# The Intent Layer, a manifesto from Setu

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

> Machines learned to understand us faster than we learned to reach them. BCI is a few years from its ChatGPT moment. This is why, and what Setu is building.

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

The Setu vision

# The Intent *Layer*

Machines learned to understand us faster than we learned to reach them. We think BCI is a few years from its ChatGPT moment. This is why, what it will take, and what Setu is building.

- Short version: 2 minutes

- Full version: 32 minutes

- 19 figures

- 62 sources

- Written by the team at Setu

[Image: Two wavering ink traces on a strip of paper: the first recorded human electroencephalogram above a regular timing wave.]

*The first recording of electrical activity from a human brain, made by Hans Berger in 1924. Upper trace: the brain. Lower trace: a 10 Hz timing signal. A century later this is still the signal. Image: [Hans Berger, public domain](https://commons.wikimedia.org/wiki/File:1st-eeg.png).*

The short version · 2 minutes

## TL;DR

1.  I

### The channel

Every computer still waits for a human body to type, tap or speak. That wait is now the slowest part of the system.
 [Read →](https://setu-labs.co/manifesto-1.html)

2.  II

### The signals

Intent leaves physical traces from brain to muscle. Laboratories have already decoded handwriting, speech and images from them. EEG is the one sensor everyone could wear.
 [Read →](https://setu-labs.co/manifesto-2.html)

3.  III

### The distance

Decoding in a laboratory is not a product. The body is a superb incumbent, so the first wins come from silence, privacy and free hands, not from speed.
 [Read →](https://setu-labs.co/manifesto-3.html)

4.  IV

### The learning problem

Hardware sets the ceiling, data is the fuel, and the winner is the system that learns each person fastest. That is why we bet on EEG, on scale, and on building the full stack.

[Diagram: A pyramid of four layers: products at the top, then algorithms, data, and hardware at the base. Attention goes to the top; the real limit is the hardware at the base.]

 [Read →](https://setu-labs.co/manifesto-4.html)

5.  V

### The source: data and alignment

Text, code and images are compressed products of thought. Neural data is the source. AI labs need it now, as training data and as a faster way to align their models. Those are the first two wedges.
 [Read →](https://setu-labs.co/manifesto-5.html)

6.  VI

### The road

BCI reaches the world through two things people will actually wear: silent communication, and brain wellness and recovery. Its ChatGPT moment is a few years away.
 [Read →](https://setu-labs.co/manifesto-6.html)

7.  VII

### Agency

None of this is worth building unless it can be switched off. Off by default, on the device, never for advertising, and the user always has the last word.
 [Read →](https://setu-labs.co/manifesto-7.html)

 [Begin with Part I →](https://setu-labs.co/manifesto-1.html)