You typed your question. The answer came back. Coherent. Confident. Wrong, somehow — not factually wrong, just... not what you needed. So you tried again with more detail. Then again with even more. By the fifth try, you were writing a small essay just to coax the answer to land where you needed it.

You don't need to write an essay. You need three lines.

Every AI interaction is doing the same three things underneath. Most people only show up with one of them.

There's the gap — what's missing. The thing you're asking AI to fill in.

There's the field — the world the gap belongs to. The whole that the missing piece is a part of.

And there's the completion — AI moving through the field to fill the gap.

When prompts disappoint, it's almost never because the question was written badly. It's because the prompter showed up with a gap and skipped the field. AI completes anyway — that's what it does — but now it has to guess the field. It picks the most statistically likely one. Default field, default completion, default answer. That's what "generic" means. It's not that AI is being lazy. It's that you didn't tell it what world your question lived in, so it chose the most probable one. Which is rarely yours.

The 3-Line Method is what every good prompt is doing implicitly, but rarely on purpose:

1. State the gap. What's missing.

2. Define the field. What world it belongs to.

3. Ask for completion. What you actually want.

That's the whole method. Three lines. Maybe thirty seconds longer than what you were doing.

The middle line — the field — is the one almost everyone skips. It's also the one that does almost all the work. Once you put it back in, the difference between mediocre AI and useful AI mostly disappears.

Here's why. The same gap, placed in different fields, produces totally different completions. "Sales are down" inside a marketing field gives you ad-spend recommendations. The same gap inside a trust field gives you a customer-relationship audit. The same gap inside a product field gives you usability questions. AI can do any of these competently. It cannot pick which one is right. That's not an AI problem. It's structurally not what AI does.

You pick the field. AI does everything else.

There's a deeper point hiding here. People say "better prompt = better answer." That's the framing the entire prompt-engineering industry lives inside. It's not wrong, but it's pointed at the wrong thing. Phrasing is downstream. Better field selection = better reality. A perfectly worded question in the wrong field produces a perfectly useless answer. A roughly worded question in the right field produces something you actually use.

Once you see this, the whole game changes. You stop trying to improve your prompts and start trying to improve your fields.

Watch the same gap, twice.

One line, no field:

What comes back? A general checklist. Audit your funnel. Run a customer survey. A/B test your messaging. Optimize your ad spend. Maybe rethink your pricing. It's all reasonable. None of it is sharp. None of it surprises you. None of it tells you anything you didn't already know.

That's what generic looks like. AI completed inside the most likely field — marketing, broadly — and gave you the most likely path inside that field. It's not wrong. It's just default.

Three lines, with a field:

Different world entirely. Now AI is asking different questions about your business. What changed in customer service quality? What recent policy shifts might have signaled disrespect? Have refund or return processes gotten harder? Are your top reviewers cooling off? Is the founder still visible, or has the brand gone corporate-feeling? The recommendations point at relationship repair, not channel optimization.

You didn't change the gap. You didn't write better prose. You didn't use a longer prompt. You named the field — three extra words — and the entire space of possible answers reorganized.

The fix wasn't in AI. The fix was in declaring the world.

The bare method is one prompt:

That's enough to fix most disappointing AI interactions in your life.

The two-minute upgrade is worth knowing too. Before you commit to a field, ask AI:

Pick one. Then run the 3-Line Method using that field. This adds a field-exploration step that costs you ninety seconds and routinely produces answers you couldn't have reached by tightening prompts for an hour.

One diagnostic test. When AI's answer feels obvious, you're probably in the wrong field. Obvious answers mean AI is completing in a default, generic space. The moment you shift to the right field, the answers stop feeling like things you already knew and start feeling like things you needed to hear.

The framework points at field selection as the irreducibly human move. Completion is computable; field selection is what humans bring. The 3-Line Method is essentially: do that move on purpose, and AI handles the rest.

But field selection is a kind of pattern recognition too. You see a situation, and the field that activates is some function of what you've been exposed to, what's been pressuring you, what you're paying attention to, and what you're sensitive to. That sounds — uncomfortably — like a thing AI could eventually do.

Maybe most field selection is already learnable. Maybe AI will get good at suggesting "have you considered framing this as a trust problem?" before you ask. And maybe that's fine — that's just better tooling.

But maybe some part of field selection isn't learnable. Maybe the moments where the right field appears — the click before the answer — are doing something prior patterns can't predict. I don't know which it is. I do know that for now, the field is your move. So make it on purpose.