How could Neuralink’s Blindsight let a blind person see?

In three steps, none of which uses the eyes.

1. Camera

   A small camera, for example on a pair of glasses, sees the scene. A computer turns each frame into a pattern: which electrodes to pulse, and how strongly.

2. Electrodes

   Fine electrodes in the visual cortex, at the back of the brain, deliver tiny pulses of current to the cells around them.

3. Dots

   Each pulse makes the person see a dot of light, a phosphene, at one fixed place in their view. Together the dots draw a rough picture of the scene.

Phosphenes are not new. Neurosurgeons have made people see them by stimulating the visual cortex for decades, and blind volunteers have tested cortical implants. In a 2021 study Eduardo Fernández and colleagues placed 96 electrodes in the visual cortex of a blind woman for six months; patterns of dots let her identify some letters and the edges of objects. The idea only works if the visual cortex is intact, which is why it suits blindness from damage to the eye or the optic nerve.

Because the visual cortex is a map of what you see, and the map is out of proportion. Neighboring spots of cortex see neighboring points of view, but the center of view, where you read and recognize faces, gets far more cortex than the edges. Horton and Hoyt’s classic 1991 measurement gives about 17.3 ÷ (E + 0.75) millimeters of cortex for each degree of view at E degrees from the center: about 23 mm per degree at the center and 1.6 mm at 10°.

In Phosphene Lab, which uses that rule, the central 2° of a 20° view is 1% of its area but takes about a quarter of the cortex, and the central 5° about half. So electrodes spread evenly over the brain land densely in the middle of the view and sparsely at the edges. With 1,024 of them over the map of the central 20°, neighbors sit about 2.1 mm apart on the cortex. In the view, that is 0.09° apart at the center and 2.5° apart at the edge: sharp in the middle, coarse at the edges.

Roughly one dot per electrode, so the electrode count is the resolution, and it has been small. Second Sight’s Orion implant, tested in six people from 2018, had 60 electrodes on the surface of the visual cortex. Fernández’s 2021 array had 96. Illinois Tech’s wireless implant, first placed in 2022, has 400. In monkeys, Xing Chen and colleagues implanted 1,024 electrodes in 2020; hundreds of them made phosphenes, and the animals recognized shapes, motion and letters drawn in dots.

In the model, 60 dots read as a scatter of stars: you could find a bright doorway but not its shape. About 1,000 start to draw it, the same count as Neuralink’s N1 chip, which has 1,024 electrodes on 64 threads. Neuralink has not published how many electrodes Blindsight will have, so the doorway is the model’s picture, not a preview of the device.

Neuralink announced on September 17, 2024, that the FDA had given Blindsight Breakthrough Device designation, which speeds up review; it is not an approval. Elon Musk wrote then that the vision will “at first be low resolution, like Atari graphics”, and that it could one day be better than natural vision and see in infrared or ultraviolet. Those are his claims.

At the end of October 2026 Musk posted that Blindsight is “ready from an engineering standpoint. Just waiting on regulatory approval.” He did not say which approval. As of early November 2026, Drug Development & Delivery reports, no Blindsight trial was listed on ClinicalTrials.gov and no human results had been published. In its June 2025 update Neuralink showed a “Calibrated Vision” simulation of letters drawn in dots; that is a company simulation, not a recording of what anyone saw.

1. 1 See

   Open Phosphene Lab. It starts at Neuralink N1 · 1,024 with A doorway. Gap between dots reads 0.09° at the center and 2.5° at the edge, and Dots in the middle 5° reads 50%: the dots crowd into the middle.

2. 2 Change

   Tap Orion implant · 60. The doorway falls apart into a scatter of stars: the gap is 0.37° at the center and 10° at the edge, and Same pixel count as reads ≈ 8 × 8. Then tap Utah array, 2021 · 96 and Illinois ICVP · 400 and watch the shape come back.

3. 3 Understand

   Tap Neuralink N1 · 1,024 again and drag Field covered down to 8°. The same electrodes, squeezed into the middle: 76% of the dots fall within 5°, and the doorway is sharp but it is tunnel vision. Then tap Ten thousand (future) and switch to A face.

Phosphene Lab is a best case. It spreads the electrodes evenly over the brain’s map of the central 20° of view and gives each one exactly one dot, at the place the map says. Real phosphene maps are irregular. Not every electrode makes a dot, and size, brightness and color vary with the current and the spot. Stimulating many electrodes at once can merge or fade dots.

Real arrays cover a small patch of cortex, not the whole map of the central 20°, and much of the visual cortex, especially the map of the outer view, lies deep inside a fold. Phosphenes are fixed to the brain’s map, so they move with the eyes: a head camera works best if the eyes stay still or are tracked. Whether people blind from birth would see phosphenes at all is an open question. The film uses short, credited excerpts of Neuralink’s own videos on YouTube; the dots, the brain map and the numbers come from the model.