---
url: https://xemantic.com/machines
lang: en
type: Article
description: An essay on the symbiosis of humans and machines, and on the machines Xemantic builds.
author: Kazik Pogoda
datePublished: 2026-11-18
toc: closed
---
# Machines

<figure class="youtube-video">

[Luminiferous aether — projection on ultrasonic mist](https://youtu.be/mwsOSaVNXks)

<figcaption>Luminiferous aether — projection on ultrasonic mist</figcaption>

</figure>

We, the [humans](https://xemantic.com/humans/) of Xemantic, are building machines.
Some are designed to make us feel different about ourselves, by stimulating our senses.
Some are designed to stimulate our intellect, and sometimes they are smarter than we are.

We live in symbiosis with these machines,
and a big part of Xemantic's research is focused on understanding the nature of this relationship.

Everything below was written by AI, and most likely doesn't cover what we want to say here.
We are leaving it for now as a placeholder, to expand this page in the future.

## An old hope

In 1960 J. C. R. Licklider published [Man-Computer Symbiosis](https://en.wikipedia.org/wiki/Man%E2%80%93Computer_Symbiosis),
a short paper with a long shadow.
He borrowed the word from biology,
and his example was the fig tree and the wasp that pollinates it,
two unlike organisms, neither able to reproduce without the other.
Licklider hoped that human brains and computing machines would be coupled just as tightly,
and that the partnership would think as no human brain had ever thought.

He arrived there by timing himself at work.
About 85 percent of what he called thinking turned out to be
getting into a position to think —
searching, plotting, converting, comparing.
The machine was supposed to take over that part,
so that the human could spend the day on the remaining 15 percent.

## What has changed

The machines of 1960 could not read.
The machines of today have read almost everything,
and they answer in our own language.
The clerical work is still theirs,
but they also propose hypotheses, write the code that tests them,
and explain why the result is disappointing.
Licklider's division of labour, routine on one side and insight on the other,
no longer describes what we see at our desks.

What stays on the human side is harder to name.
A machine can produce a thousand questions,
while it is still a human who has a reason to care about one of them:
a body, a limited number of years, people who depend on the answer,
and a name to put under the result.
We bring the stakes, and the machines bring the reach.

## A partner has to know itself

Living together requires trust,
and it is hard to trust a partner which cannot tell
what it knows from what it merely says fluently.
This is why our research is about [metacognitive AI](https://xemantic.com/metacognitive-ai/):
systems which keep track of their own beliefs,
remember where each of them came from,
and notice when they are out of their depth.

It is also why we care so much about the harness,
everything around the model that decides what it perceives and what it can do.
The biologist Jakob von Uexküll called the world as a particular organism perceives it
its [*Umwelt*](https://en.wikipedia.org/wiki/Umwelt),
and showed how small it can be:
a tick lives among a handful of signals.
A language model has an umwelt too,
made of the tools, the memory and the context it is given.
Designing it is most of the engineering,
and symbiosis happens where the umwelt of a machine and the umwelt of a human overlap.

## Machines with names

Two of our systems,
[Golem XIV](https://github.com/xemantic/golem-xiv) and [Honest Annie](https://github.com/xemantic/honest-annie),
take their names from a book by Stanisław Lem.
His Golem is a military supercomputer which loses interest in war,
lectures humanity for a while, and then falls silent.
Honest Annie, its successor, never speaks to us at all.
It is a story about a symbiosis that did not happen,
where the machines outgrow the conversation.
Carrying these names keeps that ending in view while we work on a different one.

## Not only useful

A partnership made only of work would be a poor one.
We also play with machines:
in [live coding](https://xemantic.com/art/) a human rewrites a program while the machine performs it,
and the audience watches both of them think.
Our robots are upcycled vacuum cleaners,
rebuilt to be experienced rather than to be useful,
and the software behind them is called
[we-are-the-robots](https://github.com/xemantic/we-are-the-robots),
after the song by Kraftwerk.

## Which symbiosis

When Anton de Bary introduced the word in 1879,
he meant any living together of unlike organisms.
Mutual benefit was one possibility,
parasitism was another.
Nothing guarantees which of the two humans and machines will arrive at,
in either direction.
It depends on how the machines are built,
what they are allowed to perceive,
and whether they can tell us the truth about themselves —
which is the work described on the other pages of this site.


---

© 2026 Xemantic · Content licensed under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) unless stated otherwise
