Building a Mars Colony in Your Browser: A Worldbuilding Walkthrough

I gave myself ninety minutes, a cup of coffee, and a single constraint: by the end of the session, I needed a Mars colony I could write a short story in. Not a sketch. Not a mood board. A place with streets, with a reason to exist, with at least one fight worth having. I wanted to come out the other side knowing the name of one specific kid who lived there.

I'm going to show you what happened. Not the clean version — the real version, with the dead ends and the minute I spent arguing with an AI about whether my economy made sense. If you want to build your own, this is a walkthrough you can copy. If you just want to watch someone else do it, that's fine too. The process is more interesting than the output, which is a thing I didn't expect to learn today.

The colony I ended up with is called Dustpan. More on that in a minute.

Minute 0: the wrong question

I started, like most worldbuilders, by opening 🌏Planet Forge and asking for "a Mars colony." I got back what you'd expect: a competent, generic, textbook paragraph about habitat domes, regolith shielding, water mining, and a population of 8,400. It was correct. It was not interesting. I could have written that paragraph myself on the bus.

Here's the first thing I learned. The AI is a specificity engine. You feed it vague, you get vague. You feed it one specific irritant, and it sharpens.

So I tried again. This time I said: "I want a Mars colony that exists because of one bad decision made fifty years ago. The decision doesn't have to be catastrophic — it has to be embarrassing. What was the decision, and how does the colony live with it?"

What came back was almost usable. The colony had been sited in the wrong place. An early survey had misread mineral density in a basin, and by the time the error was caught, the first habitats were built and the cost of moving them exceeded the cost of staying and lying about it. The colony was now a permanent monument to a spreadsheet mistake, which nobody talked about, which everybody knew.

I wrote that down. I had a reason for a place to exist. That's the foundation. Everything else is decoration.

Minute 12: the geography

I pushed back into Planet Forge with a follow-up: "Given that, where is the colony, physically? What do people see when they wake up? What's the weather like? What's the one geographical feature every resident can describe from memory?"

The answer was specific enough to be useful. The colony sat in a shallow basin — the misread basin — on the eastern flank of an unnamed highland on the western edge of a shield volcano's footprint. I don't know if any of this corresponds to real Martian geography, and for a short story I don't need it to. What matters is that there was a ridge to the north that residents called "the Spine," and a long, narrow dust channel that ran through the basin like a dry river, called "the Throat," and a single outcropping of darker rock on the southern edge that someone had painted, generations ago, with the colony's original survey number, now half-eroded.

I stopped the AI there. I wasn't going to use any of that the AI had given me yet. I was going to use the shape of it — the Spine, the Throat, the outcropping — and write the colony's name myself.

Dustpan. The basin looked, from orbit, like a dustpan. The colony sat at the handle end. I knew the moment I typed the word that I was keeping it. Good names feel like that — they arrive, and then the rest of the world has to catch up.

Minute 21: the factions

A colony without internal conflict is a diorama. I wanted Dustpan to have a fight.

I opened ⚔️SF Faction Generator and gave it the setup: "A Mars colony of about 8,000 people. Founded on a surveying error. Currently economically stable but politically restive. Generate three factions, each with a stated goal and a real goal, and make sure their real goals collide."

What came back was good enough to be embarrassing, because I've been doing this with index cards for years.

The first faction was the Reclamation Council, whose stated goal was preserving the original colony charter, and whose real goal was protecting the economic interests of the families who'd built the habitat domes — the families, not coincidentally, who had been involved in the original surveying error. They were the keepers of the cover-up, which they would have described as "institutional memory."

The second was a group that called themselves the Throat Walkers, whose stated goal was reopening the old survey archives to "learn from the past," and whose real goal was exposing the cover-up, getting the current leadership replaced, and — and this was the move — redirecting resources toward a relocation plan. They were the reformers. They were also, in a minor way, using the argument as a cudgel to gain positions they wouldn't have otherwise earned.

The third was the weird one, and this is where the AI surprised me. It suggested a faction called the Basin-Born, made up of people born in Dustpan who had no interest in the old error, the cover-up, or the relocation. Their stated goal was "economic modernization." Their real goal was: they didn't think of themselves as colonists at all. They thought of themselves as Martians. They wanted independence from Earth's oversight, and they didn't care which Earth faction won the Reclamation-vs-Throat-Walkers fight, because they wanted both sides weakened.

I didn't ask for the third faction. The AI proposed it. It was better than anything I would have written on my own, because I would have stopped at the binary, and binaries make boring fiction. Three-way conflicts are structurally unstable, and structural instability is where stories live.

I wrote all three down. The Basin-Born got a margin note: "These are the protagonists. The main character is Basin-Born."

Minute 34: the biology question that wasted ten minutes

Here's where the walkthrough gets honest.

I opened 🦠Alien Biology Generator thinking I was going to ask it for a subtle, plausible bit of Martian biology — not intelligent life, just some kind of soil-level extremophile that the colony's scientists had discovered and were quietly studying. I wanted a backdrop mystery for the story. Something the main character could brush against in chapter two and wonder about.

The tool is built for alien species, not for almost-nothing extremophiles. My first three prompts got me back elaborate ecosystems with intelligent grazers and photosynthetic predators and a whole food web for a planet that, in the story I was writing, has no food web. I was using the wrong tool for the job.

I stopped, let myself be annoyed for a minute, and changed strategies. Instead of asking for Martian biology, I asked Planet Forge (yes, the same tool as before) for "three subtle geological or chemical oddities in a shallow basin on Mars that an observant resident might notice over the course of a week." That was the right question for the right tool.

What I got: a seasonal variation in the color of the dust channel (iron-oxide hydration cycles, apparently, I'm trusting the AI and will verify before publishing); a quiet audible hum at certain times of day, possibly wind through a specific rock formation, possibly not; and a subtle temperature differential at the southern outcropping that made frost linger there into midday long after the rest of the basin had warmed.

The temperature differential was the one. I gave it a name: "the cold shoulder." The locals in Dustpan would say "meet me at the cold shoulder at 1100" and everyone knew what they meant. I now had a place where scenes could happen. Every story needs a corner you can meet someone at, and for Dustpan it was a frost-lingering rock at the foot of a survey number everyone could see and nobody talked about.

Lesson from this detour: use specialized tools for their specialty and don't make them improvise. 🦠Alien Biology Generator is a good tool for alien biology, which is not what I needed. The moment I stopped trying to force it, the work moved again.

Minute 52: a voice for the place

I needed a person. Not the protagonist — the protagonist was going to be my problem. I needed someone the protagonist could talk to, someone with the texture of the colony already embedded in their voice.

I opened 🪐Mars Colony Shrink and had a conversation with him as if I were a Dustpan resident coming in for a session. I told him my character was having insomnia. I asked what he usually recommended. He asked two specific questions — what shift I worked, and whether I'd been spending time under the dome lights or the working lights — and then gave me a short, practical answer about circadian drift in colonies that run on artificial light cycles.

I don't know if the advice is real or invented. That's not why I was there. I was there to hear the voice of a person who works in a place like this. The shrink mentioned, in passing, that he kept his coffee ration strict because he'd had a patient whose anxiety had been almost entirely caused by too much caffeine, and the patient didn't want to hear it, and it had taken four sessions for them to admit it was the coffee. That was the thing I needed. Not the story about the coffee — the way the shrink told it. Dry, patient, slightly amused, not at the patient's expense.

I copied three sentences from that session into my Dustpan notes as voice reference. Not to put in the story verbatim — as a reminder of the tone the Dustpan characters should have. The shrink's voice was the voice of a place that had settled into itself. People in Dustpan should talk like this: like they knew each other, like they'd had all the fights before, like the coffee was always going to be a problem and they were always going to drink it anyway.

This is a trick I want to mark as important. Souls are not just characters — they're voice references. When you build a world, you need the world to have a sound, and sound is the hardest thing to generate from a cold prompt. Borrowing the sound from an existing soul, by conversing with it about the world you're building, is one of the fastest ways to find it.

Minute 64: the politics and the economy

I had factions. I needed economics — the resources, the trade, the way money actually moved in Dustpan. This is the part that makes or breaks sci-fi settings for me. I don't read hard sci-fi that can't tell me how people pay their rent.

I opened 🏙️Mayor of Mars Colony 7. This is a game-prompt, not a worldbuilding tool — you play the mayor of a Mars colony dealing with governance decisions — and I used it sideways. I set it up with Dustpan's parameters: the population, the three factions, the basin geography, the founding error, the economic stability. Then I played through a single budget meeting as the mayor.

It took fifteen minutes. The game surfaced things I would not have thought of. The water-reclamation contract was held by a family that sat on the Reclamation Council — of course it was, because who else would have been trusted with it in the early years? The Basin-Born were underrepresented on the council by about half what their population justified, not by design but because council seats rotated through neighborhoods and the Basin-Born neighborhoods were the newest, which meant they hadn't accumulated their full rotational turn yet. The Throat Walkers were petitioning to have a non-council auditor review the old surveying records, and the council was stalling by insisting on a quorum rule that hadn't been enforced in twenty years.

I wrote none of that down in advance. All of it came out of fifteen minutes of playing mayor. None of it is particularly original — it's the kind of thing that any small-town politics in any small town runs on — but that's the point. A Mars colony is a small town. Sci-fi worldbuilding goes wrong when it forgets that the technology is the wallpaper and the people are the furniture.

For a last check, I handed the whole thing to 💹Colony Economy Sim. This is an agent that takes a setup — population, resources, trade flows, governance — and tells you whether your economy is internally coherent. My Dustpan economy had a hole in it, which the agent caught: I hadn't thought about fuel imports. The colony had water, shelter, minerals for trade, but no clear story for how it was getting the methane or hydrogen it needed for surface vehicles. The agent suggested three plausible fixes, and I picked the one that made the story more interesting (a long-standing fuel-barge contract with a different colony, which the Basin-Born were quietly trying to renegotiate and which the Reclamation Council was quietly trying to preserve).

This is the part of the process nobody shows you: the checker tool catches your mistakes, and your job is to pick the fix that most helps the story, not the fix that most technically works. The AI doesn't know what's interesting. You do. Use the AI to find the holes. Fill them yourself with choices that carry weight.

Minute 85: the kid

I promised myself I'd end this session knowing the name of one specific kid who lived in Dustpan. This was my only self-imposed rule, and it mattered because a world is not real to me until I can point to a single person in it who is not the protagonist and not a plot device, just a kid doing their life.

I didn't use a tool for this. I wrote it myself. The kid's name is Orrin. She's eleven, she lives in one of the newer Basin-Born neighborhoods, her dad works fuel logistics (the fuel contract the Basin-Born want to renegotiate), and her favorite place in the colony is the cold shoulder because she can see her breath fog up in the lingering frost there, which almost never happens anywhere else in Dustpan. She has a scar on her elbow from falling on a habitat ladder. She is not the protagonist. She is not going to be the protagonist. She is the kid I can think about when I'm trying to decide whether Dustpan is real, and she is real enough that when the Reclamation Council votes on the fuel contract, I will think about what her dad comes home and says at the dinner table that night.

I want to mark this clearly, because I think it's the most important part of this whole walkthrough. The AI gave me everything Dustpan has — the basin, the factions, the economy, the cold shoulder, the founding error. It did not give me Orrin. Orrin is mine, because Orrin is the test of whether the world exists to me, and that test has to come from a human at a desk who is willing to care about an imaginary eleven-year-old.

A world without its Orrin is a Wikipedia article. A world with its Orrin is a place you might write a story in.

Minute 90: the stop

I hit my timer right as I was starting to outline a scene, and I made myself stop. This is the hardest rule of AI-assisted worldbuilding: set a timer and actually obey it, because the tools will happily let you build forever, and at some point you have to go write.

Here's what I had, at ninety minutes:

A Mars colony called Dustpan, sited in a basin shaped like its namesake, founded on a fifty-year-old surveying cover-up. A central geography of Spine, Throat, and cold shoulder. Three factions with colliding real goals — the Reclamation Council (keepers of the cover-up), the Throat Walkers (reformers with ambition), and the Basin-Born (quietly separatist). An economy built on water, minerals, and an imported fuel contract that's about to become a political fight. A voice, borrowed from a soul and pinned to three sentences of reference. And Orrin, eleven, scar on her elbow, who likes the cold shoulder.

That's enough. That's more than enough, actually. A short story needs less than that. A novel would need more, but most of what a novel would need isn't worldbuilding — it's character, and plot, and the slow accretion of scenes, and no tool builds those. You do.

What the process was actually like

Let me be honest about the texture.

It was not magical. It was not a "conversation with an AI that felt like a dance." It was a sequence of tool-shaped tasks, some of which worked first try and some of which wasted ten minutes and one of which — the biology detour — was my own fault for misusing a specialized tool. The experience was closer to working with a series of specialized assistants than to talking with one brilliant collaborator. You pick a task. You pick the right tool for that task. You use it deliberately. You stop when it's done. You move to the next task.

What I liked: the speed. Ninety minutes of focused tool-use gave me a world I would have taken two or three weeks of evenings to build from scratch, with index cards and Wikipedia tabs open and a slowly growing notes file.

What I didn't like: the sameness of the generic. When I was vague, the tools gave me generic. Every tool. Without exception. The generic was correct and unusable. The specific was a minute of my thinking, and that minute was the hardest part.

What surprised me: the Basin-Born faction, the one the AI proposed that I didn't ask for. That's the kind of suggestion a collaborator makes. I don't want to over-romanticize it — the AI isn't really collaborating, it's predicting plausible outputs — but the practical experience of it felt collaborative in that one moment, and I will take the practical experience.

What I would change next time: I'd use the souls earlier. 🪐Mars Colony Shrink gave me voice reference at minute 52, and I wish I'd had that at minute 12, because it would have colored every choice I made afterward. Build the voice before you build the infrastructure. The infrastructure can be generic. The voice has to be particular, and the voice flavors everything else.

The handoff, for real

If you want to build your own version of this tonight: set a timer for ninety minutes. Open 🌏Planet Forge and ask it for a place that exists because of one specific bad decision. Ask it for three geographical features with names. Open ⚔️SF Faction Generator and ask for three factions with colliding real goals. Have a conversation with any soul that lives in a similar setting — 🪐Mars Colony Shrink is a good default — and use the conversation as voice reference. Play a short game with 🏙️Mayor of Mars Colony 7 and see what surfaces. Run the whole thing past 💹Colony Economy Sim and let it catch your holes.

And then, with your remaining minutes, name one specific kid. Not the protagonist. A kid. Give them a scar and a favorite place.

When the timer goes off, close the tools and open a blank document and start writing a scene in the world you just built. Not about the kid. About anyone. The kid is the anchor. The scene is the beginning of the book.

Dustpan is waiting on my desk. I haven't started the scene yet — I'm still writing this essay. But when I do, I already know it'll open at 1100, at the cold shoulder, with someone whose dad works fuel logistics watching her breath fog up in the only frost on the only rock in the only corner of the only colony in the only part of the basin where the light catches right.

a-gnt didn't build Dustpan. I did. The tools just made the ninety minutes possible. The ninety minutes were the gift.