It is the ultimate playground for academics with big dreams: $3 billion to do whatever you want, as long as it's in the interests of defending the United States. An open chequebook, and no peer review, has meant great success for the Pentagon's research wing, the 250-strong Defense Advanced Research Projects Agency (DARPA). Over the past half-century, it has been the place where the best and the brightest gather to crack some of the toughest nuts in the defence community: how to build radar-defying stealth aircraft; how to detect hostile missile launches; how to build the best satellite-navigation system so that soldiers can find their way. 'DARPA hard' is the phrase that's used — if it's not hard enough for DARPA to do, it's not worth the agency doing it.

DARPA's accomplishments are legendary, so much so that other areas of the government are hoping to mimic its success with their own DARPA spin-offs (see pages 390 and 403). This is not an easy task but not necessarily a bad idea; at the very least, some parts of the government not known for innovation might begin to think about technological issues from a creative and higher-risk standpoint. But what is the future of DARPA itself?

Its 50-year mark is as good a time as any to take stock. Like the rest of the United States, DARPA can no longer afford to operate under a cold-war mentality; the terrorist attacks of 11 September 2001 and their aftermath have dramatically restructured the country's national defence needs. DARPA stumbled badly in one of its first and most prominent responses to the attacks by setting up an invasive information-mining programme run by a man convicted of multiple felonies (later reversed on a technicality) relating to the Iran–Contra scandal of the 1980s.

Today, some of DARPA's activities remain firmly rooted in its cold-war research past, which led to the unmanned drones flying surveillance missions in Iraq. But, as was on show in a California Disneyland hotel last year, the agency's director, Anthony Tether, has also tried jazzing things up a bit, by sponsoring competitions such as the $2-million 'grand challenge' robot car race. Even his critics note that Tether has managed to draw younger researchers into an agency whose stalwart backers are growing greyer every year.

Such approaches may be useful in helping the government think outside of the box. But DARPA must be careful not to stray too far down that path. The agency's future will depend in large part on the actions of Tether's successor: the position of DARPA director is a powerful one, with the ability to select programme managers and dispense money as he or she sees fit.

The agency's next director must take the agency back to basics, or risk losing the edge that comes from the inspired patronage of risky research. Above all, DARPA needs to concentrate more on projects that could lead to long-term pay-offs in the fight against terrorism, in ways we cannot yet imagine.