Science Corporation says its PRIMA retinal implant has been placed in a first patient with Stargardt disease. Surgery took place at Sydney Eye Hospital in May 2026, months before the October 13 announcement at TechCrunch Disrupt. The company reports recognition of a letter in the first rehabilitation session and a word in the second.
That is an early response in a new disease population, not established treatment efficacy. PRIMA already has more substantial evidence in age-related geographic atrophy, but the inherited-disease expansion must demonstrate its own repeatable vision, safety and daily usefulness. European approval reported for geographic atrophy does not establish approval for Stargardt.
The implant replaces a sensory function, not a genetic cause
The first-patient report describes a system that captures a scene with glasses, processes a simplified image and projects near-infrared light onto a photovoltaic implant beneath the retina. The chip turns that light into electrical stimulation; surviving retinal cells convey signals through the optic nerve to the brain.
Stargardt is an inherited disorder damaging central photoreceptors, often involving the handling of vitamin A byproducts, with varied genetics and progression. PRIMA does not repair the gene or stop the disease process. Its rationale is that lost photoreceptors can be bypassed when inner retinal cells and the remaining visual pathway still work.
Max Hodak described a two-by-two-millimeter implant and a roughly one-hour procedure. Its current image is a sharp, high-contrast black-and-white patch, with a field he compared to looking through a straw. It is not full-color natural vision or restoration of the whole visual field.
The registered PRIMAlia study estimates five participants and covers inherited macular degeneration including Stargardt and retinitis pigmentosa. Severe central loss, a functioning visual pathway and an atrophic area large enough for the implant narrow eligibility. One participant cannot validate outcomes across that population.
Earlier AMD evidence offers a rationale, not a Stargardt result
A prospective study enrolled 38 people with geographic atrophy due to age-related macular degeneration. Thirty-two completed the 12-month assessment; 26 had clinically meaningful acuity improvement using PRIMA, or 81% of those assessed. An analysis accounting for missing data estimated 80% of all implanted participants would meet that threshold.
The peer-reviewed study results also reported peripheral natural visual acuity comparable to baseline. Participants used the system for letters, numbers and words. This supports coexistence of prosthetic central input with remaining natural vision, not automatic transfer of the same percentage to Stargardt.
The study was open-label and single-group, with participants compared to their own baselines, not randomized against sham surgery or another intervention. That can provide useful implant evidence, but expectations, rehabilitation and differences between settings remain relevant to interpreting effect size.
Safety must appear beside the gain: 26 serious adverse events occurred among 19 participants. Twenty-one occurred within two months of surgery; 20 of those resolved within two months of onset. Resolution matters, but does not erase the surgical burden or make implantation risk-free.
Stargardt can begin earlier, progress differently and leave varied damage patterns. Younger patients may need work, mobility and communication benefits over decades, not only a clinic reading task. Shared photoreceptor loss makes the experiment plausible; it does not guarantee equivalent tissue response, durability or patient priorities.
Why Science places a retinal prosthesis inside its BCI strategy
Hodak co-founded Neuralink and was its president before leaving in 2021. Science's broader work includes surface brain sensors and biohybrid interfaces combining engineered cells and electronics. PRIMA is the nearer clinical business, not proof those more speculative programs already work.
Science acquired PRIMA assets from France's Pixium Vision in 2024 rather than originating the entire program. In July it reported European regulatory approval for geographic atrophy. A March 2026 Series C supports commercialization and research; Hodak says the vision business can help fund biohybrid development.
The retina is part of the central nervous system, so Hodak uses BCI more broadly than an implant recording brain activity to move a cursor. PRIMA introduces signals through surviving eye circuitry; Neuralink's proposed Blindsight would stimulate visual cortex. A cortical route could theoretically bypass retinal or optic-nerve damage, but brings different technical demands.
These are not direct substitutes today. PRIMA depends on a viable retinal-to-brain path and an ophthalmic surgical workflow. Its early Stargardt use tests whether that focused intervention can cross diseases, not whether it can serve every cause of blindness or validate a general neural-interface platform.
Rehabilitation, risk and years of use remain part of the result
A first letter followed by a word suggests learning to interpret artificial input. Camera settings, processing, gaze strategy, contrast and training can all affect performance. An encouraging supervised session cannot separate device function, practice and favorable testing conditions on its own.
Investigators need standardized follow-up intervals, acuity and recognition measures, functional tasks, usage and adverse events for every participant, including people with little benefit. Five-person exploration can inform a larger study, but one unusually good or poor response can dramatically change its percentages.
Selection for severe loss with preserved pathways may identify likely responders while limiting generalization. PRIMA may not suit less advanced disease, extensive inner-retinal injury or other eye conditions. Gene therapy, cell replacement and optogenetics target different mechanisms or stages; they are not interchangeable with bypassing lost photoreceptors.
Everyday use also requires camera glasses, external processing and rehabilitation around a narrow monochrome window. Reading a label or recognizing a central object could be meaningful, but patients' chosen use outside testing matters more than a list of possible tasks.
Long follow-up should track implant stability, retinal response, maintenance and external components. Younger recipients may live with hardware for decades. Improved future electrodes or field size may not be deliverable to early implants through software, making upgrade compatibility another practical uncertainty.
The next evidence should be measurements, not another first
PRIMAlia updates should report visual performance, safety, rehabilitation and follow-up for the first patient and additional participants. Consistent meaningful gains would support a larger disease-specific study; mixed results would emphasize selection and anatomy rather than automatically ending the program.
European geographic-atrophy delivery is a different test: surgeon training, manufacturing reliability, rehabilitation, reimbursement and patient support after approval. That experience can improve operations, but commercial maturity in AMD must not fill gaps in Stargardt efficacy.
Hodak says Science is developing a wider field and smaller electrodes, forecasting a new generation in humans in about a year and discussing near-native acuity over five years. These are company projections, not achieved clinical milestones. Finer hardware must translate into perception without unacceptable surgical, tissue-contact or power tradeoffs.
Human implantation of a successor would begin its testing, not complete it. Processing, contrast, neural adaptation and placement also determine useful vision. Better specifications need demonstrated acuity, daily function, safety and durability.
The first Stargardt case crosses an important operational boundary: the system has reached inherited retinal disease and produced an initial response. The unresolved question is whether several patients can gain durable, useful vision with acceptable risks. Cohort measurements, real-world use and qualified hardware results—not a compelling stage demonstration—will answer it.