Maria was beginning to wish she'd kept her mouth shut. It had taken her almost an hour on the phone to persuade Durham that it was worth trying to give Lambert a proper astronomical context, and a geological history that stretched back to the birth of its sun.

"If we present this world as a fait accompli, and say: "Look, it can exist in the Autoverse" . . . the obvious response to that will be: "Yes, it can exist -- if you put it there by hand -- but that doesn't mean it's ever likely to have formed." If we can demonstrate a range of starting conditions that lead to planetary systems with suitable worlds, that will be one less element of uncertainty to be used against us."

Durham had eventually agreed, so she'd taken an off-the-shelf planetary-system modeling program -- irreverently titled The Laplacian Casino -- and adapted it to Autoverse chemistry and physics; not the deep physics of the Autoverse cellular automaton, but the macroscopic consequences of those rules. Mostly, that came down to specifying the properties of various Autoverse molecules: bond energies, melting and boiling points versus pressure, and so on. Aqua was not just water by another name, yellow atoms were not identical to nitrogen -- and although some chemical reactions could be translated as if there was a one-to-one correspondence, in the giant fractionating still of a protostellar nebula subtle differences in relative densities and volatilities could have profound effects on the final composition of each of the planets.

There were also some fundamental differences. Since the Autoverse had no nuclear forces, the sun would be heated solely by gravitational energy -- the velocity its molecules acquired as the diffuse primordial gas cloud fell in on itself. In the real universe, stars unable to ignite fusion reactions ended up as cold, short-lived brown dwarfs -- but under Autoverse physics, gravitational heating could power a large enough star for billions of years. (Units of space and time were not strictly translatable -- but everybody but the purists did it. If a red atom's width was taken to be that of hydrogen, and one grid-spacing per clock-tick was taken as the speed of light, a more or less sensible correspondence emerged.) Similarly, although Planet Lambert would lack internal heating from radioisotope decay, its own gravitational heat of formation would be great enough to drive tectonic activity for almost as long as the sun shone.

Without nuclear fusion to synthesize the elements, their origin remained a mystery, and a convenient gas cloud with traces of all thirty-two -- and the right mass and rotational velocity -- had to be taken for granted. Maria would have liked to have explored the cloud's possible origins, but she knew the project would never be finished if she kept lobbying Durham to expand the terms of reference. The point was to explore the potential diversity of Autoverse life, not to invent an entire cosmology.

Gravity in the Autoverse came as close as real-world gravity to the classical, Newtonian inverse-square law for the range of conditions that mattered, so all the usual real-world orbital dynamics applied. At extreme densities, the cellular automaton's discrete nature would cause it to deviate wildly from Newton -- and Einstein, and Chu -- but Maria had no intention of peppering her universe with black holes, or other exotica.

In fact, gravity had been seen as an irrelevant side effect of Lambert's original choice of automaton rules -- since running an Autoverse large enough for it to make the slightest difference was blatantly impossible -- and several people had tried to remove the redundancy, while leaving everything else intact. Nobody had succeeded, though; their "rationalized" versions had always failed to generate anything remotely like the rich chemistry of the original. A Peruvian mathematician, Ricardo Salazar, had eventually proved that they shouldn't have bothered: the Autoverse rules were poised on the border between two radically different levels of algorithmic complexity, and any tinkering in the hope of improved efficiency was necessarily self-defeating. The presence or absence of gravity, in itself, had no bearing on Autoverse chemistry -- but the roots of both phenomena in the simple automaton rules seemed to be inextricably entwined.

Maria was aiming for a star with four planets. Three small worlds, one giant. The seed-world, Lambert, second from the sun -- with a decent-sized moon if possible. Whether or not tidal pools had been a driving force in real-world evolution, life's bridge from sea to land (and even though the sun itself would cause small tides, regardless), it couldn't hurt to make Lambert as generally Earth-like as possible, since Earth was still the only example to turn to for inspiration. With so much about terrestrial evolution still in dispute, the safest policy was to cover every factor which might have been significant. The gravitational effects of the other planets would ensure a reasonably complex set of Milankovitch cycles: minor orbital changes and axis wobbles, providing long-term climate variations, ice ages and interglacials. A belt of comets and other debris would complete the picture; not merely supplying an atmosphere, early on, but also offering the chance of occasional mass-extinctions for billions of years to come.

The trick was to ensure that all of these supposedly evolution-enhancing features coincided with a version of Lambert which could support the seed organism in the first place. Maria had half a dozen possible modifications to A. lamberti in mind, to render it self-sufficient, but she was waiting to see what kind of environments were available before making a final decision.

That still left unanswered the question of whether the seed organism -- or life of any kind -- could have arisen on Lambert, rather than being placed there by human hands. Max Lambert's original reason for designing the Autoverse had been the hope of observing self-replicating molecular systems -- primitive life -- arising from simple chemical mixtures. The Autoverse was meant to provide a compromise between real-world chemistry -- difficult and expensive to manipulate and monitor in test-tube experiments, and hideously slow to compute in faithful simulations -- and the tantalizing abstractions of the earliest "artificial life": computer viruses, genetic algorithms, self-replicating machines embedded in simple cellular automaton worlds; all trivially easy to compute, but unable to throw much light on the genesis of real-world molecular biology.

Lambert had spent a decade trying to find conditions which would lead to the spontaneous appearance of Autoverse life, without success. He'd constructed A. lamberti -- a twelve-year project -- to reassure himself that his goal wasn't absurd; to demonstrate that a living organism could at least function in the Autoverse, however it had come to be there. A. lamberti had permanently side-tracked him; he'd never returned to his original research.

Maria had daydreamed about embarking on her own attempt at abiogenesis, but she'd never done anything about it. That kind of work was open-ended; in comparison, any problems with mutation in A. lamberti seemed utterly tractable and well-defined. And although, in a sense, it went to the heart of what Durham was trying to prove, she was glad he'd chosen to compromise; if he'd insisted on starting his "thought experiment" with a totally sterile world, the uncertainties in the transition from inanimate matter to the simplest Autoverse life would have overwhelmed every other aspect of the project.

She scrapped the desert Planet Lambert and returned to the primordial gas cloud. She popped up a gadget full of slider controls and adjusted the cloud's composition, taking back half the increases she'd made in the proportions of blue and yellow. Planetology by trial and error. The starting conditions for real-world systems with Earth-like planets had been mapped out long ago, but nobody had ever done the equivalent for the Autoverse. Nobody had ever had a reason.