Compared to other engines

These comparisons focus on model differences — the decisions you can’t configure away — rather than feature checklists, which date quickly and which a prototype would lose anyway. Corrections from the projects mentioned are welcome.

Draft

These summaries are honest but brief, written from public documentation of the other engines. They’ll be expanded — and fact-checked with people who know each engine better — before being treated as definitive.

vs. CRDT-native engines (Automerge, Jazz, Loro)

CRDT-native engines make all data convergent: any two replicas merge, automatically, without a server. datadata deliberately doesn’t — it keeps a server in charge of structured data and uses guards and staged previews to make conflicts explicit, reserving CRDTs for text, where silent convergence is what you want. You give up offline/P2P merging; you get server-validated schemas, simple reasoning about authority, and conflicts a human or agent reviews instead of an algorithm deciding.

vs. database-replication engines (Zero, ElectricSQL, PowerSync)

These engines sync a database — queries, rows, and partial replicas of something Postgres-shaped. datadata syncs documents: the unit of subscription, validation, and history is a JSON document, and the schema layer is documents too, not DDL. If your app is fundamentally relational with an existing database, they’re the natural choice; if it’s document-shaped collaboration with evolving schemas and agent participation, that’s the corner datadata is built for.

vs. event-sourcing engines (LiveStore)

LiveStore syncs an ordered event log and derives state by materializing events through application-defined reducers; datadata syncs state deltas (patches) and keeps the log as history rather than as the source of truth. datadata’s model asks less ceremony per change (no event vocabulary to design) and keeps the log readable without the application’s code — why we chose self-applying patches over events-plus-reducers — while LiveStore’s gives you richer replay and derived-state guarantees. Both are server-ordered and library-shaped — the closest philosophical neighbor on this list.

vs. collaboration platforms (Liveblocks)

Liveblocks is a hosted product with batteries — presence, comments, notifications — around collaborative state. datadata is a self-hosted library honing a smaller core: documents, schemas, sessions. If you want to buy collaboration UX, buy it; datadata is for building the data layer yourself, including the parts (agent staging, schemas-as-documents) that platforms don’t model.