"My responses are limited. You must ask the right questions." I, Robot.
Many talk about crafting the perfect prompt for AI or SI. Get the wording right, add a role, set some constraints, and the output improves.
It may be right, but is it what you want?
The bigger problem is asking the wrong question in the first place. Or the right question without enough context.
Say you want flight information for Starship to the Moon. Payload, transit time, delta-v, that kind of thing. You get a solid answer back. Numbers look reasonable. Assumptions seem consistent. Nothing jumps out as wrong.
Except the model used current V3 capabilities because you never said otherwise. You were actually thinking about what V4 might do in the same scenario.
The answer is correct. It's just correct for a different version than the one you had in mind.
If you don't notice the mismatch, you start building plans around the wrong baseline. Options that look viable aren't. Timelines shift. You waste effort chasing something that was never on the table.
AI will fill gaps with whatever it treats as current or default. It doesn't know which details you left out on purpose and which ones you simply forgot to mention.
A sharp prompt can give you a precise response to the question you asked. It can't automatically supply the context you didn't provide.
The real skill is making those assumptions explicit before you start. Version, timeline, constraints, what counts as success.
Otherwise you can get a beautifully accurate answer that still points you down the wrong path.
Sound familiar? Have you ever followed an answer that was right on the facts but wrong for what you actually needed?
Ask the right question, and maybe the responses aren't so limited.馃