Neuralink Details Telepathy and Blindsight: The Numbers Behind the Pitch

Neuralink's motor implant Telepathy carries 1,024 biocompatible electrodes on 64 flexible polyimide threads, each thread thinner than a human hair — and quadriplegic participants in the PRIME Study are already using it to move cursors, type, play video games and drive robotic arms by imagining movement alone.

That is the concrete core of the announcement Elon Musk's neurotechnology firm has now laid out across its two flagship brain-computer interface programs, Telepathy and Blindsight. The company positions both as direct bridges between human neural tissue and external machines — one for lost motor control, one for lost sight.

How Telepathy records the brain

The device captures real-time electrical activity from the cortex through its thread-and-electrode array. Custom ASIC chips amplify, digitise and wirelessly stream the neural firing patterns to an external receiver at low latency. The thin polyimide threads are a deliberate design choice: minimising long-term tissue inflammation and surgical micro-trauma is the whole point of their geometry.

The target population is specific. Telepathy is intended for people who have lost limb function — severe spinal cord injury, amyotrophic lateral sclerosis — and it grants control of computers, smartphones and physical peripherals purely through intent. Beyond cursor control and typing, Neuralink claims an ambition of output bandwidth exceeding human speech or typing speeds, using pretrained neural encoders trained on tens of thousands of hours of clinical brain data. That last figure is a company claim about training data, not a demonstrated result, and deserves the distinction.

What stage is Telepathy actually at?

Human testing, under the PRIME Study — Precision Robotically Implanted Brain-Computer Interface. Participants with quadriplegia have received the N1 implant to test cursor control, typing and gaming. The trials now measure safety, wireless connectivity, real-time decoding accuracy, and how the thread electrodes hold up inside brain tissue across multi-year periods. This is investigational work in humans, not a cleared device.

How does Blindsight propose to restore vision?

Blindsight takes a different route: it bypasses the eyes entirely. The design covers complete sight loss, including severed optic nerves, ocular trauma and congenital blindness.

• External cameras capture high-definition imagery of the environment.
• Processing hardware translates light, depth and edge data into custom spatial electrical frequencies.
• Micro-electrode arrays implanted in the primary visual cortex (V1) deliver small pulses.
• Those pulses evoke phosphenes — individual points of light — and driving thousands simultaneously builds a real-time, low-resolution visual map.

Neuralink concedes the initial experience will resemble coarse pixel matrices. Future iterations, it says, could scale electrode density toward resolutions exceeding natural human sight, with possible overlay modes in infrared or radar spectra. A generous promise. The present data are pre-clinical, and the testing roadmap runs through non-human primate models before initial human trials of cortical micro-stimulation and basic phosphene mapping.

One regulatory fact stands apart: the US FDA has granted Blindsight Breakthrough Device Designation, an administrative status meant to speed development and review of life-changing technologies. It is not an approval.

The surgical robot does the delicate part

Both implants depend on the R1 Surgical Robot, built to insert ultra-thin flexible threads into brain tissue without rupturing blood vessels. The R1 uses optical coherence tomography and high-speed computer vision to map surface capillaries, then weaves the thread needles around micro-vessels to limit bleeding and tissue damage. Its throughput: up to six threads — 192 electrodes — per minute.

What does this change for the reading room and the monitoring unit today? Very little, directly. These are investigational implants under sponsor-run trials, not clinical tools. But the specs matter to anyone tracking implantable ECoG-scale recording: the electrode counts, the multi-year stability endpoints and the automated placement of sub-hair-width threads define the engineering benchmark that cranial neurophysiology will be measured against.

The next milestones are written in the trials themselves — decoding stability over years for Telepathy, and first-in-human phosphene mapping for Blindsight.