Head direction cells in the medial entorhinal cortex are usually treated as a fixed functional class, and whether that identity is hard-wired or self-organized is yet to be resolved. This study put this question to the test, comparing entorhinal activity in mice covering an arena in two ways: freely-moving, or head-fixed to an autonomous-driving cart. Here Liu et al. show that the neuron's identity proved to be substantially dynamic. Roughly half the cells that were tuned during free foraging lost that tuning on the cart, a larger group acquired tuning, and a third group held on to their tuning throughout. More surprising, the coding on the cart was not degraded but made sharper, with higher selectivity, greater stability and better decoding of heading. This is the opposite of what a directional signal built partly from self-motion should do when the animal is carried rather than walking. Only the invariant group preserved its cell-pair correlation structure across conditions consistent with attractor dynamics. Both switching groups lost it entirely, pointing to two coding schemes running in parallel rather than a single network losing its anchor. Control of the switch turned out to be sensory rather than kinematic: stripping visual and odor cues from the arena abolished the effect, while varying cart speed across four regimes had no effect. The new code also had to be learned, sharpening block by block over the first cart session, with the recruited cells developing spatial tuning aligned to nearby walls. These experiments were performed using functional calcium imaging acquired with a miniature two-photon microscope (mTPM V2, LF headpicece, GINKGO-mTPM software, TVS-FL-01 laser; TRANSVISTA (formerly Transcend Vivoscope)) using 920 nm excitation and collected at 9.5 Hz through an implanted microprism, and recorded alongside synchronized behavioral and movement tracking cameras. The identification of  >11,000 neurons across both modes of navigation in a single environment allowed Liu et al. to show that entorhinal directional representations reorganize depending on how an animal moves through its world.