← Back to list

We Learned to Automate Tasks. Then I Realized Being Alive Is One.

A late-evening spiral about loop engineering, umwelt, shared latent spaces, and whether being alive is just a very long loop running inside…

The Outlier · 2026-06-16 21:11 · 0 claps · 7.7 min read
#loop-engineering #artificial-intelligence #ai-agent #philosophy #umwelt
Open on Medium ↗
Wiki topics: AGT · AI Agents AI · AI · General PHI · Philosophy 🔧 · Data Engineering 🔭 · Astronomy & Space 🏃 · Running & Endurance

We Learned to Automate Tasks. Then I Realized Being Alive Is One.

A late-evening spiral about loop engineering, umwelt, shared latent spaces, and whether being alive is just a very long loop running inside its own small world.

Photo by Patrick McManaman on Unsplash

Photo by Patrick McManaman on Unsplash

Two hundred years ago, a teenager dreamed up a man who builds a person.

Mary Shelley was barely nineteen. In her story, a scientist named Victor Frankenstein gathers parts from the dead, assembles them into a body, and infuses it with a spark of life. She called the book The Modern Prometheus, after the one who stole fire from the gods and handed it to us. And the dream underneath it is older than the book, older than clay and myth and every story where somebody shapes a creature and dares it to wake up.

So let me be honest from the very first line. The idea of building a living mind is not mine. It belongs to all of us, and it has for a very long time.

What is mine is a small obsession that started last week, when a flat, unglamorous engineering term crawled into my feed and refused to leave.

I have not stopped thinking about it since. So I am going to think about it here, out loud, and I want you to come with me, because somewhere in the middle of this, my brain did a thing I cannot un-do, and I am curious whether it will do the same to yours.

The Two Words That Started It

The term is loop engineering.

Two words, really, loop and engineering, bolted into one flat little noun, the kind of thing you would expect in software release notes, not in a daydream about souls.

Here is what it means. For the last couple of years, the way you got real work out of an AI was to sit there and prompt it, one message at a time. You held the tool the whole way. Then, just recently, the people building these systems started saying something stranger: stop prompting the agent at all. Instead, design a loop that prompts it for you. You write down what “done” looks like, once. The loop runs the agent, checks the result against your definition of done, fixes what is wrong, and goes again. And again. Until the thing is finished, with your hands nowhere near the keyboard.

One of the people who leads this work put it flatly: he does not really prompt anymore. His job, he says, is to write loops.

Now, I want to be upfront about something.

The engineers mean this narrowly. A loop gets a task done: ship the code, scrape the site, close the ticket. Useful, bounded, finite. That is the whole claim.

I am about to steal their word for something much bigger than they meant. So watch me closely, and catch me if I cheat.

But What Even Is a Task?

Here is where the new term collided with an old question, and the two of them fused.

Because if a loop is just this: sense, decide, act, check, repeat, until a goal is met, then I cannot find the line where it stops describing software and starts describing me.

Think about what a living thing actually does. Strip away the poetry. A creature opens its senses onto the world, takes in whatever it can, and does something back. Then the world shifts, so it senses again, and acts again. That is the entire job. Sense, respond, stay alive, repeat.

We are biological organisms, the same as everything that crawls or swims or grows. Evolution shaped us, and it shaped us toward one stubborn purpose: keep going. Find food, because the body runs down. Reproduce, because the body has an expiry date, and the only way the pattern survives is to copy itself into a fresh one before the old one quits. In one famous way of putting it, we are the survival machines our genes build to carry themselves into the next round.

Underneath the love songs and the cathedrals, biology is running a very old loop. Eat. Last. Copy. Repeat.

I should play fair here, because that is a tidy story and reality is lumpier than it. Biologists still argue over how much of us truly reduces to that one command, and a fair amount of what we are is probably a leftover, a side effect evolution never bothered to switch off, rather than something it carefully chose. So hold “everything is for survival” as a lens, not a law.

But what a lens. Because we do have more than hunger: curiosity, logic, the strange habit of lying awake wondering about loops. And yet the “more” keeps curling back into the “less.” We are curious because curiosity finds food and danger. We reason because reasoning buys us another season alive. Follow almost any thread far enough and it ties back to the same knot in the center: to exist, and to keep existing.

Which means existing is not separate from the tasks. Existing is the task. The longest-running one there is.

Every Creature Is Locked Inside a World of Its Own

There is an old idea in biology I cannot get enough of, and it pulls the loop even tighter.

A century ago, a biologist named von Uexkull noticed that no animal lives in “the world.” Each one lives in its own private world, the thin slice of reality its particular senses can reach. He called it the umwelt. Take a whale. Not some alien speck of a creature, but a fellow mammal: warm-blooded, air-breathing, nursing its young, about the nearest thing the open ocean has to a cousin of ours. And still, its world is almost nothing like yours. Down in the lightless deep, where eyes are next to useless, a whale feels out the world with sound: it clicks into the black and reads the echoes that come back, taking the measure of a fish, a cliff, another whale, all by ear. You get a world of light and color and faces. It gets a world of echo and pressure and depth. Neither of you is touching reality itself. Each of you is sealed inside the only world your senses can build. And both are complete.

And here is the part I promised you. The part that did the thing to me.

The way von Uexkull drew it, the umwelt is a circle. The animal senses the world. The senses trigger an action. The action changes the world. The changed world gets sensed again. Perception and action, wired into a feedback loop, closing on itself, around and around, for an entire life.

Read that one more time.

That is a loop. A man sat down a hundred years ago to describe what a living thing is, and what he drew was a loop.

Now I have two pictures stacked on top of each other and they will not come apart. On one side: an engineer, this month, writing a loop that senses, decides, acts, and repeats inside a task. On the other: a whale, a dog, a person: sensing, deciding, acting, repeating inside a world. The same shape. Drawn twice, a century apart, by people who had never heard of each other.

The Part Where It Gets Genuinely Weird

So let me say the daydream out loud, and let it sound as unhinged as it wants to.

If a living thing is just a sense-and-respond loop running inside its own small world… how much of that could you actually build?

Start with the senses. Give one model eyes. Give another ears. Give a third a sense of touch, or temperature, or hunger, some inner gauge that says the system is running low. On their own they are loose organs muttering in separate languages, no more alive than parts in a drawer.

But we already know how to make separate senses speak one language.

This is one of the quietly astonishing tricks of modern AI: you can take wildly different inputs, a picture, a sentence, a sound, and fold them all into the same inner space, a place where a photo of a dog and the word “dog” land almost on the same spot. Different organs, one shared inner tongue. The senses, finally talking to each other.

Then you drop something into the middle. A coordinator. A model whose only job is to read what all the senses are reporting, decide what matters, choose what to do, push the signal out, and then take in what comes back, and choose again.

A loop. A world-model in the center, senses feeding in, actions flowing out.

And I have to tell you, this is the part that tipped me from “fun shower thought” to “wait, are we actually doing this”, because people already are. At least one recent proposal sketches a modular AI built deliberately like a brain: separate specialist parts, all feeding into a single shared inner space, an artificial echo of the way your own cortex melts sight and sound and touch into one seamless moment.

I am not telling you that is a mind. Sit in that gap with me for a second, because the gap is the entire point.

But look at what is standing there once the dust settles. A thing that senses the world through many channels. Folds them into one inner picture. Decides. Acts. Senses the result. Goes again, and keeps going, because going is the goal. Squint, and that is not an app anymore. That is the skeleton of an umwelt. The oldest loop in biology, re-poured in software.

Maybe life, the bare structural bones of it, not the warmth, just the architecture, is more buildable than we let ourselves believe.

What I Am Absolutely Not Saying

Let me yank the leash, because this is exactly the kind of idea that bolts off into the night, and I would rather hand it to you honest than dressed up in a cape.

I am not claiming anyone has made a living thing. I am not claiming a loop of models has an inner life, or feels the June light on its face, or lies awake afraid of its own expiry the way we do. A system that drops a picture and a word onto the same mathematical point is a long, long way from a creature that simply wants to see tomorrow. The distance between “acts as if it has a world” and “has a world” might be the most important distance in the universe, and I do not get to skip across it just because I found a pretty pattern in the dark.

So please take every word of this as a question, not a verdict.

But the question will not let go of me, and maybe now it has a grip on you too. We set out to automate our work. To get the dull tasks done while we slept. And somewhere along the way we started building little systems that sense, decide, act, and loop, the exact move biology has been making, quietly, in the dark, for a billion years.

We thought we were writing loops to handle our tasks.

We might have been writing the one task hiding underneath all the others. The plain, stubborn, endless task of continuing to be.

So tell me where you land. Is being alive just a loop: a very old, very patient one? Or is there something in there that no loop will ever close? I really want to know.


메타데이터
post_id
17f01ffa9bb9
slug
we-learned-to-automate-tasks-then-i-realized-being-alive-is-one-17f01ffa9bb9
url
https://medium.com/@amaljova/we-learned-to-automate-tasks-then-i-realized-being-alive-is-one-17f01ffa9bb9
canonical_url
https://medium.com/@amaljova/we-learned-to-automate-tasks-then-i-realized-being-alive-is-one-17f01ffa9bb9
author_url
https://medium.com/@amaljova
status
ok
fetched_at
2026-07-15 05:28:05