Alpha is the AI Potluck's first fully open source AI product: an AI assistant built and owned by the public, designed to replace proprietary tools. Our roadmap tracks how we're building it: what's shipped, what's in progress, what's next, and where we still need help. Alpha is a model for how to build and contribute to the Potluck.
A 12-month sprint to build a public option for AI
We ship continuously, with key milestones along the way.
Next Milestone Release – 28 Oct 2026
Alpha v0.2 — Make AI Yours
Your language, your data, your rules. With this release we introduce new open models, personal accounts, safety updates, personal data storage, an agent service and a significant focus on language.
6 epics
- Partner needed
- In progress
Your Language
- Partner needed
- In progress
Your Language
With this epic we focus on ensuring that anyone can speak to AI in their language. We start with a model picker that detects your language and routes you to the highest-scoring model, then build workflows to contribute to existing datasets through Common Voice and Mozilla Data Collective, add new ones, and rate the models in an in-app evaluation tool.
1 contributor
- In progress
Your Rules
- In progress
Your Rules
With this epic we focus on making Alpha work the way you want it to from the first visit. Create an account, set your custom instructions, model and interface language in settings.
- In progress
- Experimental
Your hardware
- In progress
- Experimental
Your hardware
With this epic we focus on letting you run the model on hardware you own. Pair a laptop, phone or GPU board with your account, and the models on it appear in the model picker, with a label on every chat answered locally.
- In progress
Your Data
- In progress
Your Data
With this epic we focus on giving you control over where your files are stored. Every account starts with free storage in a Hugging Face bucket, and you can link your own bucket or choose a location for each file. Download your data as a ZIP at any time, or delete it in 2 clicks.
1 contributor
- Partner needed
- In progress
- Experimental
Your Agents
- Partner needed
- In progress
- Experimental
Your Agents
With this epic we are building the beginnings of an agent service for Alpha. It writes code, and with your permission, runs it in a sandbox and returns the result.
1 contributor
- In progress
Safety
- In progress
Safety
With this epic we focus on making Alpha safe to use. ShieldStral, from Mistral, screens messages and attachments, and Coop, from ROOST, gives human reviewers a place to work through what gets flagged. Both sit in a layered stack, so no single tool carries the load.
1 contributor
looking ahead
What's next with Alpha
Alpha is the system around an open model: your documents, your rules and your local knowledge. You can swap the model and keep the rest. Looking ahead, we focus on making that system fit the work you need done. We'll continue to prioritize inclusive access with smaller models and local compute.
4 possible focus areas
- Partner needed
- In progress
- Experimental
Shape Alpha Together
With this epic we give teams and communities control over how their own instance of Alpha behaves. The shared admin panel shows the system instructions, safety rules and model list behind an instance, editable from a dashboard or from the chat. Workspaces give a team a shared place to talk to Alpha, and later a space to share documents.
1 contributor
- Partner needed
Community Safety
With this epic we focus on safety that each community can shape. We study how federated platforms split moderation between the platform and its communities, publish starter safety recipes that a community can adopt and adjust, and put in place the security and legal review that a larger stack needs.
1 contributor
- Partner needed
Your Culture
With this epic we focus on helping you speak to AI in your culture as well as your language. A good answer reflects your dialect, your customs and what is available where you live. We start by measuring how models handle cultural context and by building ways for communities to bring local collections and guidelines into a conversation.
Put Your Data to Work
With this epic we focus on putting your data to work. Alpha v0.2 begins with controlling where your data is stored; next comes retrieval over your own documents, adapting a model to your field, and ways to use data that cannot leave your organization.
1 contributor
