https://iguessthatworks.com/tags/culture/Recent content in Culture on It Might be WorkingHugo -- gohugo.ioenJeff MayeurTue, 02 Jun 2026 07:00:00 -0700https://iguessthatworks.com/posts/06-2026/a-and-r-block/Tue, 02 Jun 2026 07:00:00 -0700https://iguessthatworks.com/posts/06-2026/a-and-r-block/ <p>Today’s musing centers around <a href="https://bsky.app/profile/carnage4life.bsky.social/post/3mnatmvrw7s2q">this response</a> to Steve Yegge’s Opus 4.8 critique. I’ve pulled back to Sonnet for most of what I do these days; partly that’s a cost choice and partly because I don’t have a meaningful use case for an advanced thinking model.</p> <p>Steve’s critique and Dare Obasanjo’s summation fit within one of the limits I’ve been seeing. I’ve identified it as the inability of a model to know when to be intentionally wrong. I understand that, with additional training data or reasoning layers, a model could be tuned to a category of voice, but it can’t reproduce my voice; even if trained on my writings, it couldn’t reproduce how I will want to express myself in a year. Yesterday my voice may have been incandescent irreverence; tomorrow it might be spurned arrogance. I’m not even a professional writer, yet I’ve found it too easy to confound Opus (<a href="https://iguessthatworks.com/posts/05-2026/aiq-scale/?query=spurge">AIQ Scale</a>).</p>