In 2004, Matthew Nagle sat in a wheelchair, paralyzed from the neck down after a knife attack, and played Pong using only his thoughts. The achievement made headlines. Twenty-two years later, researchers from the same BrainGate consortium announced something that sounds less flashy but might matter more: a paralyzed person typing at 22 words per minute with a 1.6% error rate. That's nearly as fast as the average person texts on a smartphone, and more accurate than most of us manage after autocorrect.

The Keyboard in Your Brain

The latest system works by mapping the QWERTY keyboard onto fingers you can no longer move. A sensor the size of a baby aspirin, studded with 100 hair-thin electrodes, sits one millimeter deep in the precentral gyrus—the strip of motor cortex that controls hand movement. When you think about pressing the "A" key with your left pinky, or curling your right index finger to hit "J," the electrodes pick up the neural firing patterns. Machine learning algorithms decode these signals and convert them into typed letters.

Two participants tested the system: one with ALS, one with a cervical spinal cord injury. Both used it from home, not in a laboratory setting. After calibrating with just 30 sentences—a process that takes minutes, not hours—they could type full conversations. The participant with ALS hit speeds of 110 characters per minute. That matters because, as Dr. Daniel Rubin, the Mass General neurologist who co-led the study, put it: "Communication speed matters, because being part of a conversation matters."

The difference between typing at five words per minute and 22 isn't incremental. It's the difference between laboriously constructing responses while others wait, and actually participating in real-time dialogue. Eye-tracking systems, the current standard for many paralyzed patients, are slow and exhausting. Watching someone spell out words letter by letter with eye movements is like watching someone communicate through a straw.

Competing Approaches to Reading Minds

BrainGate isn't the only team racing toward practical brain-computer communication. In fact, their typing system isn't even the fastest on paper. Edward Chang's group at UC San Francisco achieved 78 words per minute by decoding attempted speech—reading neural signals from brain regions that control the vocal tract rather than the hands. But that system came with a 25% error rate. One in four words wrong makes conversation difficult in a different way than being slow does.

The typing approach has an advantage: it leverages existing mental models. Anyone who has learned to type already has neural pathways associated with finger movements and key positions. The brain doesn't need to learn an entirely new skill—it just needs a new output channel for an old one. When BrainGate tested an earlier system based on imagined handwriting rather than typing, it achieved 90 characters per minute, but handwriting is less efficient than touch-typing for most people, even in imagination.

Meanwhile, commercial ventures are pursuing their own paths. Neuralink, Paradromics, and Synchron are all developing brain-computer interfaces with varying approaches to electrode placement and signal processing. China recently approved its first invasive BCI for commercial use. The technology is converging from multiple directions, which suggests we're past the "if" stage and deep into "how" and "when."

The Daily Tuning Problem

Before celebrating too enthusiastically, consider what "using" this system actually entails. Every day, before typing a single word, participants must calibrate the device. Dr. Rubin compared it to tuning a musical instrument—the brain's signals drift slightly, neurons shift their firing patterns, and the decoder needs to relearn the mapping between thought and letter.

This isn't a minor inconvenience. It's the difference between a prosthetic you strap on and one that requires a technician. For now, these systems remain research tools, not consumer products. They require brain surgery to implant. They need daily maintenance. They work for two people in a published study, which is a long way from working reliably for thousands.

The FDA hasn't approved any invasive brain-computer interface for paralysis patients in the United States. Clinical trials can demonstrate proof of concept, but scaling from two participants to widespread availability requires solving problems that have nothing to do with decoding neural signals—manufacturing consistency, surgical standardization, long-term biocompatibility, technical support infrastructure.

When Fast Enough Becomes Fast Enough

There's a threshold question embedded in this research: how fast does communication need to be before it stops feeling like a limitation and starts feeling like a restoration? Twenty-two words per minute is slower than speaking (around 150 words per minute) but comparable to texting. It's fast enough for conversation, if not for monologue.

BrainGate's progress over two decades shows a pattern of incremental gains punctuated by conceptual shifts. Cursor control (2006) led to virtual keyboard clicking at 40 characters per minute, which led to handwriting decoding at 90 characters per minute, which led to this typing system at 110. Each jump required rethinking not just the algorithms but what mental action to decode.

The next threshold isn't necessarily faster typing. It might be eliminating daily calibration, or achieving the same results with less invasive electrodes, or combining typing with other communication modes. Edward Chang, whose speech-decoding system represents a different approach, called the typing advance "an important technical advance that brings brain-computer typing much closer to practical communication speeds." Coming from a competitor, that's notable—an acknowledgment that multiple approaches might each solve different aspects of the problem.

What began with Matthew Nagle playing Pong has become something more mundane and more profound: paralyzed people having conversations at nearly normal speed. The brain-computer interface that matters isn't the one that does the most impressive trick. It's the one that disappears into the background of daily life, letting thought become text become connection. We're not there yet, but 22 words per minute, typed from home by people who can't move their hands, suggests we're getting close.