By skyforbes I've been using ChatGPT a lot for technical questions, planning, and coding. One of the most frustrating and time-wasting issues is when it goes crazy and runs off on a wild tangent. If you're not focused on an objective it can be interesting, but usually it's a huge waste of time.
I've tried various tricks like adding "if you're not sure about context, stop and ask questions", but these snippets do nothing. I know you can ask it to be more focused and factual and that can help, but I'm wondering about a strategy that can reduce crazy tangents (not hallucinations).