simply incoherent", I'm guessing they used a Condorcet method with no fallback rule.