A chip that sits under the retina got marketing approval in Europe in July. First sales happen in the coming weeks.
The company is Science Corporation. Its founder Max Hodak talked about it on No Priors yesterday. He co-founded Neuralink before this.
The detail I keep coming back to is from the clinical trial. Patients filling in Sudoku puzzles. Patients reading books.
These are people who went blind from macular degeneration.
How the chip works
Worth stating the mechanism, because it does not look like what most people picture when they hear brain-computer interface.
The chip is implanted under the retina. The patient wears glasses with a laser projector that beams an image onto the chip, and the chip stimulates the retina directly, bypassing the dead rods and cones to get a visual signal back into the brain.
Hodak's own analogy: think of it as a cochlear implant for the eye.
He calls cochlear implants one of the biggest impacts in all of medicine. If you have seen a video of a newborn having one switched on for the first time, the reaction is striking.
This path was bought, not built. In late 2022 a small French company called Pixium had by far the state of the art in electrically stimulating the retina. The technology came from an inventor at Stanford and was licensed to them. Science got to know them over a couple of years, then acquired them.
Hodak's line: "We saw something that I think kind of nobody else really saw at the time."
It took roughly two more years post-acquisition to reach the point where approval was possible. They got European marketing approval in July, and the first sales happen in the coming weeks.
Not "expected within a few years." It can be sold now.
The limits he did not skip
This section is worth more than the result itself, because of how restrained he was about it.
He called the retinal prosthesis a good proof of concept that they are on the right track, then immediately listed what is wrong with it.
The field of view is small, like looking through a straw. It is black and white only.
What needs adding is depth of grayscale. He thinks red and green are reachable, and blue is trickier.
When Sarah Guo asked about variation across patients and where the ceiling sits, he did not produce a clean number. He said the main thing was the existence proof, because success had previously been an impossible outcome.
Then the line that carries the most weight:
"Nobody had previously ever been able to restore a form vision image in the mind's eye of a blind patient in this way."
He also said something human: "This is one of those things that seems too good to be true." So he watched the videos himself, met one of the patients, and talked to the surgeons.
Why he thinks this is easier than drugs
This is the section I most want to pull out, because it applies to anyone doing engineering.
He set up a comparison.
Small molecule drugs: you can do drug discovery for a decade, run a clinical trial, turn over a card, the answer might be no, and then everybody goes home.
Electrodes: "If I put electrodes in M1, you will probably be using a computer in an hour."
The difference is not which one is smarter. It is the shape of the feedback.
The biology path means betting ten years and receiving a single yes or no, and after a failure you often do not know what to fix. The device path means you know how to make the thing better. His phrasing was that they have a clear sense of how to improve it.
Things that can be engineered converge by iteration. Things that cannot only get luck.
That applies well beyond medical devices.
He treats the brain as a computer, literally
He was direct about this from the opening.
"Very literally, very clearly, plainly as a computer. You can solve computational problems by arranging matter in a certain way and then taking your hands off and pressing go."
Then a framing I had not heard before. We all know the brain in a vat thought experiment.
He says the skull is the vat.
"The brain is connected to the environment through a small number of wires, the cranial and spinal nerves, these little cables that carry your interaction with the world."
He even gave the numbering: the optic nerve is cranial nerve two, and the vestibulocochlear nerve that carries hearing and balance is nerve eight.
So his conclusion: if you can get the visual signal, auditory signal, balance, and somatic sensory motor in and out of the brain, that is an end in itself. That is the central object.
And a half joking line: "If you get vision, hearing, balance, and a kilobit per second of motor control, you're halfway to the Matrix."
Why he is not building a brain keyboard
Guo asked the obvious question. Is the most commercially valuable direction in BCI right now not high-bandwidth non-invasive access to an AI model? Why is that not your focus?
His answer had two layers, and the first one I did not expect.
"Talking or writing is thinking."
He thinks the idea that a fully formed thought sits in your head, and BCI would just read it out faster, is probably not the case.
"It'll feel like it's fully formed, but until you really sit down and try to write it out, it isn't really. And that feeling is misleading."
The second layer was the famous cognitive bandwidth estimate: the brain seems to process information at roughly 10 bits per second.
Several independent lines of evidence point at it. One of his examples: put somebody with perfect memory in a helicopter over Manhattan, ask them to draw what they saw, count all the detail, and it works out to about 10 bits per second over an hour or two.
So in his view the bottleneck is deeply evolved, and it probably rolls up to language.
He did not say a brain keyboard has no value. He said it might be a different kind of product. And he gave an analogy I found unusually sharp:
"It might turn out to be like AR glasses, where our attention was already fully 100% occupied and putting it on the face didn't really change that. We were already consuming all of the available time."
The observation that links AI and neuroscience
Toward the end, Guo asked whether the surge of investor and founder interest in BCI relates to progress in AI models.
He brought up the platonic representation hypothesis, and said plainly that it is controversial in the community. But from where he sits, something real is clearly happening:
"When you look inside these big AI models, the mathematical objects that you see look a lot like the things that you see in neuroscience. If you look at how these AI models represent concepts, and you look at the parts of the brain that represent concepts, you see very similar geometry."
"That to me was one of the first clues."
I am keeping this section because it has a practical implication for anyone working on AI. If the representations really are alignable, then progress in neuroscience and progress in model interpretability may be two entrances to the same problem.
What I take away
One, on the shape of feedback. The electrode versus small molecule comparison is really about something more general. Whether a thing can be improved continuously depends on whether a failure tells you what to fix. Same in product work. Two directions can have similar odds, but if one localizes the cause of failure and the other just makes you bet again, their expected values are nothing alike. I now ask that question first.
Two, on "talking is thinking." He used it to explain why he is not building a brain keyboard, but it punctures a larger illusion. We often assume a finished thought is sitting in our heads and only the expressing part is slow. His read is that the feeling is false, and the thing takes shape during the writing. That matches my own experience. Plenty of ideas I thought I had worked out turn out to be full of holes the moment I start writing.
Three, and this is the one worth chewing on. After saying it seemed too good to be true, the first thing he did was watch the videos, meet a patient, and talk to the surgeons.
That is the same move I have been making this week. I keep verifying auto-generated summaries on X trends, and almost every one contains a claim I cannot source. Most things that sound too good turn out to be real inside a boundary nobody has told you about yet.
Hodak stated his boundary plainly: like looking through a straw, black and white only, blue unsolved. Because he drew that boundary, the line about patients filling in Sudoku puzzles becomes believable.
Sources: No Priors podcast, 2026-08-20, From Restoring Sight to Reimagining the Brain, with Max Hodak. Host Sarah Guo, guest Max Hodak (co-founder and CEO of Science Corporation, previously co-founder of Neuralink). The Prima retinal prosthesis mechanism, the macular degeneration indication, the late-2022 acquisition of French company Pixium (technology originating at Stanford), the July 2026 European marketing approval and first-sales timing, the Sudoku and book-reading observations from the trial, the stated limits (straw-like field of view, black and white only, red and green reachable with blue harder), the electrode versus small-molecule feedback comparison, "the skull is the vat" with cranial nerve numbering, the 10 bits per second cognitive bottleneck and the Manhattan helicopter estimate, "talking or writing is thinking," the AR glasses analogy, and the platonic representation hypothesis along with its contested status are all translations of statements made on the show, and the views belong to the guest. The author has not independently verified Prima's clinical data.
Top comments (0)