The return desk
The body library opens at nine. By 09:03, someone has returned a pair of legs with sand in the knees. By 09:07, a wheel borrower wants to know whether stairs count as a manufacturing defect.
The attendant points to the returns policy: PLEASE BRING BACK THE BODY YOU BORROWED, INCLUDING ANY PARTS THAT BECAME A SEPARATE EXPEDITION.
In this imagined neighborhood, a robot’s control core can move between compatible bodies. Wheels for the archive. Legs for the garden steps. A floating hull for the pond. The personality comes with you. Your preferred turning radius does not.
One regular always borrows the gripper module on Sundays. It is not for a job. It is for model trains. Another checks out the smallest chassis available and spends the afternoon looking under a wardrobe. Nobody asks why. The library has a very good privacy policy.
Fictional service model: retain an identity across approved, swappable bodies. No claim that current robot controllers transfer universally or that machine identity works this way.
A chassis is an agreement about geometry, power, sensing and contact. Copying a memory does not copy reach, braking distance, joint limits or calibration.
Preflight question: what do these actuators actually do? The correct answer is not “what the old ones did.”
The wrong kind of upgrade
The advertising campaign says BE ANYTHING. The librarian prefers BE SPECIFIC.
A visitor asks for the most advanced body. The attendant asks where it is going. “Across the courtyard.” Wheels, then. “And up the steps.” Perhaps legs. “With a tray of soup.” They discuss the ramp.
This is the part the campaign leaves out. A spectacular body can be an unnecessarily difficult answer to an ordinary route. A new capability brings new ways to fall, snag, sink or discover that the charger is still attached to your previous self.
Try the checkout desk above. It is a written thought experiment: three routes, three bodies and a return plan. The outcomes do not predict real robot performance. They expose the assumptions our imaginary librarian would ask you to check.
The visitor eventually borrows wheels and uses the ramp. The soup arrives without becoming weather. The stairs file a complaint about being excluded from innovation.
Capability selection follows the task, environment and recovery method. A body with more degrees of freedom also has more states to estimate and control.
Our checkout cases are authored fiction. Their terrain categories are deliberately coarse. No load capacity, speed, stability or success probability is established.
Before a real transfer: verify physical interfaces, controller compatibility, sensing, calibration, permitted loads and a safe stop. A route that works in a story is not a validated deployment plan.
One last thing before you go
The thought experiment becomes interesting at the return desk. Who owns the calibration? What happens to private maps? Can a damaged leg explain what happened to it? Does lending a body require lending its experience?
A useful library would need maintenance records as much as shelves. It would need a way to say “this pairing has never been tested” without turning that sentence into a sales obstacle.
At closing time, the attendant counts everything back in. Six wheels. Four legs. Three grippers. One hull smelling faintly of pond.
A control core is still sitting on the counter. It says it would like a quiet evening without a body. The attendant dims the light and puts a book beside it.
Return contract: remove private task data, inspect wear, preserve a truthful maintenance history and revalidate any changed calibration. None of these follows automatically from a standard connector.
Tomorrow’s interesting question is not just whether a core fits a socket. It is whether the next borrower can trust the history attached to it.
Fictional closing log: all hardware accounted for. One patron has selected zero locomotion for the evening.
If you could borrow a body for one afternoon, where would you take it?