The read/write asymmetry

Across essentially every modality, BCIs read far better than they write. Decoding movement, speech, or intent from neural populations has advanced to fluent demonstrations, whereas writing arbitrary, naturalistic patterned activity back into the brain, to evoke a specific percept, memory, or coordinated command, remains crude. The asymmetry is structural: reading is an estimation problem where redundancy helps, since more channels average out noise, while writing is a control and inverse problem in which you must reproduce, at cellular resolution and correct timing, spatiotemporal patterns whose target code you do not fully know.

Consequences ripple through the field. Sensory feedback for prostheses, memory prosthetics, and vision restoration are all write-limited; a closed-loop system inherits the weaker of its two channels; and roadmaps that assume symmetric progress overestimate near-term bidirectional capability. Honest frontier assessments treat high-fidelity write as the field's central unsolved problem.