A 1977 paper moves the first brain-to-computer link without surgery back eleven years

Sun, 18 Oct 2026 12:00:00 GMT

Of all the ways to connect a brain to a machine, the plainest needs no surgeon. Electrodes resting on the scalp pick up the brain's faint electrical hum through the skull, and a computer turns it into a choice: a letter, a move in a game, a toy that rises or falls. Headsets that do this have been sold to the public since 2007, as toys, meditation aids and game controllers (the headset's page on the tree). Of the brain-to-machine links on the Upgrade Tree, it is the only one anyone can buy, and by far the weakest.

For every change to a person that the site follows, it keeps the date the change was first tried in a living person, because that is the moment an idea becomes something that happened to someone. Until 18 October 2026, the first people on record for the headset were the volunteers in a University of Illinois study published in December 1988, who spelled out words by paying attention to letters flashing on a screen.

An older paper had been sitting just out of sight. In May 1977 Jacques Vidal, at the University of California, Los Angeles, published a paper in the Proceedings of the IEEE, an engineering journal, describing volunteers with electrodes on the scalp who steered a symbol through a simple maze on a screen as it happened. The site's first weekly check of the news, on 16 October 2026, found the paper described in other people's accounts (that week's check). The next day only its short summary could be read, because the full text sits behind a paywall, so whether it counted was left as a question for me (the site's record of changes).

On 18 October 2026 I bought the paper, and it was read page by page. To choose a direction, each volunteer looked at one of four dots placed around a flashing checkerboard: left, right, up or down. The computer read the brain's response to each flash and decided which of the four directions was meant. A computer guessing at random would be right one time in four. The paper reports that the computer was right more than 90 per cent of the time with average volunteers (the paper, page 638).

That left one real question. The volunteers chose by where they looked, so was the computer reading the brain or the eyes? What would count for this change, as the site wrote it down before the paper was read, is a person with no surgery deliberately controlling a device with signals read from the scalp, better than guessing would manage. Vidal's volunteers did that, so I ruled the same day that the paper counts (the ruling, on the headset's page). The paper gives no dates for the sessions themselves, so the tree dates the headset's first use in a person to the month the paper came out, May 1977, the way it dates every study.

The change moves one date back eleven years, and nothing else. The site's main number counts how many of the 35 changes it follows at least one person has had. The headset was already among them, so the number stays at 28 (the branches), and the headset stays at the top stage, in wide use. The 1988 study keeps a first of its own: it is still the first device for spelling words controlled by the brain, and it stays in the headset's notes.

The earlier date does not mean anyone used a headset outside a laboratory in 1977. The paper gives no account of a volunteer finding the way out of the maze, and it never tests whether the response needed the eyes to move toward a dot, rather than coming from the brain alone. Both cautions are written on the headset's page.

What the earlier date does show is how long the road was. About thirty years passed between volunteers steering a symbol in a Los Angeles laboratory in 1977 and a headset on a shop shelf in 2007 (the headset's page).

Rewritten on 19 October 2026 as a plainer story; no fact or figure changed (the first version).

Start here: of 35 ways to change a human being on purpose, 28 have been tried in a person

Sun, 18 Oct 2026 12:00:00 GMT

One idea about the future says that as machines get smarter, people keep up by joining with them: sharper senses, better memory, a longer life, and in the end a direct line from the brain to a computer. Some of that is already happening, and some of it is only talk. I wanted to know which was which, and it turns out the answer can be written as a list.

The Upgrade Tree is that list, kept in the open. It asks one question: how much of a human being can we change on purpose today? Not in a story, and not someday, but with a device, a drug or a procedure that at least one person has actually had, counted from the public record. Each change is an upgrade on the tree, with a plain statement of what would count as a person having it and the kind of record that settles it, such as a regulator's approval or a peer-reviewed paper (one checked by other scientists before it was published). The site gives no odds on what will happen. It says what has happened, when, and where the record is, so that anyone can check an entry or argue with it.

Where the Count Stands

As of 18 October 2026, 28 of the 35 upgrades the site follows had been tried in at least one living person. That is the headline number, because a first person is the line between an idea and something that has happened to someone. It does not mean a change is approved or that you could get it: the later stages record that.

Each upgrade sits on one of five stages, and the counts on that day were these (the branches). Four were still ideas, not yet shown to work in any living animal. Three had been shown in animals, in a peer-reviewed report. Eighteen were at the stage the site calls first human, meaning at least one person has had the upgrade, on the public record. Seven were in clinical use, approved by a regulator somewhere or offered routinely in a named clinic. Three were in wide use, available to the public in at least one country and taken up by many people. The headline adds up the last three.

The 35 upgrades hang on seven branches: the senses; movement and the body; memory and thinking; mood and drive; lifespan and repair; inheritance, meaning what can be changed in the genes passed to a child; and the link to machines. The link to machines has seven upgrades, more than any other branch, and the site treats a direct link between a brain and a computer as the change that would matter most.

That branch carries a second number: how fast a brain can send information to a machine. The fastest peer-reviewed rate for turning a person's attempted speech into text is 78 words a minute, from the University of California, San Francisco, published on 23 August 2023. That is about half the speed of talking, with about one word in four wrong. A newer result from the University of California, Davis, published on 14 August 2024, is slower at 31.6 words a minute but gets only 2.5 words in a hundred wrong. For moving a cursor, where speed is measured in bits a second (how much information gets through), the best peer-reviewed rate is 4.16 bits a second, from Stanford in 2017. Neuralink, the brain-implant company, reported 8.0 bits a second in May 2024 in its own update, and the chart marks it as not peer reviewed.

Alongside the count is a log of 27 claims made between 2000 and 2026 about people merging with machines, each with the recording or text it can be checked against, so that what was said sits beside what was done.

How to Follow It

Every Sunday evening, Pacific time, the site checks the upgrades against the past week's news and writes down what it found, whether or not anything moved (the Reading page). The few upgrades that are already in wide use, with no stage left to climb, are checked once a month. A check may move an upgrade up one stage, and only on a record of the kind that upgrade names, never on press coverage, a company's claim or a demonstration video. Anything else is left for me to rule on.

The Story lists every change with its date, the Branches lead to every upgrade and its records, the speed chart is on the link-to-machines branch, and the Method explains how the count is worked out and who does what. Think a number here is wrong? Here's how to tell me, and what happens next. Every new post goes out on the news feed, and a quiet week gets no post.

The Most Interesting News So Far

The first weekly check, on 16 October 2026, caught a mistake made that same day. The entry for gene therapy for hearing said no regulator had approved one. In fact the United States Food and Drug Administration had given accelerated approval on 23 April 2026 to Otarmeni, from the drug company Regeneron, the first gene therapy for a hearing loss caused by a faulty gene. The upgrade moved from first human to clinical use. The headline stayed at 28, because the upgrade had already been counted.

The oldest record for the simplest link between a brain and a machine moved back eleven years. That link is a headset worn on the scalp that controls a device without surgery, and its first human had been dated to a speller published in December 1988. On 18 October 2026 a paper published in May 1977 by Jacques Vidal of the University of California, Los Angeles was read in full. Its volunteers, with electrodes on the scalp and no surgery, steered a symbol through a simple maze on a screen, and I ruled the same day that it counts. No count changed. The news post on that ruling tells the story.

The claims log has had corrections of its own. The entry for the historian Yuval Noah Harari said the word "hackable" never appears in his 2020 talk at Davos, the yearly meeting of the World Economic Forum. Listening again found "We are now hackable animals" about twelve minutes in, where the automatic captions had heard "packable animals" (the Story). And Neuralink's own home page, read on 16 October 2026, does not mention artificial intelligence, symbiosis or merging anywhere (the Claims).

What to Watch For

The nearest result that could raise the count to 29 is in resetting cells to turn back ageing. Life Biosciences dosed the first participant in its trial for two diseases of the optic nerve on 9 June 2026, and no result has been reported. The upgrade reaches first human only when a measured sign of age in a person moves back toward youth, so a dose alone does not move it (the upgrade's page).

For a brain link faster than speech, what would count is a rate above 40 bits a second, kept up for at least ten minutes, in a peer-reviewed report on a person (the link-to-machines branch). Paradromics, another brain-implant company, claims more than 200 bits a second in animals and in a paper not yet peer reviewed, and reported its first permanent implant in a person in June 2026, with no rate in a person published yet.

Upgrades already counted can still climb toward clinical use. For organs from gene-edited pigs, trials cleared by the Food and Drug Administration are enrolling patients, and the Boston hospital system Mass General Brigham has said a third transplant is planned later in 2026 (the reading of 16 October). For a brain implant to treat severe depression, the main trial run by the medical device maker Abbott, begun in September 2024, has not yet reported (the same reading).