Brain-Computer Interfaces in 2025: Neuralink's Progress and What the Gap to General Use Looks Like

Neuralink has three human patients implanted with the N1 device as of mid-2025. Noland Arbaugh — the first — demonstrated cursor control through motor imagery alone: playing chess, streaming video, moving a cursor at speeds exceeding standard eye-tracking. That's real. What it tells us about the gap to broader deployment is a different question.

## What N1 Actually Does

The N1 chip carries 1,024 electrodes on 64 threads, each thinner than a human hair. Implanted by the R1 robotic surgical system into the motor cortex, it records local field potentials and spike trains from nearby neurons. Wireless data transmission runs at roughly 1 Mbps — enough for real-time cursor control with millisecond response.

> ⚡ 1,024 electrodes sounds like a lot. The human motor cortex contains roughly 15–20 million neurons. N1 samples a fraction of a percent of that population — but it's the right fraction for intent-decoding in motor impairment applications.

The robotic implantation is itself an engineering achievement. R1 maps surface vasculature in real time and routes threads to avoid blood vessels, reducing the hemorrhage risk that has historically been the primary safety concern in neural implant surgery.

---

## The Risk-Benefit Calculation for Current Targets

Neuralink's indication is ALS and spinal cord injury — patients with severely compromised or absent voluntary motor function. The risk calculation for this group differs fundamentally from enhancement applications.

Surgical risks are real: infection, hemorrhage, electrode migration, and the long-term foreign body response that can degrade signal quality over years. Arbaugh's case included retraction of some electrode threads post-implantation, reducing the functional electrode count. His functional performance remained high despite this, which is actually informative — N1's redundancy appears to tolerate partial thread retraction.

For someone who communicates through eye-tracking at 10–15 words per minute, a device enabling 40+ WPM keyboard control changes daily independence in a measurable way. The risk profile unacceptable for elective enhancement is reasonable for restoring lost function.

---

## Why Non-Invasive Alternatives Can't Close the Gap

EEG-based BCIs are fundamentally limited by physics. Skull and scalp attenuate high-frequency neural signals and spatially blur the signal across centimeters of cortex. EEG captures a weighted average, not individual neuron firing. The practical bandwidth ceiling for EEG intent-decoding is 40–80 bits per second under controlled conditions — which is why EEG-based control feels laggy and degrades rapidly as task complexity increases.

Electrocorticography (ECoG) — electrodes on the cortical surface without penetration — achieves better spatial resolution than EEG without microtrauma from depth electrodes. But ECoG still requires craniotomy. It's not non-invasive; it's less invasive.

There's no non-invasive path to high-bandwidth neural recording with current technology. Transcranial magnetic stimulation and focused ultrasound can modulate neural activity but can't read it at useful bandwidth. This isn't a gap that signal processing algorithms can bridge — it's physics.

---

## What 1,024 Electrodes Is and Isn't

Neuralink's array is the highest-density chronic neural recording device deployed in humans. It's also two to three orders of magnitude below the population-level recording that would be needed for complex applications like sensory feedback restoration or decoding speech directly from motor planning.

The path forward isn't more electrodes alone. It's electrode materials that don't trigger progressive immune responses, encapsulation that survives decades without impedance drift, and processing algorithms that scale with richer signal data. All three remain active engineering problems.

---

## The Bigger Picture

What Neuralink has demonstrated: high-quality chronic neural recording is feasible in ambulatory humans, robotic implantation reduces the surgical risk profile meaningfully, and patients with severe motor impairment can use the system for daily tasks that previously required caregiver assistance.

What it isn't: a proof point for consumer BCI or "merging with AI." Those applications aren't the next step. They require solving the biocompatibility timeline, regulatory frameworks that don't yet exist, and a risk-benefit balance that currently doesn't exist outside therapeutic applications.

The gap between "Neuralink works for ALS patients" and "Neuralink works for healthy users" is an engineering distance that the current results don't yet close.

Brain-Computer Interfaces in 2025: Neuralink's Progress and What the Gap to General Use Looks Like

// COMMENTS

ON THIS PAGE