The concept sharpens the diagnostic Ingold's framework offers for AI collaboration. The prompt-execute cycle is not merely an efficient medium; it is a structural realization of the transport mode. The user specifies endpoints (the prompt), the machine calculates the route (the generative process), the user receives the artifact at the destination (the output). The terrain between prompt and output — the dependency tangles, the logic traces, the architectural decisions, the alternative approaches that were considered and rejected — is traversed by the machine and is invisible to the user. She did not travel through it; she was transported across it.
The cost is not productivity but perceptual development. The London taxi driver who trained under the Knowledge developed measurably enlarged hippocampi from the years of wayfaring through the city's streets on a moped. Drivers who used GPS did not. The wayfaring deposited neural structure; the transport did not. The knowledge was not in conscious memory but in the architecture of the brain, built layer by layer through terrain-engagement. The AI-assisted developer who prompts for working code is the GPS-using taxi driver of the computational landscape: she arrives at the destination reliably and develops none of the perceptual structure that years of wayfaring through codebases would have deposited.
The framework does not imply that transport should be refused. Transport is valuable precisely because it is efficient, and most of daily life depends on modes of transport (literal and figurative) that work well for their purposes. The concern is that transport, when it becomes the dominant or only mode of movement through a domain, eliminates the conditions under which the complementary knowledge of wayfaring is cultivated. A civilization that uses transport alongside wayfaring maintains both kinds of knowledge. A civilization that uses only transport produces a generation that arrives at correct destinations without ever developing the capacity to find the way when transport fails.
The failure case is the test. Every transport system eventually encounters a situation for which it was not designed: the GPS that directs into a flooded road, the prompt that produces confident hallucination, the calculated route that misses the real-world constraint the model did not encode. In these moments, the traveler must fall back on wayfaring — on the perceptual capacities she developed through engagement with terrain. If the capacities were not developed, the fallback is not available. The traveler who has only ever been transported cannot navigate by wayfaring when the transport fails, because wayfaring is not a latent capacity that can be activated on demand. It is a cultivated capacity that must have been built through sustained practice.
The transport/wayfaring distinction is developed most fully in Lines: A Brief History (2007), drawing on fieldwork with Skolt Sámi reindeer herders and on comparative ethnography of navigation in circumpolar regions. Claudio Aporta's fieldwork among Inuit hunters provided the sharpest empirical documentation of what is lost when GPS displaces wayfaring.
Endpoints, not terrain. Transport valorizes arrival and ignores the landscape between origin and destination.
The traveler does not travel. In transport, the movement is performed by the system; the traveler is delivered.
Efficient but epistemically shallow. Transport produces reliable arrivals and deposits no terrain knowledge in the traveler.
AI collaboration as transport. The prompt-execute cycle is the structural form of transport applied to making, with the terrain of the problem space traversed by the machine.
Transport fails at the edges. Every transport system breaks down in situations it was not designed for, and only travelers with wayfaring capacity can navigate the failure.