WHILE the world watches the ex-Soviet Union transform into a new commonwealth, the United States is quietly making big changes in something that now seems a cold-war anachronism: nuclear weapons planning and strategy.This week the Department of Energy announced it will drastically reorganize the US atomic arms production complex, slashing its work force in half in anticipation of a much smaller future arsenal. "Nobody likes nuclear bombs. We're trying to get rid of them," said Energy Secretary James Watkins. Meanwhile, defense against ballistic missiles is suddenly in vogue. In a first, Congress has endorsed the concept of deploying a limited defense system; and after years of "star wars" derision, the Strategic Defense Initiative Organization is accelerating work to meet a 1996 deadline. The congressional mandate "was a real landmark," said SDIO director Henry Cooper at a meeting with defense reporters. Thus, the outline of the US nuclear arsenal of the 21st century is beginning to take shape. No longer will it consist of a massive stockpile pointed largely at a single nation. In its place will be a small maintenance number of long-range strategic warheads - with likely a thin defense to guard against accidental launches or third-world dictators. "We're talking about a shift in national strategy that wasn't anticipated," says Secretary Watkins. The future of the US nuclear weapons production complex has been up in the air for years. At the height of the cold war it consisted of some 100 sites scattered throughout the United States, but in recent decades many facilities have aged and deteriorated. For all practical purposes much US nuclear production has been shut down since 1988 for safety reasons. The Reagan administration, counting on continued rapid modernization of nuclear weapons, at one point planned a massive $81 billion renovation of the production complex. Continued management upheaval, revelations of environmental damage, and budget cuts derailed that program. Now the Bush administration has decided to consolidate the current 15 nuclear complex sites around the nation instead. Two plants that produce nonnuclear bomb parts will close by 1996: a Clearwater, Fla., site and the Mound Plant near Dayton, Ohio. Nuclear fabrication work at the controversial Rocky Flats, Colo., plant may end soon, as well. Watkins indicated that only the new Trident 2 submarine missile warhead would be made there, and that the administration was reviewing the need for Trident 2. The total number of workers assigned to nuclear activities will be reduced from about 30,000 now to 15,000 by the year 2005. But during that same time frame Department of Energy nuclear cleanup employment is projected to rise by at least 16,000. With the need for new warheads greatly reduced, the weapons-design activities of the four US nuclear development laboratories might even be consolidated, Watkins indicated. The failure of the Soviet coup attempt was important, he said, "because this gave us the confidence ... to go ahead and continue moving down the nuclear weapons stockpile." Insofar as it hastened the rise of the Soviet republics, the failure of the coup has also improved prospects for deployment of strategic defenses. According to SDIO director Cooper, officials from the republics "seem to be more interested" in some sort of mutual move away from mutual assured nuclear destruction to a mixture of offensive nuclear weapons and missile defenses. Cooper said he personally believes it would be possible for a commonwealth of independent states and the US to build complementary star-wars systems that "provide a common defense against the third-world missile problem." Much of SDI's recent rise in fortune can be attributed to fear of rogue dictators. Revelations about Saddam Hussein's nuclear program made this fear seem justified, and pictures of Patriot missiles downing Scuds made defense against ICBMs seem feasible. Critics point out that defense against ICBMs is much more difficult than blocking slower Scuds, and that no nation hostile to the US can be expected to marry the difficult technology of a nuclear weapon with that of an intercontinental missile until well into the next century, if then. Congress this year voted to move toward deployment of a single-site anti-missile system, probably in Grand Forks, N.D., by 1996. Some key technologies will have to be accelerated by as much as two years to make that deadline, according to Cooper. Total cost of building this system will be about $25 billion, according to Pentagon estimates. That doesn't include the research money spent on the SDI program since it began in 1983. A larger system capable of defending the whole nation would require some changes in the ABM Treaty. A six-site system with 750 interceptor rockets, plus 100 space-based interceptors, is projected to cost anywhere from an extra $50 billion to $120 billion.