A building is not one thing but several layers that change at different rates: Site (eternal), Structure (30-300 years), Skin (20 years), Services (7-15 years), Space plan (3-30 years), and Stuff (daily). Trouble comes when a fast layer is trapped by a slow one, so good design lets each layer slip independently rather than binding them together.
Cheap, low-status, unloved buildings (warehouses, sheds, garages) adapt most freely because no one guards their form, so occupants rip into walls and reroute them at will. High-status designed buildings resist change because their form is treated as precious, making them brittle as needs shift.
A building's real performance only becomes visible over years as occupants and weather expose what works and what fails, information invisible at the moment of design or opening. Judging a building by its debut photograph mistakes a hypothesis for a result.
What gets built is shaped less by use than by how the money arrives: speculative developers, grant cycles, tax rules, and resale expectations dictate layout and quality. The building embodies the financing structure that produced it more than the activity it houses.
Because future needs are unknowable, Brand argues you should not design for a predicted use but build in slack, generic spaces, and room for maneuver so the building can absorb whatever actually comes. Adaptability is bought by refusing to optimize too tightly for present requirements.
MIT's temporary wartime Building 20, a flimsy plywood structure meant to be torn down, became its most productive research space for decades precisely because occupants felt free to knock out walls, run cables through ceilings, and reshape it overnight. Its throwaway status licensed the experimentation a prestigious building would have forbidden.