Three prompts. Three completely different reviews. One theme.
Three prompts. Same theme. Completely different outcomes. The difference isn’t the AI—it’s the prompt. This pattern shows up when systems appear to work but are never forced to reveal their real gaps. The failure appears late, but the cost starts earlier, in the structure of the question itself. At that point, there are two paths: validate what exists, or force the system to expose what it cannot do. The first preserves confidence and guarantees failure later. The second restructures the problem around what must be true for the output to matter. The result is dramatically different output quality. The cost of delay isn’t incorrect answers—it’s building on assumptions that were never challenged. If your system consistently confirms what you expect, you’re already in that pattern. And at that point, the decision isn’t whether to change—it’s whether you change intentionally or under pressure.