BCI / 01

BCI /
Reading and writing
to the brain.

Bypass eyes and ears. Read intent from cortex, write signals back. Once medical-only — now an arms race.

13 SLIDES // NEURAL INTERFACES // 2026
02 / THE SIGNAL

~86 billion neurons. We want to listen.

Each neuron fires action potentials at variable rates — millivolt spikes lasting roughly a millisecond. Cognition emerges from the patterns across populations.

Neurons
86,000,000,000
cortical + subcortical
Synapses
~1014
connections per brain
Spike width
~1 ms
action potential
03 / NON-INVASIVE

Listening from outside the skull.

No surgery, no risk — but the bone and scalp blur the signal. You measure populations of millions, not individual neurons.

EEG
Electrodes on scalp
Cheap. ~100 ms temporal, cm spatial. Used in BCI gaming and meditation headsets.
fMRI
Blood-oxygen MRI
~1 mm spatial — but slow (~seconds). Tracks blood flow, not spikes.
MEG
Magnetic fields
Excellent timing. Requires giant shielded room and superconducting sensors.
fNIRS
Near-infrared light
Portable, measures oxygenation through skull. Limited depth.

All four trade fidelity for safety. None will give you a smooth cursor.

04 / INVASIVE

Inside the skull. Closer to the spike.

  • ECoG Grid of electrodes laid on the cortical surface. Used in epilepsy mapping; rich signal without penetrating tissue.
  • Utah arrays ~100 silicon microelectrodes pushed into cortex. Workhorse of academic BCI since the late 1990s.
  • Neuralink threads ~64 flexible polymer threads, 16 electrodes each — placed by a sewing-machine-like robot.
05 / READING THE BRAIN

From thought to cursor. From cortex to words.

2006 // BrainGate
Cursor control
Matt Nagle, paralyzed from the neck down, moves a computer cursor by imagining hand motion. Utah array in motor cortex. The first proof.
2021+ // Stanford / UCSF
Speech decoding
Patients with ALS or stroke "speak" via implants reading premotor cortex. Up to ~78 words / minute by 2023; trending toward conversational pace.

Reading isn't telepathy. It's measuring intent — the motor signals that would have moved a hand or shaped a sound — and translating them into text or motion.

06 / WRITING TO THE BRAIN

Stimulate. The brain interprets.

Send pulses into nerve or cortex, the brain learns to read them as sound, sight, or sensation. Write-mostly BCIs are the oldest commercial neurotech — and the most successful.

Cochlear
~700,000
implant users worldwide. Stimulates auditory nerve directly.
Retinal
Early stage
Argus II, Prima — phosphene-level vision restored. Argus retired 2020; newer systems advancing.
07 / NEURALINK

Neuralink: more channels, robot surgeon.

Founded
2016
Musk + neuroscience team
Channels
1,024+
across 64 polymer threads
First human
Jan 2024
Noland Arbaugh, paralyzed; plays chess by thinking
  • Robot inserts threads thinner than a human hair to avoid blood vessels
  • Wireless, rechargeable, fully under the skull — no protruding hardware
  • Several threads retracted in patient one; software compensated. Real-world feedback in progress.
08 / SYNCHRON

Synchron. No craniotomy.

The Stentrode is a mesh of electrodes placed via blood vessels — threaded up the jugular into the superior sagittal sinus, sitting against motor cortex from inside a vein.

  • No drilling, no skull surgery — placed in a few hours like a stent
  • Lower channel count than Neuralink, but lower risk
  • FDA breakthrough device; multi-patient trials in US + Australia
09 / PRECISION

Precision Neuroscience. A film, not a needle.

Founded by Neuralink co-founder Ben Rapoport. Lays a thin polyimide film with 1,024 microelectrodes onto the cortical surface — through a slit-like incision, no penetration of brain tissue.

Approach
Surface film
Sits on cortex, like a sticker
Channels
~1,024
per film, stackable
Reversible
Removable
Non-penetrating; can be replaced

Trade-off: ECoG-class signal — not single neurons, but huge spatial coverage. The "safe upgrade" pitch.

10 / WHAT WORKS

Medical BCI already changes lives.

Paralysis
Movement + cursor
Tetraplegic patients control robotic arms, computers, and exoskeletons. Some achieve functional independence with daily-use systems.
Speech
Restoring voice
ALS / locked-in patients regain conversational speech via cortex-to-text decoders. Trending past 70 wpm.
Parkinson's DBS
~250,000
people with deep brain stimulators. Tremor and rigidity reduced for decades. Approved 1997.

Also: epilepsy responsive stimulation, treatment-resistant depression trials, tinnitus suppression. The medical track record is real and growing.

11 / SPECULATIVE

The consumer dreams. Mostly unproven.

Memory
Hippocampal boost
Wake Forest experiments show modest recall improvement via stimulation in epilepsy patients. Generalization unclear.
Telepathy
Brain-to-brain
Lab demos route bits between people via TMS. Bandwidth is laughable. Don't expect mind-DMs soon.
AI-cognition
Co-processor
A chip that "thinks with you." No hardware, no signal-shape, no neuroscience yet supports it. Pure pitch.

The gap between "works in a lab on five patients" and "costs $300 at Best Buy" is wider than the marketing suggests.

12 / HONEST

The honest assessment.

Two completely different timelines collapsed into one word.

Real // Now

Medical BCI works. Paralysis, speech, deafness, Parkinson's — measurable wins, multi-decade track record. The science is competitive, well-funded, and shipping.

Decades // Off

Consumer BCI — meaningful enhancement of a healthy brain — is decades away. Surgery risk, signal density, biocompatibility, and FDA pathways all bind. Don't believe the demo reels.

The right framing: BCI is a medical revolution that will eventually spill into consumer use. We are early, and the early stories are the medical ones.

13 / END / REFS
The brain is no longer a sealed box. We are listening in — and beginning to talk back.
YOUTUBE SEARCH

Companies cited: Neuralink, Synchron, Precision Neuroscience, BrainGate consortium, Medtronic (DBS), Cochlear Ltd. Decoding speech: UCSF / Stanford labs (Chang, Henderson, Shenoy).

01/13