And they certainly didn't look human. The first was only a set of rollers that rotated a synthetic crystal a millimeter at a time, then lowered onto it a hemispherical cover that was filled with golden contacts. The central computer tested each circuit within the crystal through those contacts, checking continuity, resistance, power input versus output, and a host of other electronic characteristics. After fifteen minutes, the rollers tilted the crystal out onto the padded belt that carried it to the next robot—or into the garbage can, if it had failed any of its tests.

The next robot was very similar, except that it connected microscopic filaments to each contact point.

Then came a robot that looked like an octopus, with fifteen arms sprouting from a central globe that held its computer. Its job was assembling fifteen crystals into one globular cluster joined by filaments, then immersing it in a chemical bath. After two hours, enough silicon had adhered to the filaments so that the robot could withdraw its arms and start assembling another cluster, while the first rested in its bath for a week, slowly growing together into a single giant crystal.

Meanwhile, another robot—a single bench that grabbed, folded, and held metal and plastic while a steel arm welded joints—was casting and assembling the housings for the completed machines. Then came the assembly line—a final robot which took the finished giant crystals out of their baths and fastened them inside the housings, then connected the contacts to the terminals for the mechanical attachments that actually did the work.

All of it faster than he could do. All of it better than he could do. And most of it much, much smaller than he could see.

Dar surveyed the area, feeling totally useless.

"You really should take a finished robot out for testing, Dar."

"Yeah, I know—but I've done two already today, and there's plenty of time to check the other three."

"Still, it must be done, or you will have a dozen untested robots at week's end."

"I know, I know—but there doesn't seem to be much point to it. You know they'll work perfectly."

"I do not, Dar. True, if the individual crystal circuits are sound, the finished computer will be fine…"

"Of course, because the central computer tests each stage of the work as it's being done." There were contacts embedded in the "holder" on each bench, and in each arm, allowing testing while production was in process.

"But you may find a mechanical flaw, Dar."

"Yeah, sure. Last week I found a pinhole in a suction funnel, and the week before that, there was a hum in a lifting fan. Never in the computers, of course."

He watched the process, shaking his head with dissatisfaction.

"What displeases you, Dar?"

"Huh? Oh. I keep forgetting you're programmed for gestures, too. Nothing, Fess—or nothing that should, anyway. We make dam good household robots—but blast it, all we make is household robots!"

"True, Dar, but, as you say, you make them very well—and you always manage to offer an automaton that will do more than your competition's product."

"Well, that's true. We started out with a little canister that could dust, and speak a few simple responses such as 'Yes, ma'am,' 'No, sir,' 'Good morning,' and 'Please move…' "

"Which every other company's could also do, of course."

"Yeah, but we figured out a way for ours to scrub floors and polish furniture, too."

"Then you added the abilities to pick up, clear a table, load a dishwasher, one by one—and always a year or two before the other companies."

"Yeah, because they wait till we come out with it, then buy one of ours and copy the new feature—but that always takes six months at least, while we make another hundred thousand in sales. Which reminds me, we'd better finish debugging that breakfast-delivery program, or we'll lose our edge."

"I would not be terribly concerned, Dar. You can still add many features before you will have perfected the ideal household helper."

"What do you mean?" Dar frowned.

"Why, your robots cannot yet replace a closure seam in a garment, or light a fire—or fight one, for that matter."

"Oh. Right. And they don't do windows." But Dar was gazing off into space. "Let's see, now…"

"You will find ways to accomplish them all," Fess assured him.

"True, true. And there are other improvements I'm itchy to get to."

"Such as?"

"Well, they could be a lot smaller, for one thing.""

"I do not know, Dar—there is a lower limit on size, for accomplishing mechanical tasks."

"Oh, not the robots themselves, Fess—we've got those as low as they can go, and still be practical. Anything less, and the householder will be tripping over them every time he turns around. No, I meant the computers. They're still bigger than my fist."

"I fail to see how they could be smaller, Dar. You are already working with the smallest crystal lattice that can carry an adequate number of differences in electrical potentials."

"Are we?" Dar's tone sharpened. "A crystal has a regular shape because its molecule has, Fess. Why can't the differences in electrical potential that made a crystal lattice function as a circuit, be made to operate with only a single molecule?"

Fess was slow in answering, which meant his computer, which worked in nanoseconds, had analyzed the problem thoroughly, and made a preliminary try at resolving it. "In theory, there would be no reason for it not to, Dar—but the complexity of the circuitry would be limited by the number of electrons available."

"All right, so it might take a dozen molecules, or maybe even a single giant molecule—but you're still talking about something microscopic, or just barely visible."

"Are you seriously intending to research the possibility?"

"Not without telling Lona—and I don't know enough physics yet, to know if it's worth investigating. I mean, okay, it might turn out to be possible, but not marketable, especially if it wound up using giant molecules. After all, who wants a computer made of U-235 in his living room?"

"The mass would not be critical, Dar."

"No, but the customer would. Call it atavism, call it superstition, but the stuff has a bad reputation."

"Even if it were, why would you wish to do it? The current generation of computers is certainly small enough for all practical purposes."

"Not all—I can think of a few applications where microbrains would come in handy. Especially in the line I want to get into."

"Which is?"

"Industrial robots." Dar rubbed his knuckles against his palm. "We need to branch out, Fess. There's just so far we can go with household robots, and right now, the real money is in industry. If we could offer smaller, more compact computers, that would cut down on size and give flexibility a big boost. Factory managers are always complaining about having to replace all their robots with new models, every time they retool. If we could figure out how to grow a single molecule circuit, we could sell them new brains for more generalized robots."

"An excellent idea," Fess said slowly. "You must tell Lona."

Dar felt a surge of irritation, but reminded himself that Fess was, after all, Lona's robot. "No. I just don't know enough yet."

Which made him feel even more useless.

He turned away, closing the factory hatch behind him, and went into his den. "I'm going to brood—uh, study, for a while, Fess."

"I shall not disturb you, Dar." But he would wait for Dar's call. That went without saying.

The lights came on, and Dar sat down at his computer with a sigh of relief. Here, at least, he had something to do, and the illusion that it might actually be of some use, even though that was highly unlikely. Of course, he was only experimenting with computer simulations of radio sound waves and FTL drives, not with the real thing—but he might hit on a workable idea.