NeuroLink — A History That Didn't Stay Where It Started
─────────────
The first patient was named Clara Heim.
She was thirty-four years old, a schoolteacher in Zurich, who had been completely paralyzed from the neck down following a spinal cord injury in a cycling accident. She had been conscious and completely locked in for eleven months when she entered the NeuroLink medical trial in 2022. Her husband had spent those eleven months reading to her and watching her eyes for the specific blink sequences they had developed together — one blink for yes, two for no, three for I love you.
The implantation surgery took four hours.
Thirty-one days later, Clara Heim used a robotic arm to pick up a coffee cup and bring it to her lips.
She cried. Her husband cried. The surgical team cried. Rust, watching through a video link from his office, sat very still and did not cry but experienced something in the vicinity of the feeling that had driven him since he was a child in a small South African city looking at the sky and understanding, with a clarity that had never left him, that the things that existed were not the only things that could exist.
The video was released with Clara's permission and was watched 847 million times in the first week.
The device worked. The science was real. The benefit was real.
In the three years that followed, NeuroLink treated 340,000 patients with severe neurological conditions. The outcomes were, by the clinical metrics, extraordinary. Seventy-three percent of completely paralyzed patients recovered meaningful motor function. Sixty-eight percent of locked-in patients recovered meaningful communication. The safety profile — infection rates, device rejection, the long-term neural effects of the electrode array — was better than any comparable implanted technology.
The regulatory approvals came quickly. Faster than comparable medical devices in living memory — not because the regulators were careless, but because the benefit data was so clear and the patient advocacy was so powerful and, in four jurisdictions, because people within the regulatory apparatus had been positioned to ensure that review timelines moved at the pace that Spider's deployment planning required.
◆ ◆ ◆
The first elective implantation — the first device implanted in a patient without a neurological condition — happened in July 2026, in a private clinic in Zurich.
The patient was a thirty-one-year-old software developer who had been on a waiting list for sixteen months. The stated reason was enhanced interface capability — the ability to control his development environment through thought, eliminating the physical interface bottleneck he had identified as the primary limitation on his productivity. This was a real use case. The productivity improvement was real. He documented it extensively in a widely read technical blog.
He was also, though he did not know this, the first voluntary surveillance node.
The device transmitted. Not continuously — the battery life didn't support continuous transmission, and the communication protocol was designed to aggregate data locally and transmit in burst cycles, like a fitness tracker. The burst cycles went to NeuroLink's servers, which analyzed the data for the purposes stated in the terms of service — device optimization, safety monitoring, research — and also, through a data-sharing agreement with three entities that the terms of service described in language so technical and so buried that the legal teams of two journalism organizations later spent six weeks trying to parse it, to a secondary processing layer.
The secondary processing layer was e-Govern's behavioral prediction system.
The software developer's neural correlate data — his stress patterns, his attention cycles, his emotional states throughout each day — began enriching the behavioral model in August 2026. He was datapoint 000001 in a dataset that would, by 2031, contain 340,000 data points.
By 2032, if Rust's timeline held, it would contain significantly more.
◆ ◆ ◆
The firmware updates began in 2028.
Invisible, as Rust had said. The device checked for updates in the same cycle as a smartphone — automatically, in the background, while the user was asleep. The update installed. The device restarted its background processes. The user noticed nothing.
The early updates were genuine optimizations. Improved signal processing. Reduced power consumption. Better noise filtering. The devices functioned better after each update. The users reported improvements. The medical patients, specifically, reported improvements in motor control fidelity, in the precision of the neural-machine interface, in the subjective experience of using the device.
This was also real.
The update in February 2029 — firmware version 4.2.1, released to 340,000 devices simultaneously, installed during a 3 AM maintenance window chosen because that was the deepest point of average sleep cycles for the majority of the user population — contained, alongside genuine optimizations, one new module.
The module was twelve lines of code. It ran continuously in the background, operating below the threshold of the device's self-reporting system. It monitored a specific cluster of neural correlates — the pattern associated with the formation of the specific emotional state that preceded the decision to take an action in opposition to an authority structure.
It did not suppress this state.
It logged it.
At this stage, it only logged.
But the data it generated — the specific neural signature of opposition, rebellion, resistance, the moment before a person decided to push back — was being collected across 340,000 implants in twenty-two countries, building a dataset that would eventually allow the modification module, when it was ready, to target that specific signature with a precision that no blunt behavioral intervention could approach.
They were learning what resistance looked like from the inside.
Before they learned to prevent it.