Lately, I have been asked some version of the same question over and over:

“What prompt did you use?”

I understand the instinct. It feels like there must be a magic incantation, a carefully crafted block of text that unlocks the right answer from a language model. If only we could bottle the perfect prompt, we would get perfect outputs every time.

But here is the uncomfortable truth.

Crafting the “perfect” prompt is mostly a vanity exercise.

Language Models are built for conversation

Modern language models are designed to follow the flow of natural dialogue. They do not require ritual. They do not demand precision engineered syntax. They respond remarkably well to ordinary, iterative conversation, the same way a colleague would.

Think about how we actually work with humans:

  • “What if we conflate metric X into the report, does that change the conclusion?”
  • “No, I meant extracting the entire query result into a DTO, not just the settings part.”
  • “Can you simplify that explanation?”
  • “Let’s take a different angle.”

We do not stop to design the perfect three paragraph instruction before speaking. We interact. We refine. We clarify.

LLMs work the same way.

Iteration is cheap

The old mental model treats prompts as expensive artifacts. That made sense when model calls were slow, costly, or unreliable.

That is no longer the world we are in.

Responses are fast. Iteration is cheap. Clarification takes seconds.

Instead of spending five minutes crafting a perfect prompt, you can ask naturally, see what comes back, and adjust. The feedback loop is tight enough that over optimisation upfront is often wasted effort.

Prompts are not spells

There is a subtle status game in sharing prompts, like sharing a secret recipe. It implies that results come from hidden craftsmanship rather than from collaborative iteration.

In reality, most good outcomes emerge from back and forth. Adjusting assumptions. Challenging outputs. Injecting missing context. Correcting misunderstandings.

The prompt is just the starting point. The conversation is the work.

The next time someone asks, “What prompt did you use?”, it may be worth reframing the question. The more interesting story is rarely the first message. It is the sequence of clarifications, corrections, and refinements that followed.

That is where the value lives.