Potluck · The idea
Intelligence people can own.
Everyone brings what they can. Nobody goes hungry.
AI should be something you can work with on your own terms.
People and organizations should be able to run useful models on infrastructure they control, inspect the work, and carry their knowledge with them. Potluck is building toward that future, starting with local AI and machines you own.
Our current direction is Company Brain: an evolving system for working with an organization’s sources, reviewing outcomes, and retaining useful knowledge. It is in development; see the roadmap for what is available today.
An open foundation.
The Potluck core is MIT licensed and free for personal and commercial use. Companies can use the core without payment. We intend it to support complete, useful workflows, including work on owned compute and reviewed knowledge as those capabilities mature.
Potluck is a founder-owned company. Our planned commercial offering is separate: additional organization administration, operations, and support as those capabilities are delivered. See pricing for the current direction.
Basic privacy, security, and the ability to retain and export your work belong in the core, not a premium tier. Some of those capabilities are still being built; we will be clear about what is available.
A network people can help shape.
Alongside work on connected machines that organizations own, we want people to be able to contribute compute by choice and, over time, help shape a shared network. The public contributor network and its governance are not launched. A member-operated network is a long-term ambition, separate from ownership of the company and from a customer’s control of its own deployment.
Participation should be understandable, voluntary, and grounded in the realities of the machines involved. A remote machine processes the request it receives; routing choices and their limits should be clear.
Read how we think about governance and the technical approach.
What we believe
- AI should work for the people who use it.
- Useful infrastructure should be open to inspection and use.
- People should make informed choices about their data and compute.
- Evidence and human review matter more than confident-sounding answers.
- Honesty about what works today matters more than hype.