(FORTUNE Magazine) – The Defense Advanced Research Projects Agency -- Darpa, for short -- may have & done more for U.S. competitiveness than any other organization. The Pentagon agency started or funded some of the most spectacular high-tech weapons that brought Saddam Hussein to his knees. In the process, Darpa successfully shepherded underlying technologies that have spilled over into the commercial world: supercomputers, graphics workstations, artificial intelligence, composite materials, digital gallium arsenide circuits, and plenty of other products and systems. Even the PC mouse was born in a Darpa project. Darpa, based in Arlington, Virginia, has no labs of its own. It dispenses an annual budget of $1.5 billion largely through development contracts with universities, private companies, or federal labs. Says former director Robert Cooper: ''Darpa probably has the largest pot of unrestricted money in the government or even in industry to do that. That's where its power lies.'' In what the Pentagon calls ''dual use'' technology, valuable in defense and commerce alike, Darpa's contribution to U.S. aviation alone would probably justify the agency's existence. Darpa poured hundreds of millions into complex supercomputer software programs that allow aerodynamic testing of large portions of airframes inside computers, instead of wind tunnels. One big reason for Darpa's success is that it operates unlike any federal agency. About half its tiny staff of 140 are program directors who act as technical entrepreneurs, not turf-protecting bureaucrats. These scientists and engineers come to Darpa for three years or so from industry, the military, or universities; they leave when their programs are complete. The agency's gambles on nascent technologies have helped create dozens of outstanding high-tech companies. Support of early research at Stanford led to the creation of such successful pioneers as Silicon Graphics, Sun Microsystems, and MIPS Computer Systems. ''We could not have done it without Darpa,'' says MIPS chief scientist John Hennessy. One payoff for U.S. competitiveness: Silicon Graphics alone earned nearly half its $482 million in revenues last year from overseas sales. Craig Fields, Darpa's outspoken director from 1989 to 1990, now heads MCC, the information technology consortium in Austin, Texas. He mused recently that a Darpa contract has often served as a ''Good Housekeeping seal of approval'' for the technology, management, and technical staff of struggling startup companies. Fields helped persuade Congress to pass legislation that allowed % Darpa to share in a company's future profits. But when he invested $4 million last year in Gazelle Microcircuits, a Silicon Valley gallium arsenide manufacturer, ''Sununu went bananas,'' says Daniel Greenberg, editor of the Washington newsletter Science & Government Report. To the Administration, the deal amounted to an industrial policy; soon afterward, Fields was forced out. Darpa went through some months of turmoil and uncertainty, but now it seems solidly back on course. Concludes Edward E. David Jr., a former Bell Labs executive who served as President Nixon's science adviser: ''Among all the outfits that dispense public money, this one has produced the most.''