A Question Called 'Is It Possible?'

“It should work”

There’s a question you hear a lot at work: “Is X possible?” And an answer you hear just as often: “It should work.” “That seems hard.”

On the surface it’s a perfectly normal exchange — a question asked, an answer given. But I’d say nothing actually happened in that conversation. The person who asked walked away with no grounds for a decision, and the person who answered just voiced a hunch. Yet the meeting steps over that sentence and moves on to the next item, and a few weeks later someone ends up saying “but you said it would work.”

Half the problem is in the shape of the question itself. “Is it possible?” assumes everything in the world falls neatly into possible or impossible. But the real world’s answer is almost always somewhere in between.

Possibility is not an opinion

For some X to be possible, several conditions usually have to hold at once. Call them A, B, C, D, and E. Take them apart one by one and they tend to split three ways: things that work today, things that work if you spend money and time, and things that are physically off the table.

A and B work today. C and D cost something, but nothing rules them out. E, though, has no path with current technology. Then the answer about X falls out on its own: “not possible right now, because of E.”

There’s quite a distance between that answer and “that seems hard.” The first is a conclusion you can check by retracing it; the second is just a hunch. A conclusion can be examined by others — is anything missing from the condition list, is E really a dead end, does each call hold up? A hunch allows none of that. The only thing that can stand up to a hunch is another hunch, so all you get back is “well, I think it’ll work.”

Possibility is not an opinion; it’s a function of conditions. And if it’s a function, you should be able to compute the answer, not memorize it.

A duel of hunches

Everyone has seen how a meeting goes when only conclusions get exchanged, with no conditions attached.

The moment “it should work” collides with “that seems hard,” the meeting stops being an exchange of reasoning and turns into hunches arm-wrestling. Hunches can’t verify each other, so the contest gets decided by all the wrong things: whose voice is louder, who has more seniority, who looks more confident. And a conclusion settled that way starts wobbling the moment people walk out of the room.

Bring decomposition in, and the same collision plays out differently. “I’m negative on this because C looks hard” against “couldn’t we solve C this way?” is a healthy collision — it’s a dispute you can settle by checking. What’s left at the end of the meeting isn’t “who won” but a piece of work: “let’s find out what C actually costs.” Turning clashes of opinion into verifiable work — that, I think, is the biggest change decomposition brings to a meeting.

Most impossibility has an expiration date

Decomposition gives you one more thing, and I’d say this one matters more: you learn when the answer changes.

“Not possible right now, because of E” is a conclusion with an expiration date written on it. The moment E comes unstuck, X moves from a question of impossibility to a question of cost. And looking back at the history of technology, E comes unstuck more often than you’d think. When GPS and mobile internet landed on smartphones, “hail a passing car with an app” went from fantasy to business. When LLM inference costs dropped by orders of magnitude, products that had been impossible to build on top of them poured out. The people who had memorized only the conclusion walked right past the moment the condition gave way.

So I like to compress all of this into one line: cache the conditions, not the conclusion.

Store only the conclusion “X is impossible,” and that cache never expires, no matter how much the world changes. Years later, whenever X comes up, “that won’t work” fires off reflexively. But for someone who remembers it with the condition attached — “X doesn’t work because of E” — a news item skimmed in passing becomes a signal. A price cut, a newly opened API, a regulation change, and they notice on their own: “huh, E just got unblocked. Time to look at X again.” That’s why the same news is mere information to one person and an opportunity to another.

Verdicts allocate other people’s time

None of this means the question “is X possible?” is itself bad. Someone who doesn’t know asking someone who does is only natural, and for a junior, asking it is part of learning. But when a senior passes that question along to the next person without decomposing it, that’s a different matter. Breaking a question down into conditions is precisely the work expected of a senior. The same sentence is a question from one person and, from another, an act of handing off their own job.

The higher the seat of judgment, the heavier this gets. A verdict of “possible” or “impossible” is, in itself, an act of allocating other people’s time. A leader who says “impossible” without decomposing erases options the team could have had; a leader who says “possible” without decomposing makes promises the team will have to keep. Either way, the price is paid not by the one who ruled, but by the ones who execute.

And if you think about which judgments people end up trusting, the answer is the verifiable ones. A verdict presented along with its conditions can be examined by others, and trust builds when it survives the examination. Authority can come from a title, but trust comes from the number of times you’ve passed verification.

Even so

That said, decomposition doesn’t guarantee the right answer.

The condition list itself can be wrong. You might lay out A through E, finish your verdict, and find there was a hidden condition F all along. The E you declared impossible might turn out to have a workaround. You aren’t always given time to decompose every judgment either, and some decisions are rightly made fast, on gut feel.

But a decomposed judgment, even when it’s wrong, lets you trace where it went wrong. A condition list can be fixed, and the fixed list gets reused in the next judgment. A hunch leaves nothing behind when it misses. With no material to review beyond “that’s what I thought at the time,” you can repeat the same mistake without ever noticing.

So decomposition is less a technique for being right than a technique for leaving your judgment in a form you can learn from when you’re wrong. The more people speak in that form, the more meetings drift from duels of hunches toward verification of conditions. And I believe it’s the conclusions from those meetings that survive outside the room.