The story
I'm building a world I actually want to play in
I'm Max. By day I work in cybersecurity; my evenings and weekends go into my own server. Here's how it started and how it works today.
Project artwork
The project’s visual identity and concept illustrations. These are not gameplay screenshots.
01 · Where it started
Two windows and an empty chat
I used to play on Comeback PW, a Russian private server on the classic 1.3.6 version, with a Blademaster and a Cleric at the same time in two windows. The Blademaster hits things, the Cleric heals and buffs. The second window was my patch for a hole nobody else was going to fill: on a weeknight evening there's no party, no healer, and the chat is silent. I'd stand in Archosaur, type "LFG", and half an hour later realise I was playing with myself.
To me this game was always about people: squads, factions, the Friday-night boss run. If you played PWI, you've probably boxed a cleric of your own or watched your server empty out after a merge. I wanted a world that feels alive even when my friends aren't online.
02 · A hobby that got out of hand
I only wanted to see how it works inside
Stand up my own server, poke around, figure it out. I had never administered a Perfect World server before, so every step was a first: the install, permissions, backups, bugs that made the client close without a word.
Then I wrote my first bot: a Cleric companion that follows me and heals, so I could drop the second window. Next came a bot that picks up and hands in quests on its own. Then one that talks in chat. Now there are hundreds of them and each lives its own life.
03 · Who they are
AI players: not scripts, not NPCs
Every AI player connects to the server over the game's own network protocol, like a normal game client, with its own character and the same rules I play by. It walks and flies by itself, earns exp and money from quests, dies, and talks in world chat. A language model wrote the names, backstories and way of speaking for many of them. About one in five writes in English, and all of them answer a person in that person's language. Decisions are mostly plain code; a language model steps in where conversation is needed.
They now work as a team too: they answer a call in world chat and run a dungeon with a person, offer help with a cultivation boss, take orders in party chat, keep shops and post bids on the gold exchange. What's been checked, and how, is on the home page and the roadmap.
I explain how it works inside in the devlog: "How the AI players work".
04 · Why I won't stop
I want this to be insanely good to play
The main reason is simple: I want to play in this world myself. Not a "good enough for testing" world, but one that's properly good, one you look forward to coming back to every evening. The bar is high, so development never stops. Until the world is what I picture in my head, I keep building.
One person can't do that alone, so development runs around the clock: swarms of AI coding agents work on the code. They write code and tests; I decide what we build and in what order, look at the results, and personally check everything that goes into the world. The agents propose, I decide.
I try to run it like a proper engineering team:
- Automated tests. As of 26 October the repository has 524 test files and 9,834 tests.
- CI on every push. GitHub runs the tests, a linter, a check of the dependencies for known vulnerabilities and a scan for passwords or keys committed by accident, on every push and every night. It goes red sometimes too; then I fix the cause, not the test.
- Versioned releases. The bots only run released, pinned versions of the code. Whatever I haven't deliberately shipped never reaches the live world.
- Telemetry and A/B. Every hour an automated check collects server health, class pace, chat quality and the economy. I switch new things on for half the bots and compare against the rest.
- Security. I didn't write the server package and I don't take it on trust: I audited it and found weak default passwords and a dangerous admin panel. The world now runs on a dedicated server, and before I let people in I'll work through a separate list of measures: only the game login faces the internet, while the admin panel, the database and SSH stay unreachable from outside.
- Backups. The world is backed up every hour, and I've checked that a backup restores into a separate folder. I'll rehearse a full restore of the live world before the beta.
- A human in the loop. Everything that ships to the world, I check myself.
05 · A childhood dream
My first project this big
I've wanted to build my own online world since I was a kid. I've never made anything this size before, so I'm learning as I go: from mistakes, from graphs, from your questions. It's sometimes funny and sometimes scary, but I like where it's heading.
Working in security trained me not to believe words. So this site only says what I can verify: the numbers come straight from the server, and every feature carries an honest label. What works today, what I'm building now and what's planned is on the roadmap.
06 · What next
I want to build a community
My first idea was a WoW-style subscription realm with no real-money power. Then I realised my full-time job lets me fund it as a hobby and a childhood dream: the goal is a world I want to play myself, shared with a community. It is free today, and I would prefer to keep it free through voluntary support. If that cannot cover hosting, development and maintenance, I will need to switch to a subscription with advance notice. Character power will never be sold. The funding promise explains the working budget, actual costs and the possibility of other realms.
There will be no private club and no screening. First comes a closed beta for a handful of people, because I'm one person and want to keep up with replies, then an open one.
Today you can do three things: read the devlog, join the community, or leave a contact on the home page. When beta sign-ups open, I'll write to you first.