History suggests that science follows its own timetable, often producing results long after the politicians who approved the funding have left office. Yet curiosity without the prospect of imminent practical application is something biotech investors are increasingly loath to pay for. When the Nature study began, Diversa was on the cutting edge of “ethical bioprospecting”—searching the world for novel environments and enzymes. After merging with a biofuels company, it became Verenium last year, and shifted to the more prosaic task of making commercial enzymes involved in the development of products including animal feed, paper, and fuels.

David Weiner, the assistant director of enzyme technology at Verenium, gave me a tour of the labs, showing me what he calls the “giant funnel”—the process the company uses to sift through nature’s intellectual property for enzymes that can be converted to profits. “We’re not really interested in DNA,” he said, meaning that the focus is on an enzyme’s performance, not its origins.

Whereas the Joint Genome Institute began by sequencing the termite-gut DNA—learning about its underlying structure—and only then tried to identify what might be useful, Weiner’s colleagues threw all the material from the Costa Rican expedition directly into testing, using the funnel approach to separate the most-useful enzymes from the millions of useless ones. Researchers inserted gene fragments into lab bacteria that had been genetically “tamed” to produce whatever enzyme the fragments were programmed to make. They then tested those enzymes on cellulose, to see if they would attack it. Only the winners made it to sequencing. You might think of the Joint Genome Institute as a group of diligent librarians, studying every step along the way. In contrast, a Verenium senior researcher told me, the company takes a “Julia Child approach”—once it has thrown together the ingredients (like termite guts and cellulose), it turns its attention to the final product, with far less focus on the stages in between.

Much of the action takes place in a machine—a type of robot, really—called the GigaMatrix. Clad in steel, the Giga­Matrix looks like a copier from the late 1980s, with two flat TV monitors on top and a door on the side. It can screen up to a million enzymes at a go, easily exceeding in a single day the lifetime performance of a human lab tech. The Giga­Matrix and other machines took the 500 or so most interesting enzymes from the termite gut and narrowed them down to fewer than 100 with potentially practical applications. Those were then tested for their effects on cellulose, modified, and inserted into “factory” bacteria trained to produce large quantities of enzymes while dining on cheap food, such as corn syrup. As the enzymes made their way through the process, every parameter of their growth and efficacy was measured. Only a small percentage proved powerful enough to merit continued investigation; these were stirred into multiple-enzyme “cocktails” to evaluate their speed and efficiency in combination. By the end, Weiner said, just a few enzymes remained in the running for further testing.

Geoff Hazlewood, a former senior vice president and now a consultant to Verenium, told me that the company has currently put aside studying termites for biofuels and has moved on to other potentially lucrative efforts. “You could screen ad nauseam,” he said, “but you can’t commit an infinite amount of resources.” Whatever the termites are doing may be too complicated and fragile to be useful in a large industrial process. There may be genius in the termite gut—Weiner calls it, admiringly, “a whole town”—but the wonders of symbiosis, in themselves, mean little to companies focused on the bottom line. “We want faster, cheaper, more efficient,” Weiner told me.