*Edited 5/7/2018. Thanks to Sam Hopkins for several corrections and suggestions.*

I am revising my notes from the course on “better-than-worst-case” analysis of algorithms. With the benefit of hindsight, in this post (and continuing in future posts) I would like to review again how one applies spectral methods and semidefinite programming to problems that involve a “planted” solution, and what is the role of concentration results for random matrices in the analysis of such algorithms.