Everybody hates Opus 5, but I don’t

# מהמקור

First off, I haven’t noticed a a significant difference in O5’s interactions with me compared to other models. Most of my work was a knowledge acquisition and synthesis, however (I don’t code). I’ve run extensive (“deep”) Research comparisons and found that Opus 5 at Medium effort produces high quality Research reports consistently (although Sonnet 5 at Low effort beats it when synthesizing a report from multiple Research reports due to a) higher variability in results and b) lower cost per run). I finally got around to updating the “Cross-Instance Review” function of my Cross Conversion skill, which I primarily use for building and updating skills and Project instructions, which I last updated after the release of Opus 4.8. At that time, the