The Chip, the Cursor and the Count Nobody Can Check

Imagine a room in Toronto, or in Abu Dhabi, where a man who cannot move his hands sits before a screen. Nothing touches a keyboard. A cursor drifts across the display, hesitates over a square, and settles. He thinks about the move, and the arrow obeys.

That scene is imagined. Its ingredients are not. Neuralink's first product, which the company calls Telepathy, is described in one recent report as a system that "allows people with paralysis to control a computer cursor and keyboard using brain signals." According to the company and Elon Musk, trial participants have moved a cursor, played chess and used design software by imagining movements. The report attributes those performance claims to the company and cites no independent verification (Business Today).

The strangest number in this story is not the chess game. It is the unit. In October 2026 Neuralink said a participant reached 11.32 bits per second of cursor control, which it described as a new record (allthings-elon.com). A bit is the basic unit of information, the answer to a single yes-or-no question. The same report says the company credited training on more than 50,000 hours of neural recordings for the result. The result was presented as a technical update rather than a peer-reviewed paper, and it has not been independently checked.

So this is less a story about whether a chip can move a cursor than about what happens when a mind becomes measurable in bits, updated in software and counted by a company. Before that question, we have to count the people. The counting is harder than it sounds.

Twenty-One Patients, Counted by One Company

The most recent count attributed to Neuralink itself is 21 people enrolled in trials worldwide, as reported for January 2026 (The Star). According to the Reuters report carried by The Star, Neuralink said it was maintaining a record of zero serious device-related adverse events at that point.

Then the numbers go soft. Neura Pod, a third-party tracker, reports 27 or more participants as of October 2026 (Neura Pod). That is an outside estimate, not a company count; the latest company figure in this record is the January one. In my judgment, a number you cannot trace to a company statement is an estimate built from public traces, and it should be read as one. Call it what it is: a floor with a spreadsheet.

The geography is as uneven as the count. The first human trial in the United States began in September 2023 at the Barrow Neurological Institute in Phoenix. Health Canada approved a Canadian trial in November 2024, and the University Health Network in Toronto implanted two patients in August and September 2025 (Wikipedia, Neuralink). University College London reported on 28 October 2025 that its first UK patient, Paul, had controlled a computer by thought only hours after the implant procedure (Wikipedia, Neuralink).

None of the trial sites named in this piece is in Africa. That is a statement about the record, not about what is possible.

What the Chip Shows, in Order of Proof

Neuralink has made claims about three capacities: moving a cursor, speaking and seeing. The evidence for each is unequal. The honest way to read it is by the strength of the proof, not the size of the headline.

The first is the cursor and the keyboard. The N1 implant, marketed as part of Telepathy, carries 1,024 electrodes across 64 threads (Business Today). The 11.32 bits-per-second record is the best-known number, and it is a company result that has not been peer-reviewed (allthings-elon.com).

The second is speech. In September 2026 Neuralink posted a video of a trial participant saying "I love you" through the device. A blog report on the video says the company's disclaimer describes the devices as investigational, not approved by the FDA, and warns that the results may not reflect all participants or future outcomes (explainx.ai). A video is worth watching, but it is not a study. In my judgment, the most rigorous speech results in this record come from academic BrainGate groups rather than from Neuralink, as the next section shows.

The third is vision. Neuralink's Blindsight project received FDA Breakthrough Device Designation in September 2024. It targets people with total vision loss from optic-nerve damage whose visual cortex is intact, and it stimulates the visual cortex directly (Wikipedia, Blindsight). In an X post dated 30 October 2026, reported by Business Today, Musk called Blindsight the company's second product and wrote that it is "ready from an engineering standpoint." He added:

"Just waiting on regulatory approval."

That line may be the most important sentence in the Blindsight story. The designation is confirmed. Musk's account stops at engineering. No human Blindsight implant is described in any source this piece can confirm, so none is claimed here. Imagine a blind person waiting for a letter that has a date, and then another date after that. The gap between "ready" and "approved" is where the years go.

The Rivals and the Quiet Academic Lab

Neuralink is not alone, and the rivals are making different bets.

Synchron, with its Stentrode device, closed a $200 million Series D in November 2025, led by Double Point Ventures. Synchron will use the money to accelerate pivotal trials and prepare for commercial launch. The coverage gives 16 electrodes for Stentrode and 1,024 for Neuralink's N1 (MedTech Dive). Electrode counts are not performance, but they show the two companies betting on different designs. The article gives no trial dates or patient numbers, so none are given here.

In my judgment, the rivalry matters less for which device wins than for which company's numbers get checked first. Synchron's round was framed around trials that are still to come. Neuralink's latest figures arrived as a technical update and a video.

The academic side of the field publishes differently. In March 2026 BrainGate published a typing study with two participants. One reached 22 words per minute with a 1.6 percent word error rate. Brown University stressed that the device remains investigational (Brown University). The sample is tiny, but it is a published academic report that others can examine, which a company video is not.

My judgment is that a published academic study of two participants is a different kind of evidence from a video of one participant, posted by the company that makes the device. Setting the two side by side as equals is advertising.

What Has Not Yet Been Shown

The fastest way to be fooled by this field is to read the headline and skip the ledger of what is missing.

Safety over years comes first. The first human trial was approved in May 2023, after the FDA rejected the company's first application in 2022, citing battery safety, wire migration and removal concerns (Wikipedia, Neuralink). A few years of follow-up in 21 enrolled people says little about a lifetime in one brain. Public reporting does not say how the 2022 concerns about battery safety, wire migration and device removal were answered in full, and in my judgment that question should be asked before anyone enrols.

The second gap is the threads. In a later interview, the first participant, Noland Arbaugh, said that about 85 percent of the threads had detached, because his brain had moved about three times as much as Neuralink had expected. The FDA later approved the company's proposed fixes, and software updates restored much of the lost performance (Wikipedia, Neuralink). The 85 percent figure is Arbaugh's account, not a company disclosure.

The third gap is independent data. The 11.32 bits-per-second record is a company result, presented as a technical update, and it has not been independently checked. No peer-reviewed Neuralink paper on the device is cited in this essay.

The fourth gap is the valuation. A $650 million Series E in June 2025 is reported at about $9 billion pre-money (Wikipedia, Neuralink). Public reports of that round disagree on the amount, so the figure should be read as approximate. It is a priced round, not a market quote for a device that is not yet approved for sale. In my judgment, a price for an unapproved device is a bet on what the next five years will prove, and the bet is larger than the public evidence.

Whose Mind Is the Interface?

The cursor hides a harder question. The device reads signals from a brain and returns them to a screen. It is also a company's software, trained on tens of thousands of hours of data and updated over time. In my judgment, part of what the patient gets back comes from the decoder, the model that turns neural signals into cursor movement, and not from the hardware alone. Who owns that model? Who owns the hours of neural data it learned from? If the patient wants the device removed, what happens to the software that learned their particular signals?

The questions are open. The Pawanax view is simpler and stricter: a consent form for a brain implant should cover the data, the decoder and the updates as separate things, not only the surgery.

The thread story makes the point physically. A brain moves. The threads were meant to stay put, and in the first participant's case many of them did not. The body is not fixed, and neither is the company's model. Both move, and the patient carries the weight.

Ask these questions before enrolling, whether you are a patient or a hospital in Nairobi that one day considers this technology. Ask how many people have been implanted, and who counted them. Ask what the threads do when the brain moves. Ask who owns the decoder and the data it learned. Ask what happens to the device if the company changes its software, stops supporting it or runs out of money.

In the imagined room in Toronto, the cursor stops above the square. The man thinks the move once more, and the arrow waits. The chip will obey. The question that remains is whose signals the model learned from, and who keeps them.

Editorial note: This essay expresses the founder’s ideas, developed with AI-assisted research and writing (parallel Claude Haiku 5.5 agents) and checked against the linked sources on 30 October 2026. Convictions, proposals and forecasts are identified in the text; scenes framed as imagined are illustrative.

Cover photograph / illustration: Laurens R. Krol (Wikimedia Commons user Lrkrol). CC BY 4.0. Original source. Illustrative image: a brain-computer interface experiment, with a participant wearing an EEG cap. Photograph: Laurens R. Krol, via Wikimedia Commons. CC BY 4.0. This image does not document the events described.