Electrical retinal: epiretinal and subretinal
The first road stimulates the surviving retina electrically. The epiretinal Argus II (around sixty electrodes) reached the clinic and let blind patients with retinitis pigmentosa detect motion and slowly read large letters — then was commercially discontinued, stranding implanted users and becoming a lasting ethics lesson about device abandonment. The subretinal photovoltaic approach, PRIMA, tiles a wireless photodiode array under the retina, powered by near-infrared light projected from glasses; recent trials in patients with geographic-atrophy macular degeneration reported them reading letters and words — a notable milestone for a degenerative disease with no other restorative option.
Cortical: writing straight onto V1
For patients with no eye or no optic nerve, the retina is not an option, so the second road writes straight onto primary visual cortex. Intracortical microstimulation through Utah arrays in V1 let a blind volunteer recognize letters and simple shapes in a landmark study; surface arrays (the Orion cortical device) take a lower-resolution, lower-risk approach. ICMS detection thresholds and per-electrode phosphene maps are measured for each participant before any image can be written.
The Shannon safe-stimulation criterion relates charge density per phase D (in microcoulombs per square centimetre) to charge per phase Q (in microcoulombs); stimulation is considered safer below the line, with k roughly 1.5 to 2.0. It sets the ceiling on how much you may inject per electrode before risking tissue damage — the hard constraint behind the brightness-size coupling of Guide 2.
The cortical advantage is that it works regardless of eye disease — the largest addressable population. The costs are steep: you must recreate the entire retinotopic map from nothing, accept higher surgical risk on the brain itself, and confront the anatomy of V1, much of which is folded deep inside the calcarine sulcus and hard to reach with surface electrodes. Cortex is the doorway of last resort, and the most demanding to encode.
Optogenetic restoration
The third road changes the physics of stimulation. Optogenetic restoration delivers a light-sensitive opsin into surviving inner-retinal cells and then drives them with light-projecting goggles. In 2021 a patient with retinitis pigmentosa recovered partial, goggle-mediated vision using a red-shifted opsin plus amplifying eyewear — able to locate, count and touch objects on a table. This was the first optogenetic sensory restoration reported in a human, and its promise is targeting: put the opsin only in the intended cell type and you could finally deliver, physically, the encoder that Guide 3 could only compute.
The preclinical frontier of writing light
The fourth road is largely still in animals and dishes, and it defines the ceiling of what precise writing might become. Near-infrared and upconversion optogenetics uses nanoparticles that convert deeply-penetrating infrared into visible light in situ, easing the scattering and delivery problem. Chemical photoswitches are small molecules that make surviving retinal cells light-sensitive without any gene therapy. And two-photon holographic ensemble writing — using holographic two-photon stimulation — sculpts light to activate chosen neurons with near-single-cell precision, even 'playing in' specific percepts by driving defined ensembles in animals.
Be exact about status. Upconversion, photoswitches, and holographic writing are largely preclinical — animal, ex vivo, or in vitro. They tell you what the physics could ultimately allow, not what a patient can receive today. Holding both truths at once — genuine excitement and clear-eyed staging — is the intellectual honesty this volume demands.