The Trump administration often talks about North Korea policy as if it’s an on-off switch. President Donald Trump thundered Tuesday that the U.S. would “totally destroy” North Korea to defend itself and its allies. But Defense Secretary James Mattis blandly insisted the next day that it’s “still a diplomatically led effort.”

Somewhere in this maze of public statements – including Thursday’s announcement of new economic sanctions on Pyongyang – there’s a nuanced U.S. policy. But the seeming binary options are weirdly reminiscent of the nuclear standoff of the Cold War, when the only choices seemed to be a conflict with massive loss of life – or surrender to the adversary’s demands.

To escape this straitjacket, strategists in the 1970s and ’80s began to devise new conventional and nuclear weapons, and ultimately, missile defenses. A similar creative re-examination is needed now.

We can always hope that the Trump administration’s strategy will work: Maybe Trump’s threat to Pyongyang of “fire and fury” will convince China to halt oil deliveries; perhaps the North Koreans will enter negotiations; maybe an interim peace agreement will stabilize the situation so “final status” talks begin about eventual denuclearization of the Korean Peninsula and withdrawal of American troops.

That’s the best outcome, certainly. But to be prudent, U.S. officials and their allies should assume it won’t work. They need to be planning other options, with a coldblooded rationality that is the opposite of schoolboy taunts about “Rocket Man.”

U.S. officials need, first, to decide how serious a threat North Korea truly poses to America. If major cities are at risk, and Kim Jong Un’s erratic behavior can’t be deterred, then perhaps the U.S. should indeed be planning to denuclearize North Korea by force.

If the U.S. adopted this maximalist strategy, it would begin a buildup of forces that, by most estimates, would take at least two months. Japan and South Korea would begin intensive civil-defense programs to protect their populations and minimize civilian casualties.

This is a nightmare scenario, but if you believe Kim is truly a nightmare leader, then you must think about the unthinkable. Japanese Prime Minister Shinzo Abe seems to be doing just that. His op-ed in The New York Times Monday was more polite than Trump’s bombast, but no less firm. He urged “concerted pressure” against “an unprecedented, grave and imminent threat from North Korea.”

Suppose you take a less drastic view of Kim and conclude that the real target of his antics is China. He clearly fears Beijing’s influence: He brutally murdered his uncle and half-brother, both of whom were said to be close to China. His missile and nuclear tests defy Trump, but even more the repeated warnings he has received, and ignored, from President Xi Jinping.

If we see Kim as a regional threat, rather than a global one, then perhaps the right response is an intelligence strategy that begins with the reality of his split with China – and takes off from there. Back in 2003, China suspended oil deliveries for several days (blaming the problem on a supposed pipeline malfunction) and North Korea quickly began negotiations. In the deniable realm of intelligence operations, it’s always possible that a pipeline could “malfunction” again, or that other crippling difficulties could arise for Pyongyang and its mercurial leader.

Suppose, instead, that the U.S. and its allies decide that North Korea isn’t worth the risks of either military or covert action. What then? To be cynically honest, we must recognize that sometimes it’s less costly to bribe an adversary than to go to war. What blandishments would get Kim to agree to halt his testing program? Is there a “freeze” option, as suggested by Robert Einhorn of the Brookings Institution, that would stop escalation, prevent proliferation and stabilize the situation – but leave denuclearization for the distant future?

Finally, are there defensive measures that can sharply reduce the North Korean threat? For the past decade military planners have been touting a “boost-phase” intercept, which could destroy North Korean missiles in the first several minutes after launch.

The Pentagon has started a new program to build a lightweight, powerful laser carried on big, high-altitude drones that could loiter outside North Korean airspace. The lasers won’t be ready until 2023, at the earliest. But how about a simpler version that would shoot fast interceptor missiles from existing drones? Some Pentagon planners say such a system could be deployed soon.

President John F. Kennedy famously solved the Cuban missile crisis in 1962 by thinking outside the box. Similar creative thinking is badly needed now on North Korea.