Neuralink has put chips in at least twelve human brains. The story of what happened next is more interesting than anything Elon Musk has posted about it.

Start with the hardware problem. Within a month of Noland Arbaugh’s surgery in January 2024, roughly 85% of the electrode threads pulled away from his brain tissue. The N1 implant uses 64 polymer threads thinner than a human hair, carrying over a thousand electrodes that need to sit close to neurons to read their signals. When tissue shifts, swells, or heals around the foreign material, those threads can retract. In Arbaugh’s case, most of them did.

Neuralink’s engineers avoided a second surgery by doing something clever: they pushed software updates that redistributed signal processing across the electrodes that still had contact. It worked. Arbaugh kept browsing the web, playing games, making video calls — all by thought. But a software patch is not a hardware fix. The question of how you anchor microscale filaments inside a living brain that actively tries to wall off foreign objects remains open, and it matters enormously for any company planning to put these devices in thousands of people.

Which is exactly what Neuralink says it will do. In December, Musk announced plans for “high-volume production” and a nearly automated surgical procedure in 2026. A redesigned threading approach would punch through the dura mater without removing it — a change that, if it works, would shorten surgery and reduce risk. The company also holds FDA Breakthrough Device Designation for Blindsight, its visual prosthesis program targeting the visual cortex. Human Blindsight trials are expected before year-end.

The numbers are moving. Trial enrollment has grown steadily, with estimates now approaching twenty participants across multiple sites.

But the most striking thing about Neuralink’s clinical program isn’t the engineering. It’s what the patients say. Arbaugh, paralyzed from the neck down since a diving accident, described his pre-implant life in terms that don’t require medical interpretation: he stayed up all night and slept all day because nothing made waking up worthwhile. The implant changed that. Brad Smith, the third recipient and an ALS patient who’d lost his voice, used the device to produce and edit a YouTube video through thought and eye tracking alone — narrated by an AI reconstruction of his original voice. A paralyzed veteran implanted in April 2025 described the experience not in terms of bits-per-minute or cursor accuracy but purpose.

These stories don’t answer the engineering questions. They don’t substitute for the peer-reviewed clinical data that Neuralink has so far declined to publish, drawing criticism from researchers who argue that an implanted brain device deserves more scrutiny than an X post. They don’t resolve the animal welfare concerns that have followed the company since its primate testing program. Twelve patients — even twenty — is a case series, not a product.

But they do clarify what the stakes are. The people who volunteer to have a chip drilled into their skull aren’t abstractions in a regulatory filing. They’re making a calculation that a device with known risks and unproven durability is better than the life they have now. Getting from a dozen of them to a thousand will require Neuralink to match its production ambitions with the kind of transparent, published evidence that the rest of medicine demands. So far, the company has shown it can build remarkable technology. Whether it can build the trust to go with it is a different question.