Somebody at the table mentions Neuralink, or Synchron, or “that company threading wires into people’s brains,” and the conversation stalls. Not because the topic is boring. Because most of the vocabulary sounds like it was airlifted in from a physics lab. Electrode array. Latency. Closed-loop stimulation. Nobody hands you a glossary before you’re expected to have an opinion on brain implants.
So here’s the glossary. Twelve terms, no filler, organized the way an engineer would actually explain them to you over drinks: what’s in the hardware, how a signal gets read, why the jargon around signal quality matters, and why almost none of this is available for purchase yet, however loudly a press release insists otherwise. I’ll flag development stage on everything, because conflating a lab demo with a shipping product is the fastest way to sound like you read one headline and stopped there. 🧠
The hardware, bolted to (or near) your skull
Start with the umbrella term, because everything else nests inside it.
A brain-computer interface (BCI) is any system that reads electrical activity from the brain and turns it into a command a machine can act on, sometimes in the other direction too, sending signals back in. That’s it. That’s the whole definition. What trips people up is treating every BCI as the same kind of project, when medical, consumer, and military applications are pursuing genuinely different goals with different risk tolerances. A medical BCI is trying to restore lost function, like giving someone with paralysis control of a cursor. A consumer BCI is selling you a headband that tracks focus. A military one, like DARPA’s N3 program, is trying to build a non-surgical, wearable interface that lets a soldier control drones or cyber-defense systems hands-free. Treating those as interchangeable is, frankly, the single fastest way to lose credibility with anyone who actually works in this field. 🔬
The electrode array is the physical sensor grid doing the listening (or, in stimulation devices, the talking). 🔌 Specs vary wildly by design goal:
Neuralink’s N1 implant packs 1,024 electrodes across 64 flexible threads, according to the company’s own published specs
Precision Neuroscience’s Layer 7 Cortical Interface also runs 1,024 microelectrodes, but embedded in a film thinner than a human hair that rests on the brain’s surface instead of piercing it
Synchron’s Stentrode uses a far leaner 16 electrodes, trading raw channel count for a delivery method that avoids cutting into the skull at all
More electrodes generally means richer, higher-resolution signal. It also generally means more surgical risk. That trade-off is the whole ballgame, and it’s why invasive vs. non-invasive is the first fork in the road for any BCI design. Invasive systems penetrate or sit directly on brain tissue for the cleanest possible signal. Non-invasive systems stay outside the skull entirely, sacrificing resolution for safety and, not incidentally, for a regulatory pathway that doesn’t require anyone to open your head. Every other term in this list is really just a variation on where a given device lands on that spectrum. ⚡
Quick gut check before we move on: when you hear “device,” ask yourself whether it’s implanted, resting on the brain’s surface, or strapped to someone’s head. That single question resolves most of the confusion in this field.
How the signal actually gets read
This is where the three big recording categories live, and where the “is this a real product” question gets sharpest.
Electroencephalography (EEG) is the oldest, cheapest, and most non-invasive option on this list: electrodes on the scalp, picking up the electrical noise generated by millions of neurons firing at once underneath. It’s been a clinical staple since the 1920s, and it’s exactly what’s inside most consumer neurotech you can actually buy today, things like focus-tracking headbands. NeurotechMag’s rundown of consumer devices you can buy right now covers the Neurosity Crown, a headset that uses EEG to flag cognitive load during work sessions. The catch: EEG’s spatial resolution is genuinely poor, because the skull smears and scatters the signal before it ever reaches the electrode. Great for broad mental states. Useless for controlling a robotic arm with any precision. 📡
Electrocorticography (ECoG) 🔎 solves the resolution problem by skipping the skull entirely: electrodes sit directly on the cortical surface, under the skull but not penetrating brain tissue. Precision Neuroscience’s Layer 7 device is the clearest current example, and its regulatory status is worth being precise about. The FDA granted it 510(k) clearance in April 2025, but only for implantation durations of up to 30 days, and it’s already been used in 37 patients at institutions including Mount Sinai and the University of Pennsylvania. That’s real and commercially authorized. It is not the same thing as a permanent, take-home BCI, which the company itself says is still in development. Precise language matters here more than almost anywhere else in this glossary.
Then there’s endovascular BCI, a genuinely clever workaround pioneered by Synchron. 🩺 Instead of any craniotomy, the Stentrode device is delivered through a catheter via the jugular vein and lodged in a blood vessel next to the motor cortex, more or less the same technique used to place a cardiac stent. No skull opening, no robotic drilling rig. Synchron’s COMMAND study became the first FDA-approved investigational device exemption (IDE) trial of a permanently implanted BCI, and the company raised $200 million in a Series D round in November 2025 specifically to fund the decisive 2026 trial aimed at the first premarket approval (PMA) for an implantable BCI, per reporting on the funding. Aimed at. Not there yet.
Stacked side by side, the three recording approaches break down like this:
EEG: fully external, cheapest, lowest resolution, already sold as consumer wearables
ECoG: sits on the cortex under the skull, higher resolution, currently cleared only for short-term surgical use
Endovascular BCI: delivered through a blood vessel, no craniotomy, still in late-stage trials with no product on the market
Nobody, as of this writing, has an FDA-approved, commercially available, permanently implanted BCI for paralysis. Not Synchron, not Neuralink, not Precision. Anyone who tells you otherwise at your dinner table is wrong, and now you can say so.
The signal-processing jargon you’ll actually hear
Once you’ve got a raw signal, three words come up constantly, and they’re worth having straight.
Spike sorting is the computational grunt work of figuring out which electrical “spike,” a burst caused by a single neuron firing, actually came from which specific neuron. 🔬 An electrode usually picks up overlapping noise from several nearby cells at once, so software has to untangle the mess before any of it becomes a usable command. It’s unglamorous. It’s also the reason a “1,024-electrode array” doesn’t automatically mean 1,024 clean channels of data; the decoding algorithm has to earn that number.
Latency is the delay between a neuron firing and the system responding to it. ⏱️ Low latency is what makes a BCI feel like an extension of intent rather than a laggy remote control you’re fighting with. High latency is the difference between “I thought about moving the cursor and it moved” and “I thought about moving the cursor, waited, and eventually it moved, sort of.”
Bandwidth 📶 is how much information the whole system can move per second, a function of electrode count, sampling rate, and how good the decoder is at turning raw spikes into meaningful output. This is exactly why the electrode-count gap matters in practice:
A 1,024-channel array like Layer 7 or the Neuralink N1 has, in principle, far more bandwidth to work with
Synchron’s 16-electrode Stentrode deliberately caps its bandwidth in exchange for that no-craniotomy delivery method
More bandwidth generally supports finer motor control, like individual finger movements, rather than broad cursor gestures
Neither approach is objectively “better.” They’re optimizing for different things, safety versus signal richness, and that’s a trade-off, not a hierarchy. Something to actually think about next time someone frames this as a simple electrode-count arms race. 💡
The brain’s side of the deal, and what happens when you push back on it
BCIs aren’t a one-way street. Two more terms round out how the brain participates in, and responds to, all this hardware.
Neuroplasticity is the brain’s capacity to reorganize its own connections in response to experience, injury, or repeated input. 🧬 It’s why BCI users tend to get better at controlling a device over time, and why the devices often get better at reading them right back. Both sides are adapting simultaneously, which is part of why early clinical trial data can look modest and later data can look dramatically better with the same hardware. You can read more on the mechanism itself at Wikipedia’s overview of neuroplasticity, which is a reasonably solid starting point if you want the underlying neuroscience rather than the product angle.
Optogenetics 💡 is a research technique, not a product, and it’s worth being blunt about that distinction. It involves genetically engineering specific neurons to respond to light, then using that light to switch those neurons on or off with real precision. It’s an extraordinarily powerful tool for basic neuroscience research in animal models, and it has taught researchers an enormous amount about how neural circuits actually work. It is not implanted in humans as a treatment, and it is not on a path to becoming a consumer or even near-term clinical product. If you see “optogenetic” attached to a human-ready headline, be skeptical. Check Wikipedia’s entry on the technique if you want the full picture of where it’s actually used.
Which brings us to the last term, and probably the most misunderstood one on this list: closed-loop stimulation. Most stimulation devices you’ve heard of, including classic deep brain stimulation (DBS) for Parkinson’s, run open-loop: they deliver a constant, pre-set pattern of electrical pulses regardless of what the brain is doing in that moment. A true closed-loop system reads brain activity in real time and only stimulates in response to a detected pattern. NeuroPace’s RNS System is the clearest example, and the only one with full FDA approval for this design: it’s built specifically to detect the electrical signature of an oncoming seizure and respond with targeted stimulation before symptoms hit. Researchers are now testing the same responsive approach for treatment-resistant depression, and NeurotechMag has covered how neurostimulation is already reshaping depression care for patients who’ve exhausted medication options. Worth being precise, though: the closed-loop depression trials, including one called PReSiDio, are still recruiting patients. That’s promising early-stage research, not an approved therapy.
The three flavors of stimulation, in order of how common they are today:
Open-loop: a fixed, constant pattern of stimulation regardless of what the brain is doing, the design behind classic DBS for Parkinson’s and essential tremor
Closed-loop or responsive: reads brain activity first, stimulates only when a specific pattern is detected, the design behind NeuroPace’s RNS System for epilepsy
Adaptive: a newer, still largely experimental variant that continuously tunes stimulation strength in real time rather than just switching on or off 🚀
Which of these will actually show up in your feed this year
Not all twelve terms carry equal weight for how soon you’ll see them in the news. 📈 A rough cheat sheet, if you want to sound calibrated rather than just informed:
👀 Watch endovascular BCI and ECoG closely; both have active 2026 trial milestones with real regulatory deadlines attached
Expect EEG to keep showing up in consumer gadget reviews, not medical headlines
Treat optogenetics as a research story, always, no matter how the headline is written
Use invasive vs. non-invasive as your default filter whenever a new device gets announced
File closed-loop stimulation claims for depression under “promising trial,” not “available treatment,” until a company says otherwise
The gap between what’s cleared for a 30-day surgical procedure and what’s cleared for you to take home is where most of the hype in this space quietly lives. 🧠 Learn to ask “which stage is this actually at?” and you’ll out-argue most of the people talking about brain implants online.
What’s the term on this list you’ve seen misused the most? And which of these twelve do you think actually reaches your local pharmacy first?