#2026-M05 --- I keep trying to develop a research line of my own, *"electronic organisms" inside natural ecosystems.* The aim is to build some theoretical frame for how to take my interactive objects outdoors, and how such a machine can work at all. Although it's still early, I'm trying to organise it into a structure and iterate on it step by step. *Hopefully it won't look like I'm romanticising machines (again)... *🤖 ## Technology in nature and the usefulness of machines The very idea of why we should put technology into the landscape at all is a bit naive. **But technology has long been out there in nature, we just call it infrastructure** *(transmitters, probes, high-voltage lines, or night lighting).* So the question isn't whether technology belongs in the landscape. It's already there. It's more about *what exactly does technology add out there?* And that's exactly why, for a long time, I looked for various places where a machine could even directly help an ecosystem. ![[een.jpg]] But the deeper I went, the clearer it got that most of these natural processes have long been *well optimised.* Mycelium connects with a root system and trades nutrients with it across generations of trees. Earthworms move tens of tonnes of soil per hectare a year for free. Fungi do break down an infested spruce over whole decades, but that slowness is actually fine. Any acceleration a machine could add always carries its energy cost, along with the risk of disrupting diversity and contaminating the substrate. Placing one more such technology into the landscape **usually adds nothing to biology, but it can give something to us.** A machine in nature can be useful, though not necessarily by reproducing a system that already works well. I don't mean to give up the ecological reach of the work, more to clarify the project's direction. --- ![[frame-predecessors.jpg]] ## Related machines - [Gilberto Esparza, Plantas Nómadas](https://gilbertoesparza.net/portfolio/plantas-nomadas/) - an autonomous robot roaming a polluted river - [Ian Ingram](https://www.ianingram.org/) - robots that pick up the signals of wild species and translate them - [Ken Rinaldo, Autopoiesis](https://www.kenrinaldo.com/portfolio/autopoiesis/) - A-Life robotic sculptures that change their behaviour and "communicate" with each other - [Ivan Henriques, Symbiotic Machine](https://ivanhenriques.com/works/symbiotic-machine/) - an autonomous solar machine that feeds on algae - [Edward Ihnatowicz, The Senster (1970)](https://en.wikipedia.org/wiki/Senster) - a shy cybernetic creature --- ## What the machine means I start from a shared premise: a machine built around defending (maintaining) its own conditions (equilibrium) *has something inside it can lose.* And that is what steers its behaviour. So-called *[precarious autonomy](https://ezequieldipaolo.net/wp-content/uploads/2011/10/dipaolo-jonas.pdf)* *(Hans Jonas)* says that even a very simple system that holds its own boundaries begins to have something like an interest in its own persistence, a stance toward the world. The machine isn't alive and feels nothing. It's enough that it has an inside with boundaries and a stake in keeping them. A machine with this kind of autonomy has its own *[Umwelt](https://en.wikipedia.org/wiki/Umwelt)* (Jakob von Uexküll's term). The world as it appears to it. Made of the variables it senses for itself *(battery voltage, air humidity, vibration in the soil)*. But the machine doesn't read the landscape through sensors alone, it reads it through its own body too. It knows how much energy it spends in a day, how long moisture takes to dry on its body, or what it touches while exploring. All of it is data about the place the machine is embedded in. What it comes to know about its surroundings is the robot's own diary, inseparable from the place it stands in. These objects aren't out to fake life. They're more about raising questions, because their behaviour really does sit on the edge. The same robot, when a human drives it, reads as a tool; the moment it acts from its own feedback, we read it as an actor (see [System](https://kindl.work/system-text-en)), without anything in it physically changing. Take [xenobots](https://www.youtube.com/watch?v=p3lsYlod5OU), tiny clusters of frog cells assembled by an evolutionary algorithm, now under study. They heal themselves, they can sweep loose cells into a copy of themselves and move collectively, with no one programming them step by step. These are the traits we denied *"machines"* for decades, because we built them to be *modular and predictable.* Not because a machine couldn't, in principle, manage it. *Čapek's robot, after all, was never made of non-living matter.* The machine's resulting behaviour is still authored, because I design it. What's emergent and unexpected is read into it by the observer. A mechanical machine on its own knows nothing and strives for nothing. **Personality, care, or intent are always projected onto the machine by a person.** --- ## Three poses To anchor all of this, I've defined three typological poses such an *electronic organism* can occupy. ``` MACHINE IN THE LANDSCAPE: THREE POSES A · VISITOR explores, maps mobile, seasonal → map of experience B · INHABITANT idle work stationary, slow → leaves an artefact C · TRANSLATOR listens to soil static, continuous → translation, attention ``` --- ### A | Visitor A mobile, surface-dwelling electronic organism. Conceptually it follows most closely from the hexapod *[Host](https://kindl.work/2026/The+Guest+and+the+Host)* and pushes it outdoors. **The Visitor** *does not work the land*. It doesn't plant, inoculate, or irrigate. It moves, explores, scans. On its back a LiDAR and a camera, in its body a piezo, sensors and probes. It collects data continuously and carries off *a map of the landscape* in a form no one else captures. *A map of experience:* where it stood longer and where it passed faster, where it detected movement and where it was alone. An archived testimony of time passing and of shifts in climate and vegetation. Its behavioural architecture is homeostatic. The *Visitor* tracks internal states *(battery level, internal temperature, signal quality)* and its behaviour follows from how it defends them. Its path through the forest is a by-product of the drive to keep itself going. A human observer can't tell whether the machine is drawn to the light or to a tree. They only witness *another way of being in the landscape* and persisting in it. ![[frame-visitor.jpg]] --- ### B | Inhabitant A disposed machine that in the landscape *just is*. It has no task, no measurable benefit for the ecosystem. A slow, small object that gradually *grows its own ritual*. In the morning it shifts a few centimetres to follow the sun, at noon it hides back in the shade. Sometimes it scratches a mark into the soil, other times it gathers a few twigs and arranges them into a structure I never gave it, one that means nothing in itself. It leaves behind countless artefacts of its own. This pose is *entirely useless*, and that's the point. From the ecosystem's side nothing dramatic changes. But we see a machine that *takes care of something*, even though neither of us knows what. A kind of *idle work*. Organised activity that runs and consumes without ever turning into use. In hardware terms it's an object with a tiny power draw and a handful of simple *axioms* (follow the light, retreat into shade, avoid water, save energy). This is also the only one of the three poses where it makes sense to let the behaviour *evolve* freely. The software can rewrite itself over time. The evolution serves no external task, it just reshuffles an empty gesture endlessly, and that *emptiness is the point*. ![[frame-inhabitant.jpg]] --- ### C | Translator A static vertical organism, 1 to 2 m tall. It has two vertical *light-guiding antennae* that communicate through light or sound. It also has *sensory roots,* sort of interchangeable probes giving the machine sensory access to the surrounding ecosystem. It perceives bioacoustics in the soil, contact vibrations on surrounding surfaces, humidity and light conditions, and possibly chemical or spectral inputs. The antennae work as a tool of translation. They capture what happens at frequencies and scales a person can't perceive directly. The soil sounds below the loudness threshold our hearing can reach. In the band of roughly 20 Hz to 8 kHz an activity unfolds underground that the antennae render into a perceptible medium through a synthesis of sound and light. *(springtails moving in the litter, the peristalsis of an earthworm, a bark beetle chewing its gallery in the inner bark of a dying spruce).* At the same time it can record continuously what it hears in the soil, building a local archive of the place. The unit runs on its own. It charges from the sun, goes without maintenance for long stretches, and speaks outward only when there's someone to speak to. The machine is entrusted with a *delegated attention* a person probably wouldn't have the patience for. It senses constantly and opens a channel whenever someone passes by. The antennae brighten and the speaker begins to synthesise the sound of what it hears underground. Over time the object adapts to the rhythm of its particular place, to its daily cycle, to humidity and temperature, or to the sound of life beneath the soil. ![[frame-translator.jpg]] #### "Antennae" The type C, *Antennae,* is the furthest along in development right now. I've managed to sketch out the hardware and its basic functions. I plan to develop it out in the field this summer at [PIFcamp](https://pif.camp/) in the Soča valley, at the end of July. The project also follows naturally from [Receptor](https://kindl.work/2023/Receptor) and [Synapse](https://kindl.work/2023/Synapse). *In time I'd like to grow it into a network of several objects in the landscape, passing information between each other and forming a shared technological (eco)system.* --- > **SIDE-NOTE: An audience beyond the human species:** There's still another angle to look at this from. What if the machine didn't communicate with us, but instead deliberately *gave something back to biology?* A spectral broadcast addressed to non-human perception, narrow-band LEDs in amber 590 nm tolerated by bats, near-UV 365 nm that draws moths, polarised channels for certain beetles. In that pose the machine would choose *an audience purely beyond the human species*, and the landscape would be its real addressee. For now I'm only speculating, but it would be interesting to explore this direction too. Perhaps in the future, with someone who knows these species. --- ## Research Log - [Active Sensing Subserves Task-Level Control](https://arxiv.org/abs/2605.22988) - active sensing doesn't come from a drive to minimise uncertainty, it falls out directly from controlling the machine's own behaviour. Movement is the substrate of perception, not a by-product of it. - [Sensing Intelligence as a Trainable Metamaterial Property](https://arxiv.org/abs/2605.23967) - a metamaterial body's shape can be trained to do part of the perceptual work itself. Up to fivefold more accurate sensing, or nearly an order of magnitude fewer sensors. - [Exploratory Experience Shapes the Geometry of Predictive Representations](https://arxiv.org/abs/2605.27929) - mice and agents that explore the landscape build a more organised internal map. The moment they head straight for the reward, the map falls apart. --- ![[litotest.jpg]] > *First litograph from [Synthetic Archive.](https://kindl.work/2026-W18)*