Abstract Microlensing planet searches have discovered a total of 20 planets, including the first two Jupiter-Saturn like systems and the only 5 "cold Neptunes" yet detected. The discovery process is almost unbelievably chaotic, with the so-called "high-magnification events" being the most chaotic. I show, nevertheless, that the high-magnification subsample constitutes a "controlled experiment", which enables rigorous statistical analysis, yielding important new clues to planetary architecture. I also discuss the current transition to "second generation" microlensing experiments, which is already yielding important new discoveries, and will in the future explore domains of planet parameter space not probed by any other method.