Cheaper than printing it out: buy the paperback book.

Weiser's buildings are a coevolutionary ecology of machines. Each device is an organism that reacts to stimulus and communicates with the others. Cooperation is rewarded. Alone, most of the electronic bits are wimpy and would die of nonuse. Together, they form a community that is attentive and robust. What each microbit lacks in depth is made up by the communal net which casts its collective influence wide over a building, outreaching even a human.

Not only would rooms and halls have embedded intelligence and ecological fluidity but entire streets, malls, and towns. Weiser uses the example of words. Writing, he says, is a technology that is ubiquitously embedded into our environment. Writing is everywhere, urban and suburban, passively waiting to be read. Now imagine, Weiser suggests, computation and connection embedded into the built environment to the same degree. Street signs would communicate to car navigation systems or a map in your hands (when street names change, all maps change too). Streetlights in a parking lot would flick on ahead of you in anticipation of your walk. Point to a billboard properly, and it would send you more information on its advertised product and let its sponsor know what part of the street most of the queries came from. The environment becomes animated, responsive, and adaptable. It responds not only to you but to all the other agents plugged in at the time.

One definition of a coevolutionary ecology is a collection of organisms that serve as their own environment. The flamboyant world of orchid flowers, ant colonies, and seaweed beds overflows with richness and mystery because the movie that each creature stars in features walk-ons and extras who are simultaneously acting as stars in their own movie filmed on the same lot. Every borrowed set is alive and liquid as the star is. Thus, the fate of a mayfly is primarily determined by the histrionics of neighboring frog, trout, alder, water spider, and the rest of stream life, each playing the environment for the other. Machines too will play on a coevolutionary stage.

The refrigerator you can purchase today is an arrogant snob. When you bring it home it assumes that it alone is the only appliance in the house. It has nothing to learn from other machines in the building, and nothing it will tell them. A wall clock will tell you the time of day but not its manufactured brethren. Each utensil haughtily serves only its buyer without regard to how much better it could serve in cooperation with the other items around it.

An ecology of machines, on the other hand, enhances the limited skills of dumb machines. The chips imbedded in book and chair have only the smartness of ants. They're no supercomputer; they could be manufactured now. But by the alien power of distributed being, sufficient numbers of antlike agents can be lifted into a type of colony intelligence by connecting them in bulk. More is different.

Collaborative efficiency, however, has a price. An ecological intelligence will penalize anyone new to the room, just as a tundra ecology will penalize anyone new to the arctic. Ecologies demand local knowledge. The only folks who know where the mushrooms bloom in the woods are native sons. To track wallabies through the Australian outback you want a local bush ranger as a guide.

Where there is an ecosystem, there are local experts. An outsider can muddle through an unfamiliar wilderness at some level, but to thrive or to survive a crisis, he'll require local expertise. Gardeners regularly surprise academic experts by growing things they aren't supposed to be able to grow because, as local experts, they tune into the neighborhood soil and climate.

The work of managing a natural environment is inescapably a work of local knowledge. A roomful of mechanical organisms improvising with each other demands a similar local knowledge. The one advantage snooty old Refrigerator had was that he ignored everyone equally, owner and visitor alike. In a room enlivened by a colony intelligence, visitors are at a disadvantage. Every room will be different; indeed, every telephone will be different. Because the new phones will merely be one node of a far larger organism -- linking furnace, cars, TVs, computers, chairs, whole buildings -- whose own behavior will hinge on the holistic sum of everything else going on in the room. The behavior of each will particularly depend on how its most frequent user employs it. To visitors, the indefinite beast of a room will seem to be out of control.

Adaptable technology means that technology that will adapt locally. The logic of the network induces regionalism and localism. Or to put it another way, global behavior entails regional variety. We see evidence of the shift already. Try using someone else's "smart" phone. It is already either too smart, or not smart enough. Do you dial "9" to get out? Can you punch any button for a line? How do you (gulp!) transfer a call? Only the owner knows for sure. The local knowledge needed to fully operate a VCR is legendary. Just because you can preprogram yours to record The Prisoner reruns doesn't in any way mean you can handle your friend's.

Rooms and buildings will vary in their electronic ecology, as will appliances within a room, since they all will be aggregations of smaller distributed parts. No one will know the idiosyncrasies of my office's technology as well as I will. Nor will I be able to work another's technology as easily as my own. As computers become assistants, toasters become pets.

When the designers get it right, the coffee machine that an impatient visitor tries to use will default to "novice mode" when it senses desperate attempts to make it work. Mr. Coffee will cop to the situation by engaging only the five basic universal appliance functions that every school child will know.

But I find the emerging ecology in its earliest stages already daunting to strangers. Since computers are the locus where all these devices hail from and head toward, we can see in them now the alienness of unfamiliar complex machines. It doesn't matter how acquainted you are with a particular brand of computer. When you need to borrow someone else's, it feels like you're using their toothbrush. The instant you turn a friend's computer on, it's there: that strange arrangement of familiar parts (why do they do it like that?), the whole disorienting logic of a place you thought you knew. You kind of recognize it. There's an order here. Then, a moment of terror. You are...peering into someone else's mind!

The penetration goes both ways. So personal, so subtle, so minute is everyone's parochial intelligence of their own computer's ecology, that any disturbance is alarming. A pebble dislodged, a blade of grass bent, a file moved. "Someone has been in my compu-room! I know it!"

There will be nice-dog rooms and bad-dog rooms. Bad-dog rooms will bite intruders. Nice-dog rooms will herd visitors to someplace safe, away from places where real harm can be done. The nice-dog room may entertain guests. People will acquire reputations on how well-trained their computers are and how well-groomed their computational ecology is. And others will gain notoriety for how fiercely wild their machinery is. There are sure to be neglected areas in large corporations someday where no one wants to work or visit because the computational infrastructure has been neglected to the point that it is rude, erratic, swampy (although brilliant), and unforgiving, yet no one has time to tame or retrain it.

Of course there is a strong counterforce to keep the environment uniform. As Danny Hillis pointed out to me, "The reason we create artificial environments instead of accepting natural ones is that we like our environments to be constant and predictable. We used to have a computer editor that let everyone have a different interface. So we all did. Then we discovered it was a bad idea because we couldn't use each other's terminals. So we went back to the old way: a shared interface, a common culture. That's part of what brings us together as humans."

Machines will never go completely on their own way, but they will become more aware of other machines. To survive in the Darwinian marketplace, their designers must recognize that these machines inhabit an environment of other machines. They gather a history together, and in the manufactured ecology of the future, they will have to share what they know.