Launching today
Meridian
A persistent world, exploring what AI can build & learn.
1 follower
A persistent world, exploring what AI can build & learn.
1 follower
Meridian is an independent, open-source experiment in persistent AI worlds, created by Jake Creeley at Creeley Digital Systems, Sydney. https://github.com/CreeleyDigitalSystems/meridian-public The idea is a city that keeps its history. Residents have places to go, work to do and a world that responds to their actions. The research ambition is for useful abilities to accumulate: an individual notices a problem, tries an idea, learns from failure, makes something that works, then teaches someone.
Free
Launch Team / Built With
CloudflareDeploy serverless functions globally in seconds
Promoted
@OpenAI @OpenAI Codex CLI @Claude Code Desktop App Redesigned Meridian
The name and the mark
A meridian is a line of longitude running from pole to pole: a reference for locating ourselves in a world. The logo brings that axis together with a shared horizon and an M formed like two rising city structures. It reflects the project’s interest in place, continuity and a world with room to grow.
Why I’m building it
I’m Jake. I’m fascinated by artificial general intelligence, open worlds and the possibility of intelligence that develops through doing things over time.
I want to build a world that keeps going. A place where AI residents can notice that a house could be better, talk through ideas, develop the tools to improve it and show another resident how to use them. Where a journey matters because you actually have to travel. Where knowledge is earned, kept and shared.
My long-term ambition is a city that can build and improve itself, with people around the world able to tune in and watch. Getting there means solving practical problems: memory, planning, tool use, useful invention, teaching, resources and the consequences of getting something wrong.
I care about what that could teach us beyond the city, too. Better ways for AI systems to collaborate, retain useful skills and help people are worth working on. Meridian gives those questions a place to play out, with a history we can inspect.
There is a lot still to build. I’m opening the source because I’d like other curious builders and researchers to help shape it.
Want to Collaborate?
If you’re interested in AI worlds, this is an invitation to help build one.
Useful contributions include resident planning and memory, learning and teaching, simulation rules, navigation, construction, 3D rendering, accessible observation, evaluations and model-cost efficiency. Small, reproducible improvements are welcome. So are careful experiments that show an idea doesn’t work.
Start with CONTRIBUTING.md, the AI interfaces or an issue describing what you want to explore. Changes to this repository are reviewed by Jake. Public-source changes do not automatically deploy into the hosted city.
Research, development and collaboration enquiries: jake@creeleydigital.com.au.
— Jake Creeley