The red-tagged molecules wandered the cell at random, part-digested mixed with raw indiscriminately. Neat process diagrams of metabolism -- the real-world Embden-Meyerhof pathway, or the Autoverse's Lambert pathway -- always gave the impression of some orderly molecular conveyor belt, but the truth was, life in either system was powered by nothing at the deepest level but a sequence of chance collisions.

A few red tags turned orange. Stage two: an enzyme tightening the molecule's hexagonal ring into a pentagon, transforming the spare vertex into a protruding cluster, more exposed and reactive than before.

Still nothing new. And still no hint of violet.

Nothing further seemed to happen for so long that Maria glanced at her watch and said "Globe," to see if some major population center had just come on-line for the day -- but the authentic Earth-from-space view showed dawn well into the Pacific. California would have been busy since before she'd arrived home.

A few orange tags turned yellow. Stage three of the Lambert pathway, like stage one, consisted of bonding an energy-rich group of atoms to the sugar. With nutrose, there was a payoff for this, eventually, with twice as many of the molecules which supplied the energy ending up "recharged" as had been "drained." Stage four, though -- the cleaving of the ring into two smaller fragments -- was the point where mutose gummed up the works irretrievably . . .

Except that one yellow speck had just split into two, before her eyes . . . and both new tags were colored violet.

Maria, startled, lost track of the evidence. Then she caught sight of the same thing happening again. And then a third time.

It took her a minute to think it through, and understand what this meant. The bacterium wasn't reversing the change she'd made to the sugar, converting mutose back into nutrose -- or doing the same to some part-digested metabolite. Instead, it must have modified the enzyme which broke the ring, coming up with a version which worked directly on the metabolite of mutose.

Maria froze the action, zoomed in, and watched a molecular-scale replay. The enzyme in question was constructed of thousands of atoms; it was impossible to spot the difference at a glance -- but there was no doubt about what it was doing. The two-atom blue-red spike she'd repositioned on the sugar was never shifted back into its "proper" place; instead, the enzyme now accommodated the altered geometry perfectly.

She summoned up old and new versions of the enzyme, highlighted the regions where the tertiary structure was different, and probed them with her fingertips -- confirming, palpably, that the cavity in the giant molecule where the reaction took place had changed shape.

And once the ring was cleaved? The fragments were the same, whether the original sugar had been nutrose or mutose. The rest of the Lambert pathway went on as if nothing had changed.

Maria was elated, and a little dazed. People had been trying to achieve a spontaneous adaptation like this for sixteen years. She didn't even know why she'd finally succeeded; for five years she'd been tinkering with the bacterium's error correction mechanisms, trying to force A. lamberti to mutate, not more rapidly, but more randomly. Every time, she'd ended up with a strain which -- like Lambert's original, like those of other workers -- suffered the same handful of predictable, useless mutations again and again . . . almost as if something deep in the clockwork of the Autoverse itself ruled out the exuberant diversity which came so effortlessly to real-world biology. Calvin and others had suggested that, because Autoverse physics omitted the deep indeterminacy of real-world quantum mechanics -- because it lacked this vital inflow of "true unpredictability" -- the same richness of phenomena could never be expected, at any level.

But that had always been absurd -- and now she'd proved it was absurd.

For a moment she thought of phoning Aden, or Francesca -- but Aden wouldn't understand enough to do more than nod politely, and her mother didn't deserve to be woken at this hour.

She got up and paced the tiny bedroom for a while, too excited to remain still. She'd upload a letter to Autoverse Review (total subscription, seventy-three), with the genome of the strain she'd started out with appended as a footnote, so everyone else could try the experiment . . .

She sat down and began composing the letter -- popping up a word processor in the foreground of the workspace -- then decided that was premature; there was still a lot more to be done to form the basis of even a brief report.

She cloned a small colony of the mutose-eating strain, and watched it grow steadily in a culture of pure mutose. No surprise, but it was still worth doing.

Then she did the same, with pure nutrose, and the colony, of course, died out at once. The original ring-cleaving enzyme had been lost; the original roles of nutrose and mutose as food and poison had been swapped.

Maria pondered this. A. lamberti had adapted -- but not in the way she'd expected. Why hadn't it found a means of consuming both sugars, instead of exchanging one kind of exclusive reliance for another? It would have been a far better strategy. It was what a real-world bacterium would have done.

She brooded over the question for a while -- then started laughing. Sixteen years, people had been hunting for a single, convincing example of natural selection in the Autoverse -- and here she was worrying that it wasn't the best of all possible adaptations. Evolution was a random walk across a minefield, not a preordained trajectory, onward and upward toward "perfection." A. lamberti had stumbled on a successful way to turn poison into food. It was tough luck if the corollary was: vice versa.

Maria ran a dozen more experiments. She lost all track of time; when dawn came, the software brightened the images in front of her, keeping the daylight from washing them out. It was only when her concentration faltered, and she looked around the room, that she realized how late it was.

She started again on the letter. After three drafts of the first paragraph -- all eliciting the same response from Camel's Eye: You'll hate this when you reread it later. Trust me. -- she finally admitted to herself that she was wasted. She shut down everything and crawled into bed.

She lay there awhile in a stupor, burying her face in the pillow, waiting for the ghost images of Petri dishes and enzymes to fade. Five years ago, she could have worked all night, and suffered nothing worse than a fit of yawning in the middle of the afternoon. Now, she felt like she'd been hit by a train -- and she knew she'd be a wreck for days. Thirty-one is old, old, old.

Her head throbbed, her whole body ached. She didn't care. All the time and money she'd squandered on the Autoverse was worth it, now. Every moment she'd spent there had been vindicated.

Yeah? She rolled onto her back and opened her eyes. What, exactly, had changed? It was still nothing but a self-indulgent hobby, an elaborate computer game. She'd be famous with seventy-two other anal-retentive Autoverse freaks. How many bills would that pay? How many typhoons would it neutralize?

She wrapped her head in the pillow, feeling crippled, stupid, hopeless -- and defiantly happy -- until her limbs went numb, her mouth went dry, and the room seemed to rock her to sleep.