The phrase that signaled the revolution was "carbon electronics." When I first heard the term, I imagined a small black chip of carbon with integrated circuits on it. So what? But my opinion of the technology completely changed when an excited colleague came into my office one day. (This was back when I was a silicon IC designer, and a good one too.) He was holding a swatch of a new, buckytube-fiber-based material that had set off a fashion rage. Fabric chemists had found out how to mass-produce carbon nanotubes and weave them into the remarkably tough yet silklike material that everyone thinks is so cool today.

Naturally, we were a little worried about our future, but one fallout of the revolution was a vast increase in the demand for IC designers. This computing fabric, running "CMOS"-the Cell Matrix Operating System-was ideally suited for emulating circuits rather than for running step-by-step algorithms. So the old idea of software as an algorithm went out the window, and circuit designers became the new hackers.

But we turned out to be only the lowest level in the software hierarchy. Programmers had to master a new paradigm, and some just didn't get it. The CMOS paradigm is based on self-replication and self-reorganization rather than a deterministic cause-and-effect model of data processing.

Everyone was predicting that the big semiconductor fabs would implode, but those guys did a real fast switcheroo, converting overnight into compound-semiconductor laser I/O lighting fabs. Gallium arsenide, the material of the future, had finally arrived. Of course, there is still a fairly big demand for silicon microelectromechanical systems (MEMS) and sensors of all types. These mainly use silicon, keeping a small portion of the old silicon fabs in business.

The lighting revolution replaced conventional light sources with room-filling laser fields wired into the optical Internet. And because the atomic-scale computing fabric uses so few electrons, the light field is also the power source for today's "computers." The whole package is microscopically integrated into room decor and clothing and thus has receded from consciousness. Most people don't even know how it works or how we got here. So I have been working on an historical account of the revolution.

I guess we can blame all of this on Richard Smalley, one of the discoverers of carbon buckyballs. Buckyballs were just the start of what Smalley began to call "graphitic architectures." Carbon most readily forms into two-dimensional hexagonal sheets, which are very weak. But at high temperatures, the sheets fold and twist around themselves to form exotic three-dimensional structures. The really technologically useful structure turned out to be the carbon nanotube.

Most people don't realize that all the pieces of the puzzle were in place as early as 2002. Research groups had shown how to build linear circuits with buckytubes and enhance them by inserting buckyballs right into the tubes. Filled with electrically or optically active materials, the buckyballs introduced such critical functionality as optically active I/O points along the tubes. And a group at Belgium's Starlab had come up with the basic design for fabric-based circuitry.

No one had heard of Cell Matrix Corp. at the time, but that company had already embarked on the software concepts that would make atomic-scale computing fabric a practical proposition, at least from the perspective of designing useful computational blocks. Cell Matrix introduced a critical concept: self-replicating configurable circuitry that simply flows around defects or nonfunctioning blocks of gates.

Of course, a major hurdle stood in the way of practical applications of those discoveries: No way to mass-produce buckytubes with a specific structure had been discovered. But just as silicon technology co-opted techniques from the printing industry to create the VLSI revolution, fabric computing leveraged advanced technology from the textile industry.

The critical component was micromachined nanoextrusion chips that combined MEMS technology with genetically engineered microassemblers. Physically, the chips looked like scaled-down versions of the extrusion equipment used to make traditional synthetic fibers, such as nylon.

Buckytubes were assembled and extruded from the back of the chip, where microspinners would twist them into larger fibers that could be handled by conventional textile machinery. The result was fabric of unprecedented strength and lightness-qualities the military certainly appreciated right out of the gate.

The fabric proved to be fantastically versatile; and fashion designers loved it. Oddly, fashion became a pivotal aspect of the subsequent electronics revolution. A group of chemical and electronics giants got together and formed a consortium-Buckytube LLC-to adapt the process to electrically active buckytubes.

The really weird thing about this approach to circuit design, at least for a traditional IC designer like myself, is the absence of any attempt at controlling yield or registration. Contacts are established along the threads with optically active buckyballs, and the whole thing works via statistics. If there are enough of them, sufficient interconnect will occur to support the CMOS software.

Because this is atomic-scale fabrication, even with only 25 percent of the tubes actively connected and working there is plenty of hardware available for the software, which just flows around the nonoperational blocks of fabric. So now electronic-textile fabs are cranking this stuff out at the rate of miles per second, and the logic gates are free. The only added value is either software or fashion design.

The electronics industry has gone down the same road as the auto industry before it: They tweak the technology a bit each year, but products are sold on the basis of style. This year, synthetic-cashmere computing sweaters seem to be the hot item.