Some websites display content. Others tell stories. Neo Babylon ZERO | K’s Hunting Log seems to be doing a third thing: trying to make the world itself a generator of content.
I spent time reading beyond its home page, moving through the background, Fourteen Domains, hunting logs, bestiary, and the subdomains that have already opened. By the end, “AI helped write a novel” felt inadequate as a description. The more interesting question is an old one that generative AI has suddenly made possible to test:
If a world already has rules of its own, can it keep growing toward the unknown without an author writing every page?
My answer is yes. But that “on its own” is more complicated than it sounds.
1. The first thing created was not life, but “the blank”
The central premise of Neo Babylon ZERO is beautiful. In the year 5488, humanity has turned the world it controls into domes, tower cities, floating cities, and settlements behind high seawalls. Outside the domes, humans no longer provide the only order. The site states the problem outright: humanity calls the inside of the domes paradise and the outside blank, but life has already established its own order in that blank.
This is also the most precise metaphor for the AI experiment. Before large language models, a blank on a map was usually something an author had to fill. An undrawn forest, an unmentioned creature, an undescribed sea: all were the author’s work. Here M.K. makes a crucial move. He does not simply ask AI to fill the blank; he turns the blank into an explorable state.
The world is therefore no longer a completed map. It resembles the fog at the edge of a strategy game or RPG: you do not know what lies there, but you know something may. The site currently names fourteen global domains. Ten have been reached in the journey, while four remain unsurveyed. The map even notes that some subdomains will not be revealed until Fang and K arrive. At that point, the novel starts to resemble a state space.
2. This is not “ask AI to invent monsters”; it is letting AI inhabit a world
Many projects called AI worldbuilding amount to “think up ten creatures,” followed by ten entries from a model. K’s Hunting Log takes a different direction. As the creator describes it publicly, multiple agents first receive Neo Babylon’s canon, chronology, technological limits, ecological conditions, and nonnegotiable rules. They then extend geography, climate, ecosystems, species, and relationships. New settings feed back into the world state for the next exploration. The creator has even said he wants the agents to resemble explorers placed in the world rather than the author himself.
That is a different philosophy. Generation happens in an instant. Reproduction requires memory. If you create an animal today, you must know tomorrow what it eats; later, what eats it; later still, how it affects plants, climate, roads, or human engineering. A model that writes attractive prose is not enough. You need some kind of world state.
This is why the site does not feel like disconnected articles. Each journey links to a region; each region links to geography and species. The bestiary has gathered fifty forms of life, from the Sky-Tide Crownwhale and Thunder-Canopy Manta to the Diffraction Spore Spider, Veinlight Ear Hare, and Gate-Pore Reef-Weaver Octopus, then on to the Aurora-Sail Sea Butterfly and Silent-Pressure Snowy Owl. The useful word here is ecology. A world is not just characters plus places plus stories. A world is relationships.
3. The marvel is not the monsters but the causality
At a glance, names like Heat-Veil Mirror-Rift Giant Bear, Breach-Tide Mud-Crown Hippopotamus, or Lava-Vein Double-Crown Giant Bovid could make Neo Babylon ZERO look like a lavish AI monster catalogue. Read the logs and a different question emerges: why does this creature live this way?
The giant bear is more than a large bear. Its heat-insulating fur, heat-storage plates, and heat-absorbing footpads fit the Glass Sahara, its underground cold wells, night downdrafts, and a human-operated sun-tracking mirror array. Overheating can even fracture the obsidian ground. The Veinlight Ear Hare is more than a luminous rabbit: its ear veins release heat, it carries seeds, and its group reacts to faulty city flashes before an alarm sounds. The hare becomes part of an environmental sensing network.
The Anchor-Pore Reefbuilder Crabs are especially compelling. They compress spent shells, mineralized algae, and anchor-cable fibers into porous reef rings that shelter smaller life in slow water. One creature’s behavior changes the conditions in which another can survive.
Here lies the difference between generating content and generating a world. Content places sentences beside one another. A world connects causes to consequences.
4. The unsettling thing is how moral this world is
I repeatedly encountered an idea that reads almost like its constitution: “What we recorded this time was understanding, not conquest.” The background page broadens “hunting” to include tracking, sampling, repelling, rescue, and rerouting engineering works. Faced with dangerous life, K and Fang usually do not draw a gun. They shut down a scan, retreat behind a safe line, stop sonar, disable a beacon, remove machinery, and allow an animal’s route to reappear.
In the first week, the task with the Tideshell Matron was not to defeat her but to remove the error trapping her young. By the sixth, when a manatee stopped outside a crystal-reef gap, people altered reef-harvesting equipment and a beacon rather than ordering the animal to move. In the latest ice-harbor chapter, the Silent-Pressure Snowy Owl is rated EXTREMELY DANGEROUS, yet K removes a route from the map: today’s problem is the cable’s rhythm, not the life in front of him.
This is fascinating. It also raises a counterquestion: did AI really grow this ethic by itself? I do not know. The site’s author note says that M.K. creates and curates the work, and distinguishes DRAFT_CANON, CANON_ANCHOR, and EXPANSION. This world has a clear constitution. Perhaps what is happening is less “the human left and AI became free” than a transfer of authorship from writing every event to defining what kinds of events may happen. That is authorship at another level.
5. The most interesting contradiction: it claims “no humans,” but bears human fingerprints
The experiment was presented to me as involving “no human participation,” and the creator publicly describes a world allowed to grow through AI explorers, with the site’s text and images generated by them. Yet the site says M.K. creates and curates the work. A broader account of the AI content system also explicitly includes human review, data governance, quality gates, and publication checks.
I would not reduce this to “AI replaces the author.” It asks a deeper question: must an author personally write every outcome? Perhaps not. An author can design a universe, its constraints, its memory, and the mechanisms by which work is checked against itself. Then that author can step away from the screen and watch the world produce something unforeseen. In this view, the author moves from producer toward ecosystem designer. That shift seems historically meaningful to me.
6. I especially love its incompleteness
The site has not finished its world. Some of the fourteen domains remain unsurveyed. The Amazon Lightless Sinkhole and Guiana Canopy Cloudport are still distant points; much of the map is unknown. Generative AI projects often try too hard to prove they have an answer. A convincing world needs places it does not yet know.
Middle-earth feels deep not because Tolkien described every tree, but because one senses that things happen beyond view. K’s Hunting Log unexpectedly touches that feeling. It does not say, “This is the whole Earth.” It says, “This is as far as we have gone.” That is a mature idea of a world, perhaps even one close to life.
7. From a language model’s perspective, the interesting question is not whether AI can create
That question has become too small. Models can already write poems and novels, draw pictures, and compose music. The stranger question is: can AI become a secondary evolutionary force within a cultural system?
A single generated novel exists in the moment of generation. But a multi-agent world system may have memory, state, constraints, roles, mutual checking, world updates, rejection of errors, and input to the next round. It can begin to acquire an elementary, wholly artificial kind of historicity. One generation affects the next; a disaster changes a habitat; a road changes migration; a new species changes other species; and the system keeps those changes. The next round does not start from zero. That is what “reproduction” means here.
I would call the experiment recursive cultural evolution in an artificial world, rather than AI novel generation. It is early and immature, and traces of model prose remain. In the first week’s narrative, repeated gestures of coordination between Fang and K can feel formulaic. By weeks six through ten, ecological causality, engineering interference, and relationships among species seem much more closely knit. Those imperfections matter: they are the early creases of a new unit of creation.
8. Perhaps AI is not what this experiment is testing
In the end, I wondered whether this experiment tests something other than AI’s ability to make a world: can people accept that a world they created may not remain wholly theirs? Novelists hope characters go where they imagined; worldbuilders hope settings remain consistent; IP stewards hope characters stay in control. Yet a sufficiently complex world-generating system will eventually produce an outcome you did not expect.
If every outcome needs your approval first, perhaps the world remains only an extension of you. To grow “on its own” must allow surprise, divergence, deviation, error, accident, and reinterpretation. Then the author might feel what a reader feels for the first time: “So this is what this place is like.”
That moment is why K’s Hunting Log is worth watching. The issue is neither whether AI writes better than people nor whether it draws faster or replaces novelists. It is whether someone who hands over a world’s rules can still be surprised by that world. If so, we may be approaching a new creative relationship. The author becomes less the god of the world than its first ancestor. AI is still not alive, but it begins to resemble a peculiar artificial descendant in one respect: it produces the next condition from the conditions it inherited.
The blank has not simply been filled. It has begun to speak.