We use "frog DNA" to fill in gaps.

When confronted with ambiguity, we use our pre-existing knowledge of the world to guess at the resolution to the ambiguity.

Like in Jurassic Park where the dinosaur DNA has gaps that they fill with frog DNA.

LLMs have broad world-knowledge to draw from to fill in gaps.

Your prompt has to share all of the out-of-distribution specifics for the LLM to get the specific background and not just give you mostly a frog DNA answer.

If the thing you're trying to do is in-distribution the prompt can be small and draft off the existing understanding.