Stop Asking for Options. Ask for the Trade-Off.

Ask a model for options and it will give you options. Three of them. Each with pros and cons, weighted so evenly that no answer is ever wrong. You read it, nod, and end up exactly where you started — except now you have a document, which feels like progress.

That is not a decision. That is a menu.

The reason is structural, not a flaw in the model. "Give me options" is a request for coverage, and coverage is the one thing it can always deliver. Nothing in that request forces it to say the uncomfortable thing: that one of these costs you something you are not going to want to pay.

Four steps to get that instead.

1. Ban the list

Here is my decision: [describe it].

Do not give me a list of options with pros and cons. Give me the two real candidates and nothing else.

If there is only one real candidate, say so and tell me what I am actually deciding instead.

Three options is almost always one real choice and two decoys. The decoys exist because a list of three looks more thorough than a list of two — not because you were ever going to pick them.

The last line is the one that earns its place. A surprising number of "decisions" dissolve under it: you were not choosing between A and B, you were looking for permission to do the thing you had already chosen, or you were deciding something else entirely one level up. Finding that out costs one prompt and saves the afternoon.

2. Make it name the price

For each of the two, finish this sentence: "Choosing this means giving up ______."

The blank must be something I lose, not something I fail to gain. If you cannot fill it, the two are the same choice wearing different names — tell me that.

The distinction between lose and fail to gain is the whole step.

"Option B is faster to build" is not a cost of option A — it is a benefit of B restated as a complaint. Everything sounds like a trade-off when you phrase forgone upside as loss, which is why so many pros-and-cons tables are symmetrical and useless. Force the sentence into the shape of an actual sacrifice and half the fake trade-offs disappear.

What is left is short, and it is the real decision.

3. Find the condition that flips it

Now tell me what would have to be true about my situation for each one to be clearly right.

Write both conditions as things I can check this week — not as beliefs I hold. Then ask me the single question that separates them.

Most decisions are not close on the merits. They are close because a fact is missing, and both sides are arguing about a gap rather than about the answer.

"Check this week" is the constraint that makes this useful. A condition like "if you value speed over quality" is not checkable — it is a mood. A condition like "if more than a third of your traffic already comes from search" is. When the condition is checkable, the decision stops being a debate and becomes an errand.

Sometimes the errand is small enough to do immediately, and the decision resolves itself in ten minutes.

4. Ask which one you'd regret

Assume I chose the one you think is worse, and it went badly. Write the two-sentence post-mortem.

Then do the same for the other one. I am picking the failure I can live with, not the success I can imagine.

Every option looks fine in the version of the future where it works. That is why imagining success is worthless for deciding — you are comparing two things you designed to be appealing.

Failure discriminates. Read both post-mortems and one of them will land differently: not more likely, but more expensive to have been wrong about. Recoverable and unrecoverable failures are not the same size, and no list of pros and cons has ever told you which is which.

That is usually the moment the decision becomes obvious.

The point

A choice with no cost is not a choice. It is a wish, and the model will happily help you make wishes all day, in a table, with headers.

Asking what something costs is the only question in the set that a balanced answer cannot survive. Ask it first.

---

The four prompts are in the carousel on [Instagram](https://instagram.com/sharpaihub).

Want the calm version of AI news like this, once a week? Subscribe to the Sharp AI Hub newsletter →