Scientists use parallel processing supercomputers--some of which can perform trillions of operations per second--to tackle giant problems like the way a protein folds, how weather patterns shift, and the big bang. IBM last week announced it's putting its $29 million a year of supercomputing research under one virtual roof, called the Deep Computing Institute (DCI). Named after Deep Blue, the chess-playing program that defeated Grand Master Gary Kasparov, the institute wants "to take the modeling that scientists are used to and extend it to business decision-making," such as managing investment risk and scheduling airline routes, says William Pulleyblank, DCI's director.

The institute also wants to get academic scientists involved, starting with a Web site where it has just released the source code to IBM Visualization Data Explorer--software for turning data crunched on a supercomputer into understandable pictures. The site will also offer other freebies, like IBM problem sets for testing algorithms, and DCI will sponsor scientific conferences. That's good both for IBM--which will generate demand for its supercomputers--and for outside scientists, who can now tinker with and improve the software, says Rice University's Ken Kennedy, a member of DCI's advisory board of scientists and business leaders.