Roadmap

On the road to Beads 2.0

Here's what we're planning for Beads for the foreseeable future. All of this represents our hopes and dreams, which all of you get to help make come true! Don't be shy about sharing your thoughts about the future of Beads.

On the quality front, the big near-term focus is CI coverage. The current Dolt server implementation currently has tests that don't run in CI and routinely fail even on main. We plan to get all of those tests running, and then systematically fix the broken ones, starting with the tests most critical to Gas City. We'll also be adding a Gas City dependency test to Beads CI, so that any Beads change that would break Gas City fails loudly before it lands.

While that's happening, we're finishing the Dolt server v2 implementation. Once it's ready and well-tested, you'll be able to opt in and report any issues. We'll provide an explicit migration command for moving existing v1 repos to v2, and then move to v2 by default over time.

With v2 as the foundation, we'll turn our attention to Beads schema v2 to simplify and clean up the database schema and make intentional decisions about what belongs in a 2.0 release. That unblocks a general code hygiene pass to remove dead code and unnecessary cruft, and simplify the code base as much as possible.

We're also planning to decouple the Beads database from the Git repository. Today Beads requires one database per Git repo, which has its benefits but also real limitations. Adding support for a shared, decoupled database will make cross-project Beads usage a first-class feature.

And finally, we're adding application-level usage metrics to Beads we we can understand what features people are using and not. Metrics can ALWAYS be disabled via a simple gc command and we'll be explicit about capturing them when you first run gc so you can opt out immediately. The metrics won't touch your context or your work, but they do help us to improve the product.

All of that adds up to Beads 2.0: extremely stable, always "just works" out-of-the-box, and simple enough that users and agents get a consistent experience without needing to think about the implementation details that make the current 1.x tricky at times.