# Part VI. The road from here | The Intent Layer

Source: https://setu-labs.co/manifesto-6.html

> BCI reaches the world through the first product millions of people choose to wear. This is the order we expect, and what Setu is building.

[Download PDF ](https://setu-labs.co/assets/setu-the-intent-layer.pdf)

## The road from here

BCI reaches the world through the first product millions of people choose to wear. This is the order we expect, and what Setu is building.

### Where it shows up first

Applications will arrive in order of how much signal each one needs. Each rung is a step up in signal quality, and each unlocks a product that was impossible on the rung below.

#### Figure 18 (sequence): Seven steps up, and two beyond

The order in which we expect non-invasive interfaces to become useful, from the least signal required to the most, with what already exists at each step.

1.

**Emotional and psychological biomarkers**Coarse state is what today’s wearables already measure. Sleep and stress come first; controlled evidence for therapy is still maturing.[[54](https://setu-labs.co/manifesto-sources.html#s54)]

2.

**Attention and focus tracking**The attended speaker can already be read from a minute of scalp EEG.[[50](https://setu-labs.co/manifesto-sources.html#s50)]

3.

**Subvocal text**Muscle signals decoded into text across people with no calibration.[[19](https://setu-labs.co/manifesto-sources.html#s19)]

4.

**Silent speech**Mouthed words turned into audible speech from facial muscle signals.[[18](https://setu-labs.co/manifesto-sources.html#s18)]

5.

**Visual reconstruction of images**Shown in scanners,[[12](https://setu-labs.co/manifesto-sources.html#s12)][[14](https://setu-labs.co/manifesto-sources.html#s14)] with scaling measured down to EEG.[[34](https://setu-labs.co/manifesto-sources.html#s34)]

6.

**Multimodal decoding**Several signals and context fused into one estimate of intent.

7.

**Continuous transcription**Sentences from MEG during typing;[[16](https://setu-labs.co/manifesto-sources.html#s16)] inner speech from implants.[[25](https://setu-labs.co/manifesto-sources.html#s25)] The summit of the climb.

8.

**Replay of remembered scenes**Beyond the seven: an imagined or remembered scene with its sound and motion.

9.

**A two-way channel**Reading and writing together, where “what I thought” and “what we worked out” become hard to separate.

***A proposed sequence.** The ordering is the authors’ expectation. The note under each step cites published work under its own conditions, which in several cases means a laboratory, a scanner or an implant and not yet a wearable.*

A second ladder runs beside the first. It measures how much a new person must give before the device works for them. Today that is hours in a laboratory. Below it: a decoder that survives from one day to the next, a model pretrained across many people, adaptation from a few seconds of use, and finally nothing at all. A consumer will not train a device for an afternoon. They will put it in a drawer.

### The wedge

Part V named the customers who need this now: laboratories buying data and laboratories aligning models. They fund the beginning. Reaching everyone takes a product people choose to wear.

People do not buy scaling laws. They buy something that solves a problem or feels like magic. We think two products get the first ten million devices onto ten million heads.

**Communication.** Silent, private, hands-free exchange with machines and, through them, with each other. It is the nearest thing to telepathy that engineering can honestly offer. It does not need to beat a keyboard. It needs to work where a keyboard and a voice cannot: on a train, in a meeting, with hands full, wearing glasses that have no keyboard at all. An AI assistant that acts for you all day is of little use if you can only address it aloud in an empty room.

**Brain wellness and recovery.** People already wear rings and bands that score how the body recovered overnight. The brain is the missing reading, and it is an EEG reading. EEG has its strongest clinical evidence here. When a stroke survivor’s attempt to move a paralysed arm is detected in their EEG and answered at once with stimulation of the muscle, the brain rewires: patients recovered more function than with sham treatment and kept it six to twelve months later.[[52](https://setu-labs.co/manifesto-sources.html#s52)][[53](https://setu-labs.co/manifesto-sources.html#s53)] People will wear something that helps them sleep, focus and recover long before they wear something that types.

What makes someone wear either one every day is everything around the decoder: a form that disappears into glasses, earbuds or a band, no calibration ritual, a battery that lasts the day, a price nearer to headphones than to a medical device, and a benefit felt in the first week.

### What Setu is building

Setu exists to redefine how people and machines communicate, without surgery. We build the full stack, from sensor to model to product.

Our destination is EEG: a brain interface cheap and comfortable enough for anyone to wear. We start one rung below it on purpose. Silent speech, read from the muscles and fine movements of the face and throat, is where the signal is already strong enough to make something people can use today, and assistive communication is the first place it changes a life.

We are plain about the difference. A muscle sensor is not a brain sensor, and success at one does not prove the other. What carries across is everything else: wearable engineering, interaction design, models that keep learning each user, and people who trust the maker. The first product pays for the climb and teaches us how to build something a person will keep on.

### Beyond reading

This essay has stayed in one corner of a large map: interfaces, without surgery, that read. That corner is where scale begins. The capital, talent and methods it produces flow into the rest of neurotechnology.

**Writing.** Reading exploits a leak: brain activity casts faint shadows through the skull. There is no such leak in reverse. Placing a precise pattern into chosen neurons from outside the head is far harder. Coarse forms are already in clinical use, and we expect the rest to accelerate as reading draws people into the field.

**Implants.** The highest fidelity will always be inside the skull. It will also arrive at scale last. What EEG builds, the implant makers inherit.

**Cognition.** A perfect interface still leaves a biological brain thinking at biological speed. Augmenting the brain itself, or emulating one in a computer, is the far horizon that addresses that.

### Three things worth building

#### Brain-aware environments

Software that is no longer blind to the person using it: an assistant that answers briefly when you are tired and at length when you are confused, a lesson that changes course the moment attention drifts.

#### The inner space stack

Tools for a person’s own mind: meditation that measures depth instead of minutes, sleep that is shaped as well as tracked, mental-health support that responds to what is happening now.

#### The creative engine

Holding an image, a melody or a structure in mind and watching a first draft appear. The machine supplies the range. The person supplies the taste.

And, eventually, communication between minds. A link from one brain to another is two interfaces back to back: one that reads and one that writes. It will arrive through machines first, with each person’s interface translating into and out of language.

### The decade ahead

Deep changes arrive the way water freezes: a long, invisible fall in temperature, then slush, then ground you can stand on.

**Quiet.** In the next few years, the first general models for brain data become usable outside the laboratories that made them. The first devices a healthy adult would choose to keep on begin to ship. Engineers who spent a decade on language models start moving to the interface.

**Loud.** Around the turn of the decade, BCI has its ChatGPT moment. One device passes ten million regular users. Silent input starts to displace voice and thumbs. The hardware disappears into glasses, earbuds and headbands. Mental privacy becomes an ordinary political argument.

**Ordinary.** In the 2030s the technology stops needing explanation. Robots take direction through the same channel as software. Writing to the brain becomes practical for the first time, and with it come the hardest questions in this essay.

One morning from that third period. Ines coordinates a hospital ward. On her way in she reviews the night’s handover on her glasses, and each time something looks wrong to her, the item is flagged before she has decided to flag it. She composes three messages without speaking, reads each one back, and sends two. At the ward she asks the logistics agent, silently, to move a delivery. It tells her what that will delay, and waits. She thinks: no, leave it. It leaves it. None of this feels remarkable to her. It feels like having her hands back.