Dear friends,
First of all, thank you — sincerely.
Since MoFA was accepted into Google Summer of Code, friends from around the world have reached out to ask:
- What kind of proposals are you looking for?
- What is the future direction of the project?
- Can you assign me a task?
We appreciate your enthusiasm. In this letter, we want to answer these questions openly and thoughtfully.
Why not just give you a ready-made answer? Because with the rapid advancement of AI capabilities, MoFA is undergoing a fundamental transformation. We have redefined our project strategy and want to share it with you. GSoC work will be aligned with this strategic direction.
I. MoFA’s Goal: Composition AI
Our mission has always been:
Empower the extraordinary with Composition AI.
AI is advancing faster than anyone expected — models are growing more powerful, tools more abundant, and Agent systems are emerging everywhere.
But we see one problem with increasing clarity:
The world does not lack large models. What it lacks is the ability to compose. And this is precisely where developers can contribute — and where the results will benefit the developer community in return.
- How do we compose multiple Agents?
- How do we compose tools and capabilities?
- How do we compose models and data?
- How do we compose content and media?
- How do we compose distributed systems?
MoFA is evolving toward:
A future-oriented collection of local, distributed, and composable AI architectures, frameworks, tools, and applications.
We are no longer just a framework. We are building comprehensive AI composition capabilities — empowering ordinary people to achieve extraordinary things.
II. What MoFA Is Becoming
MoFA is no longer just a framework — it is evolving into a layered system:
- Agent Framework → mofa: Our new Agent framework, built on Rust for performance and reliability. It adopts a microkernel architecture, inherits and extends the composition capabilities of MoFA 1.0, and targets production-grade applications.
- Agent Engine → mofa-claw: MoFA’s own OpenClaw — itself an application built on mofa. We aim to make it a more efficient and controllable Agent engine.
- Content Engine → Remix Engine: The greatest contribution of LLMs to human cognition may not be producing content, but remixing it — transforming content from one form into another that better suits its audience. We believe systematically building this remixing capability is immensely valuable.
- Remix Tools → MoFA-studio, MoFA-pptx: MoFA-studio focuses on remixing in the audio domain; MoFA-pptx covers images, short videos, presentations, and other multimedia formats.
- Application Layer → mofa.fm, zhufu.ai, what’s-going-on, MoFA Wonderland: End-user applications built on top of the layers above.
- Model Layer → Moxin Models and third-party models: Through mofa, we support models deployed both locally and in the cloud. Moxin, our sibling project, provides high-quality speech models for local deployment, giving us a unique advantage in the voice domain.
We build infrastructure and real applications in parallel.
Experiment fast, build fast, iterate fast, fail fast — MoFA is not a static codebase. It is a system that continuously evolves in a rapidly changing era.
III. What This Means for You
If you plan to submit a GSoC proposal, we want you to understand: we are not looking for “simple feature additions.” Ask yourself:
- Which layer am I improving?
- How does my contribution enhance composition capabilities?
- How does it fit into the overall architecture?
- Can it connect to a real-world application?
We value architectural understanding and systems thinking over feature accumulation.
We welcome:
- Creators who love to experiment
- Engineers with systems thinking
- Builders who care about real-world impact
- Explorers who are not afraid to fail
- People who want to be part of open source for the long term
You don’t need to know everything, but we hope you are open-minded and resilient.
We are navigating a rapidly changing AI world and do not pretend to have all the answers. We must push forward while embracing uncertainty and imperfection — what matters most to us is alignment in values and vision.
No matter where you come from, we hope you will keep building, keep exploring. MoFA is still in its early stages. That means your contribution can truly shape its direction.
We look forward to creating together.
— The MoFA Team