"Okay, fine."

"I want everybody in place as soon as possible."

"Okay," Bobby said.

"I'm serious. In your places."

"For Christ's sake, Ricky, okay, okay. Now will you stop talking and let us work?" Leaving the group behind, Ricky took me across the floor to a small room. I said, "Ricky, these kids aren't the way they were when they worked for me."

"I know. Everybody's a little uptight right now."

"And why is that?"

"Because of what's going on here."

"And what is going on here?"

He stopped before a small cubicle on the other side of the room. "Julia couldn't tell you, because it was classified." He touched the door with a keycard. I said, "Classified? Medical imaging is classified?"

The door latch clicked open, and we went inside. The door closed behind us. I saw a table, two chairs, a computer monitor and a keyboard. Ricky sat down, and immediately started typing. "The medical imaging project was just an afterthought," he said, "a minor commercial application of the technology we are already developing."

"Uh-huh. Which is?"

"Military."

"Xymos is doing military work?"

"Yes. Under contract." He paused. "Two years ago, the Department of Defense realized from their experience in Bosnia that there was enormous value to robot aircraft that could fly overhead and transmit battlefield images in real time. The Pentagon knew that there would be more and more sophisticated uses for these flying cameras in future wars. You could use them to spot the locations of enemy troops, even when they were hidden in jungle or in buildings; you could use them to control laser-guided rocket fire, or to identify the location of friendly troops, and so on. Commanders on the ground could call up the images they wanted, in the spectra they wanted-visible, infrared, UV, whatever. Real-time imaging was going to be a very powerful tool in future warfare."

"Okay ..."

"But obviously," Ricky said, "these robot cameras were vulnerable. You could shoot them down like pigeons. The Pentagon wanted a camera that couldn't be shot down. They imagined something very small, maybe the size of a dragonfly-a target too small to hit. But there were problems with power supply, with small control surfaces, and with resolution using such a small lens. They needed a bigger lens."

I nodded. "And so you thought of a swarm of nanocomponents."

"That's right." Ricky pointed to the screen, where a cluster of black spots wheeled and turned in the air, like birds. "A cloud of components would allow you to make a camera with as large a lens as you wanted. And it couldn't be shot down because a bullet would just pass through the cloud. Furthermore, you could disperse the cloud, the way a flock of birds disperses with a gunshot. Then the camera would be invisible until it re-formed again. So it seemed an ideal solution. The Pentagon gave us three years of DARPA funding."

"And?"

"We set out to make the camera. It was of course immediately obvious that we had a problem with distributed intelligence."

I was familiar with the problem. The nanoparticles in the cloud had to be endowed with a rudimentary intelligence, so that they could interact with each other to form a flock that wheeled in the air. Such coordinated activity might look pretty intelligent, but it occurred even when the individuals making up the flock were rather stupid. After all, birds and fish could do it, and they weren't the brightest creatures on the planet.

Most people watching a flock of birds or a school of fish assumed there was a leader, and that all the other animals followed the leader. That was because human beings, like most social mammals, had group leaders.

But birds and fish had no leaders. Their groups weren't organized that way. Careful study of flocking behavior-frame-by-frame video analysis-showed that, in fact, there was no leader. Birds and fish responded to a few simple stimuli among themselves, and the result was coordinated behavior. But nobody was controlling it. Nobody was leading it. Nobody was directing it.

Nor were individual birds genetically programmed for flocking behavior. Flocking was not hard-wired. There was nothing in the bird brain that said, "When thus-and-such happens, start flocking." On the contrary, flocking simply emerged within the group as a result of much simpler, low-level rules. Rules like, "Stay close to the birds nearest you, but don't bump into them." From those rules, the entire group flocked in smooth coordination. Because flocking arose from low-level rules, it was called emergent behavior. The technical definition of emergent behavior was behavior that occurred in a group but was not programmed into any member of the group. Emergent behavior could occur in any population, including a computer population. Or a robot population. Or a nanoswarm.

I said to Ricky, "Your problem was emergent behavior in the swarm?"

"Exactly."

"It was unpredictable?"

"To put it mildly."

In recent decades, this notion of emergent group behavior had caused a minor revolution in computer science. What that meant for programmers was that you could lay down rules of behavior for individual agents, but not for the agents acting together. Individual agents-whether programming modules, or processors, or as in this case, actual micro-robots-could be programmed to cooperate under certain circumstances, and to compete under other circumstances. They could be given goals. They could be instructed to pursue their goals with single-minded intensity, or to be available to help other agents. But the result of these interactions could not be programmed. It just emerged, with often surprising outcomes.

In a way this was very exciting. For the first time, a program could produce results that absolutely could not be predicted by the programmer. These programs behaved more like living organisms than man-made automatons. That excited programmers-but it frustrated them, too. Because the program's emergent behavior was erratic. Sometimes competing agents fought to a standstill, and the program failed to accomplish anything. Sometimes agents were so influenced by one another that they lost track of their goal, and did something else instead. In that sense the program was very childlike-unpredictable and easily distracted. As one programmer put it, "Trying to program distributed intelligence is like telling a five-year-old kid to go to his room and change his clothes. He may do that, but he is equally likely to do something else and never return."

Because these programs behaved in a lifelike way, programmers began to draw analogies to the behavior of real organisms in the real world. In fact, they began to model the behavior of actual organisms as a way to get some control over program outcomes. So you had programmers studying ant swarming, or termite mounding, or bee dancing, in order to write programs to control airplane landing schedules, or package routing, or language translation. These programs often worked beautifully, but they could still go awry, particularly if circumstances changed drastically. Then they would lose their goals. That was why I began, five years ago, to model predator-prey relationships as a way to keep goals fixed. Because hungry predators weren't distracted. Circumstances might force them to improvise their methods; and they might try many times before they succeeded-but they didn't lose track of their goal.

So I became an expert in predator-prey relationships. I knew about packs of hyenas, African hunting dogs, stalking lionesses, and attacking columns of army ants. My team had studied the literature from the field biologists, and we had generalized those findings into a program module called PREDPREY, which could be used to control any system of agents and make its behavior purposeful. To make the program seek a goal.

Looking at Ricky's screen, the coordinated units moving smoothly as they turned through the air, I said, "You used PREDPREY to program your individual units?"

"Right. We used those rules."

"Well, the behavior looks pretty good to me," I said, watching the screen. "Why is there a problem?"

"We're not sure."

"What does that mean?"

"It means we know there's a problem, but we're not sure what's causing it. Whether the problem is programming-or something else."

"Something else? Like what?" I frowned. "I don't get it, Ricky. This is just a cluster of microbots. You can make it do what you want. If the programming's not right, you adjust it. What don't I understand?"

Ricky looked at me uneasily. He pushed his chair away from the table and stood. "Let me show you how we manufacture these agents," he said. "Then you'll understand the situation better." Having watched Julia's demo tape, I was immensely curious to see what he showed me next. Because many people I respected thought molecular manufacturing was impossible. One of the major theoretical objections was the time it would take to build a working molecule. To work at all, the nanoassembly line would have to be far more efficient than anything previously known in human manufacturing. Basically, all man-made assembly lines ran at roughly the same speed: they could add one part per second. An automobile, for example, had a few thousand parts. You could build a car in a matter of hours. A commercial aircraft had six million parts, and took several months to build.

But a typical manufactured molecule consisted of 1025parts. That was 10,000,000,000,000,000,000,000,000 parts. As a practical matter, this number was unimaginably large. The human brain couldn't comprehend it. But calculations showed that even if you could assemble at the rate of a million parts per second, the time to complete one molecule would still be 3,000 trillion years-longer than the known age of the universe. And that was a problem. It was known as the build-time problem.

I said to Ricky, "If you're doing industrial manufacturing ..."

"We are."

"Then you must have solved the build-time problem."

"We have."

"How?"

"Just wait."

Most scientists assumed this problem would be solved by building from larger subunits, molecular fragments consisting of billions of atoms. That would cut the assembly time down to a couple of years. Then, with partial self-assembly, you might get the time down to several hours, perhaps even one hour. But even with further refinements, it remained a theoretical challenge to produce commercial quantities of product. Because the goal was not to manufacture a single molecule in an hour. The goal was to manufacture several pounds of molecules in an hour. No one had ever figured out how to do that.

We passed a couple of laboratories, including one that looked like a standard microbiology lab, or a genetics lab. I saw Mae standing in that lab, puttering around. I started to ask Ricky why he had a microbiology lab here, but he brushed my question aside. He was impatient now, in a hurry. I saw him glance at his watch. Directly ahead was a final glass airlock. Stenciled on the glass door was MicroFabrication. Ricky waved me in. "One at a time," he said. "That's all the system allows."

I stepped in. The doors hissed shut behind me, the pressure pads again thunking shut. Another blast of air: from below, from the sides, from above. By now I was getting used to it. The second door opened, and I walked forward down another short corridor, opening into a large room beyond. I saw bright, shining white light-so bright it hurt my eyes. Ricky came after me, talking as we walked, but I don't remember what he said. I couldn't focus on his words. I just stared. Because by now I was inside the main fab building-a huge windowless space, like a giant hangar three stories high. And within this hangar stood a structure of immense complexity that seemed to hang in midair, glowing like a jewel.

DAY 6

9:12 A.M.

At first, it was hard to understand what I was seeing-it looked like an enormous glowing octopus rising above me, with glinting, faceted arms extending outward in all directions, throwing complex reflections and bands of color onto the outer walls. Except this octopus had multiple layers of arms. One layer was low, just a foot above the floor. A second was at chest-level; the third and fourth layers were higher, above my head. And they all glowed, sparkled brilliantly.

I blinked, dazzled. I began to make out the details. The octopus was contained within an irregular three-story framework built entirely of modular glass cubes. Floors, walls, ceilings, staircases-everything was cubes. But the arrangement was haphazard, as if someone had dumped a mound of giant transparent sugar cubes in the center of the room. Within this cluster of cubes the arms of the octopus snaked off in all directions. The whole thing was held up by a web of black anodized struts and connectors, but they were obscured by the reflections, which is why the octopus seemed to hang in midair.

Ricky grinned. "Convergent assembly. The architecture is fractal. Neat, huh?" I nodded slowly. I was seeing more details. What I had seen as an octopus was actually a branching tree structure. A central square conduit ran vertically through the center of the room, with smaller pipes branching off on all sides. From these branches, even smaller pipes branched off in turn, and smaller ones still. The smallest of the pipes were pencil-thin. Everything gleamed as if it were mirrored.

"Why is it so bright?"

"The glass has diamondoid coating," he said. "At the molecular level, glass is like Swiss cheese, full of holes. And of course it's a liquid, so atoms just pass right through it."

"So you coat the glass."

"Right. Have to."

Within this shining forest of branching glass, David and Rosie moved, making notes, adjusting valves, consulting handheld computers. I understood that I was looking at a massively parallel assembly line. Small fragments of molecules were introduced into the smallest pipes, and atoms were added to them. When that was finished, they moved into the next largest pipes, where more atoms were added. In this way, molecules moved progressively toward the center of the structure, until assembly was completed, and they were discharged into the central pipe. "Exactly right," Ricky said. "This is just the same as an automobile assembly line, except that it's on a molecular scale. Molecules start at the ends, and come down the line to the center. We stick on a protein sequence here, a methyl group there, just the way they stick doors and wheels on a car. At the end of the line, off rolls a new, custom-made molecular structure. Built to our specifications."

"And the different arms?"

"Make different molecules. That's why the arms look different." In several places, the octopus arm passed through a steel tunnel reinforced with heavy bolts, for vacuum ducting. In other places, a cube was covered with quilted silver insulation, and I saw liquid nitrogen tanks nearby; extremely low temperatures were generated in that section.

"Those're our cryogenic rooms," Ricky said. "We don't go very low, maybe -70 Centigrade, max. Come on, I'll show you." He led me through the complex, following glass walkways that threaded among the arms. In some places, a short staircase enabled us to step over the lowest arms.

Ricky chatted continuously about technical details: vacuum-jacketed hoses, metal phase separators, globe check valves. When we reached the insulated cube, he opened the heavy door to reveal a small room, with a second room adjacent. It looked like a pair of meat lockers. Small glass windows were set in each door. At the moment, everything was at room temperature. "You can have two different temps here," he said. "Run one from the other, if you want, but it's usually automated."

Ricky led me back outside, glancing at his watch as he did so. I said, "Are we late for an appointment?"

"What? No, no. Nothing like that." Nearby two cubes were actually solid metal rooms, with thick electrical cables running inside. I said, "Those your magnet rooms?"

"That's right," Ricky said. "We've got pulsed field magnets generating 33 Tesla in the core. That's something like a million times the magnetic field of the earth." With a grunt, he pushed open the steel door to the nearest magnet room. I saw a large doughnut-shaped object, about six feet in diameter, with a hole in the center about an inch wide. The doughnut was completely encased in tubing and plastic insulation. Heavy steel bolts running from top to bottom held the jacketing in place.

"Lot of cooling for this puppy, I can tell you. And a lot of power: fifteen kilovolts. Takes a full-minute load time for the capacitors. And of course we can only pulse it. If we turned it on continuously, it'd explode-ripped apart by the field it generates." He pointed to the base of the magnet, where there was a round push button at knee level. "That's the safety cutoff there," he said. "Just in case. Hit it with your knee if your hands are full."

I said, "So you use high magnetic fields to do part of your assemb-"

But Ricky had already turned and headed out the door, again glancing at his watch. I hurried after him.

"Ricky ..."

"I have more to show you," he said. "We're getting to the end."

"Ricky, this is all very impressive," I said, gesturing to the glowing arms. "But most of your assembly line is running at room temperature-no vacuum, no cryo, no mag field."

"Right. No special conditions."

"How is that possible?"

He shrugged. "The assemblers don't need it."

"The assemblers?" I said. "Are you telling me you've got molecular assemblers on this line?"

"Yes. Of course."

"Assemblers are doing your fabrication for you?"

"Of course. I thought you understood that."

"No, Ricky," I said, "I didn't understand that at all. And I don't like to be lied to."

He got a wounded look on his face. "I'm not lying."

But I was certain that he was.

One of the first things scientists learned about molecular manufacturing was how phenomenally difficult it was to carry out. In 1990, some IBM researchers pushed xenon atoms around on a nickel plate until they formed the letters "IBM" in the shape of the company logo. The entire logo was one ten-billionth of an inch long and could only be seen through an electron microscope. But it made a striking visual and it got a lot of publicity. IBM allowed people to think it was a proof of concept, the opening of a door to molecular manufacturing. But it was more of a stunt than anything else.

Because pushing individual atoms into a specific arrangement was slow, painstaking, and expensive work. It took the IBM researchers a whole day to move thirty-five atoms. Nobody believed you could create a whole new technology in this way. Instead, most people believed that nanoengineers would eventually find a way to build "assemblers"-miniature molecular machines that could turn out specific molecules the way a ball-bearing machine turned out ball bearings. The new technology would rely on molecular machines to make molecular products. It was a nice concept, but the practical problems were daunting. Because assemblers were vastly more complicated than the molecules they made, attempts to design and build them had been difficult from the outset. To my knowledge, no laboratory anywhere in the world had actually done it. But now Ricky was telling me, quite casually, that Xymos could build molecular assemblers that were now turning out molecules for the company. And I didn't believe him.

I had worked all my life in technology, and I had developed a feel for what was possible. This kind of giant leap forward just didn't happen. It never did. Technologies were a form of knowledge, and like all knowledge, technologies grew, evolved, matured. To believe otherwise was to believe that the Wright brothers could build a rocket and fly to the moon instead of flying three hundred feet over sand dunes at Kitty Hawk.

Nanotechnology was still at the Kitty Hawk stage.

"Come on, Ricky," I said. "How are you really doing this?"

"The technical details aren't that important, Jack."

"What fresh bullshit is this? Of course they're important."

"Jack," he said, giving me his most winning smile. "Do you really think I'm lying to you?"

"Yes, Ricky," I said. "I do."

I looked up at the octopus arms all around me. Surrounded by glass, I saw my own reflection dozens of times in the surfaces around me. It was confusing, disorienting. Trying to gather my thoughts, I looked down at my feet.

And I noticed that even though we had been walking on glass walkways, some sections of the ground floor were glass, as well. One section was nearby. I walked toward it. Through the glass I could see steel ducting and pipes below ground level. One set of pipes caught my eye, because they ran from the storage room to a nearby glass cube, at which point they emerged from the floor and headed upward, branching into the smaller tubes. That, I assumed, was the feedstock-the slush of raw organic material that would be transformed on the assembly line into finished molecules.

Looking back down at the floor, I followed the pipes backward to the place where they entered from the adjacent room. This junction was glass, too. I could see the curved steel underbellies of the big kettles I'd noticed earlier. The tanks that I had thought were a microbrewery. Because that's certainly what it had looked like, a small brewery. Machinery for controlled fermentation, for controlled microbial growth.

And then I realized what it really was.

I said, "You son of a bitch."

Ricky smiled again, and shrugged. "Hey," he said. "It gets the job done." Those kettles in the next room were indeed tanks for controlled microbial growth. But Ricky wasn't making beer-he was making microbes, and I had no doubt about the reason why. Unable to construct genuine nanoassemblers, Xymos was using bacteria to crank out their molecules. This was genetic engineering, not nanotechnology. "Well, not exactly," Ricky said, when I told him what I thought. "But I admit we're using a hybrid technology. Not much of a surprise in any case, is it?" That was true. For at least ten years, observers had been predicting that genetic engineering, computer programming, and nanotechnology would eventually merge. They were all involved with similar-and interconnected-activities. There wasn't that much difference between using a computer to decode part of a bacterial genome and using a computer to help you insert new genes into the bacteria, to make new proteins. And there wasn't much difference between creating a new bacteria to spit out, say, insulin molecules, and creating a man-made, micromechanical assembler to spit out new molecules. It was all happening at the molecular level. It was all the same challenge of imposing human design on extremely complex systems. And molecular design was nothing if not complicated.

You could think of a molecule as a series of atoms snapped together like Lego blocks, one after another. But the image was misleading. Because unlike a Lego set, atoms couldn't be snapped together in any arrangement you liked. An inserted atom was subject to powerful local forces-magnetic and chemical-with frequently undesirable results. The atom might be kicked out of its position. It might remain, but at an awkward angle. It might even fold the entire molecule up in knots.

As a result, molecular manufacturing was an exercise in the art of the possible, of substituting atoms and groups of atoms to make equivalent structures that would work in the desired way. In the face of all this difficulty, it was impossible to ignore the fact that there already existed proven molecular factories capable of turning out large numbers of molecules: they were called cells.

"Unfortunately, cellular manufacturing can take us only so far," Ricky said. "We harvest the substrate molecules-the raw materials-and then we build on them with nanoengineering procedures. So we do a little of both."

I pointed down at the tanks. "What cells are you growing?"

"Theta-d 5972," he said.

"Which is?"

"A strain of E. coli."

E. coli was a common bacterium, found pretty much everywhere in the natural environment, even in the human intestine. I said, "Did anyone think it might not be a good idea to use cells that can live inside human beings?"

"Not really," he said. "Frankly that wasn't a consideration. We just wanted a well-studied cell that was fully documented in the literature. We chose an industry standard."

"Uh-huh ..."

"Anyway," Ricky continued, "I don't think it's a problem, Jack. It won't thrive in the human gut. Theta-d is optimized for a variety of nutrient sources-to make it cheap to grow in the laboratory. In fact, I think it can even grow on garbage."

"So that's how you get your molecules. Bacteria make them for you."

"Yes," he said, "that's how we get the primary molecules. We harvest twenty-seven primary molecules. They fit together in relatively high-temperature settings where the atoms are more active and mix quickly."

"That's why it's hot in here?"

"Yes. Reaction efficiency has a maxima at one hundred forty-seven degrees Fahrenheit, so we work there. That's where we get the fastest combination rate. But these molecules will combine at much lower temperatures. Even around thirty-five, forty degrees Fahrenheit, you'll get a certain amount of molecular combination."

"And you don't need other conditions," I said. "Vacuum? Pressure? High magnetic fields?" Ricky shook his head. "No, Jack. We maintain those conditions to speed up assembly, but it's not strictly necessary. The design is really elegant. The component molecules go together quite easily."

"And these component molecules combine to form your final assembler?"

"Which then assembles the molecules we want. Yes."

It was a clever solution, creating his assemblers with bacteria. But Ricky was telling me the components assembled themselves almost automatically, with nothing required but high temperature. What, then, was this complex glass building used for? "Efficiency, and process separation," Ricky said. "We can build as many as nine assemblers simultaneously, in the different arms."

"And where do the assemblers make the final molecules?"

"In this same structure. But first, we reapply them."

I shook my head. I wasn't familiar with the term. "Reapply?"

"It's a little refinement we developed here. We're patenting it. You see, our system worked perfectly right from the start-but our yields were extremely low. We were harvesting half a gram of finished molecules an hour. At that rate, it would take several days to make a single camera. We couldn't figure out what the problem was. The late assembly in the arms is done in gas phase. It turned out that the molecular assemblers were heavy, and tended to sink to the bottom. The bacteria settled on a layer above them, releasing component molecules that were lighter still, and floated higher. So the assemblers were making very little contact with the molecules they were meant to assemble. We tried mixing technologies but they didn't help."

"So you did what?"

"We modified the assembler design to provide a lipotrophic base that would attach to the surface of the bacteria. That brought the assemblers into better contact with the component molecules, and immediately our yields jumped five orders of magnitude."

"And now your assemblers sit on the bacteria?"

"Correct. They attach to the outer cell membrane."

At a nearby workstation, Ricky punched up the assembler design on the flat panel display. The assembler looked like a sort of pinwheel, a series of spiral arms going off in different directions, and a dense knot of atoms in the center. "It's fractal, as I said," he said. "So it looks sort of the same at smaller orders of magnitude." He laughed. "Like the old joke, turtles all the way down." He pressed more keys. "Anyway, here's the attached configuration." The screen now showed the assembler adhering to a much larger pill-shaped object, like a pinwheel attached to a submarine. "That's the Theta-d bacterium," Ricky said. "With the assembler on it."

As I watched, several more pinwheels attached themselves. "And these assemblers make the actual camera units?"

"Correct." He typed again. I saw a new image. "This is our target micromachine, the final camera. You've seen the bloodstream version. This is the Pentagon version, quite a bit larger and designed to be airborne. What you're looking at is a molecular helicopter."

"Where's the propeller?" I said.

"Hasn't got one. The machine uses those little round protrusions you see there, stuck in at angles. Those're motors. The machines actually maneuver by climbing the viscosity of the air."

"Climbing the what?"

"Viscosity. Of the air." He smiled. "Micromachine level, remember? It's a whole new world, Jack."

However innovative the design, Ricky was still bound by the Pentagon's engineering specs for the product, and the product wasn't performing. Yes, they had built a camera that couldn't be shot down, and it transmitted images very well. Ricky explained it worked perfectly during tests indoors. But outside, even a modest breeze tended to blow it away like the cloud of dust it was. The engineering team at Xymos was attempting to modify the units to increase mobility, but so far without success. Meanwhile the Department of Defense decided the design constraints were unbeatable, and had backed away from the whole nano concept; the Xymos contract had been canceled; DOD was going to pull funding in another six weeks. I said, "That's why Julia was so desperate for venture capital, these last few weeks?"

"Right," Ricky said. "Frankly, this whole company could go belly up before Christmas."

"Unless you fix the units, so they can work in wind."

"Right, right."

I said, "Ricky, I'm a programmer. I can't help you with your agent mobility problems. That's an issue of molecular design. It's engineering. It's not my area."

"Um, I know that." He paused, frowned. "But actually, we think the program code may be involved in the solution."

"The code? Involved in the solution to what?"

"Jack, I have to be frank with you. We've made a mistake," he said. "But it's not our fault. I swear to you. It wasn't us. It was the contractors." He started down the stairs. "Come on, I'll show you."

Walking briskly, he led me to the far side of the facility, where I saw an open yellow elevator cage mounted on the wall. It was a small elevator, and I was uncomfortable because it was open; I averted my eyes. Ricky said, "Don't like heights?"

"Can't stand them."

"Well, it's better than walking." He pointed off to one side, where an iron ladder ran up the wall to the ceiling. "When the elevator goes out, we have to climb up that." I shuddered. "Not me."

We rode the elevator all the way up to the ceiling, three stories above the ground. Hanging beneath the ceiling was a tangle of ducts and conduits, and a network of mesh walkways to enable workers to service them. I hated the mesh, because I could see through it to the floor far below. I tried not to look down. We had to duck repeatedly beneath the low-hanging pipes. Ricky shouted over the roar of the equipment.

"Everything's up here!" he yelled, pointing in various directions. "Air handlers over there! Water tank for the fire sprinkler system there! Electrical junction boxes there! This is really the center of everything!" Ricky continued down the walkway, finally stopping beside a big air vent, about three feet in diameter, that went into the outer wall.

"This is vent three," he said, leaning close to my ear. "It's one of four main vents that exhausts air to the outside. Now, you see those slots along the vent, and the square boxes that sit in the slots? Those are filter packs. We have microfilters arranged in successive layers, to prevent any external contamination from the facility."

"I see them ..."

"You see them now," Ricky said. "Unfortunately, the contractor forgot to install the filters in this particular vent. In fact, they didn't even cut the slots, so the building inspectors never realized anything was missing. They signed off on the building; we started working here. And we vented unfiltered air to the outside environment."

"For how long?"

Ricky bit his lip. "Three weeks."

"And you were at full production?"

He nodded. "We figure we vented approximately twenty-five kilos of contaminants."

"And what were the contaminants?"

"A little of everything. We're not sure of exactly what."

"So you vented E. coli, assemblers, finished molecules, everything?"

"Correct. But we don't know what proportions."

"Do the proportions matter?"

"They might. Yes."

Ricky was increasingly edgy as he told me all this, biting his lip, scratching his head, avoiding my eyes. I didn't get it. In the annals of industrial pollution, fifty pounds of contamination was trivial. Fifty pounds of material would fit comfortably in a gym bag. Unless it was highly toxic or radioactive-and it wasn't-such a small quantity simply didn't matter. I said, "Ricky, so what? Those particles were scattered by the wind across hundreds of miles of desert. They'll decay from sunlight and cosmic radiation. They'll break up, decompose. In a few hours or days, they're gone. Right?"

Ricky shrugged. "Actually, Jack, that's not what-"

It was at that moment that the alarm went off.

It was a quiet alarm, just a soft, insistent pinging, but it made Ricky jump. He ran down the walkway, feet clanging on the metal, toward a computer workstation mounted on the wall. There was a status window in the corner of the monitor. It was flashing red: PV-90 ENTRY.

I said, "What does that mean?"

"Something set off the perimeter alarms." He unclipped his radio and said, "Vince, lock us down."

The radio crackled. "We're locked down, Ricky."

"Raise positive pressure."

"It's up five pounds above baseline. You want more?"

"No. Leave it there. Do we have visualization?"

"Not yet."

"Shit." Ricky stuck the radio back on his belt, began typing quickly. The workstation screen divided into a half-dozen small images from security cameras mounted all around the facility. Some showed the surrounding desert from high views, looking down from rooftops. Others were ground views. The cameras panned slowly.

I saw nothing. Just desert scrub and occasional clumps of cactus.

"False alarm?" I said.

Ricky shook his head. "I wish."

I said, "I don't see anything."

"It'll take a minute to find it."

"Find what?"

"That."

He pointed to the monitor, and bit his lip.

I saw what appeared to be a small, swirling cloud of dark particles. It looked like a dust devil, one of those tiny tornado-like clusters that moved over the ground, spun by convection currents rising from the hot desert floor. Except that this cloud was black, and it had some definition-it seemed to be pinched in the middle, making it look a bit like an old-fashioned Coke bottle. But it didn't hold that shape consistently. The appearance kept shifting, transforming. "Ricky," I said. "What are we looking at?"

"I was hoping you'd tell me."

"It looks like an agent swarm. Is that your camera swarm?"

"No. It's something else."

"How do you know?"

"Because we can't control it. It doesn't respond to our radio signals."

"You've tried?"

"Yes. We've tried to make contact with it for almost two weeks," he said. "It's generating an electrical field that we can measure, but for some reason we can't interact with it."

"So you have a runaway swarm."

"Yes."

"Acting autonomously."

"Yes."

"And this has been going on for ..."

"Days. About ten days."

"Ten days?" I frowned. "How is that possible, Ricky? The swarm's a collection of micro-robotic machines. Why haven't they decayed, or run out of power? And why exactly can't you control them? Because if they have the ability to swarm, then there's some electrically mediated interaction among them. So you should be able to take control of the swarm-or at least disrupt it."

"All true," Ricky said. "Except we can't. And we've tried everything we can think of." He was focused on the screen, watching intently. "That cloud is independent of us. Period."

"And so you brought me out here ..."

"To help us get the fucking thing back," Ricky said.

DAY 6

9:32 A.M.

It was, I thought, a problem no one had ever imagined before. In all the years that I had been programming agents, the focus had been on getting them to interact in a way that produced useful results. It never occurred to us that there might be a larger control issue, or a question of independence. Because it simply couldn't happen. Individual agents were too small to be self-powered; they had to get their energy from some external source, such as a supplied electrical or microwave field. All you had to do was turn off the field, and the agents died. The swarm was no more difficult to control than a household appliance, like a kitchen blender. Flip the power off and it went dead.