There's a new Tanstack package 👀

TanStack DB lands as a bid to fix a long-running pain point in React Query: optimistic updates. The core claim is that “optimistic updates in TanStack Query kind of suck,” because the usual approach—invalidating queries and refetching—forces extra network round trips and leaves users staring at loading states. Even when developers try to do true optimistic UI, the mechanics get brittle: query keys are just strings/arrays, type safety doesn’t stay bound to the data shape being updated, and the client ends up recreating server logic to predict what the backend will do. As apps scale—adding filters, search, and multiple places where the same data is queried—those optimistic assumptions can drift, break, or require increasingly complex client-side glue.

Electric SQL enters as the practical backdrop for why this matters. Electric’s goal is a sync layer on top of Postgres that replicates only the relevant subset of data to each client, pushing changes out in real time. That requires serious infrastructure work: row-level security for authentication, a client integration layer, and an optimistic/latency-hiding strategy that works cleanly with React Query-style data fetching. Electric has its own server-side engine (including an Elixir backend and a “PGite” library that brings Postgres-like querying into the browser via WebAssembly), plus a TypeScript client that models database “shapes” and streams updates. But Electric SQL React hooks can get complicated when developers need optimistic behavior, because they often have to merge “pending mutation results” with “query results” inside render logic.

TanStack Optimistic—now renamed to TanStack DB—tries to provide a more coherent local state model for optimistic transactions. The pitch frames TanStack DB as a reactive, normalized transactional state engine that extends TanStack Query with “collections,” “live queries,” and “optimistic mutations.” Instead of invalidating and refetching, the library overlays local optimistic mutations on top of immutable synced data so UI updates can appear immediately, then be reconciled when the real sync data arrives. The API also leans heavily on Electric’s concepts: developers define collections tied to a query function, specify a primary key so updates can be granular, and provide a schema (e.g., Zod/Valibot/Arbitrary typing). Optimistic mutations run inside a transactional “mutator,” and live queries subscribe to changes affecting the shaped data.

Still, skepticism runs through the critique. The optimistic layer appears tightly coupled to Electric’s “shape” model, and the implementation details leak into application code—especially around how transactions and mutation logic are wired to collections. The reviewer argues that this makes the ergonomics less “TanStack-like” than React Query’s simple async-function model, and more specialized: backend endpoints and data shaping must match the library’s expectations. The result is a tradeoff—more reliable optimistic behavior in the Electric-centric world, but a potentially confusing or less portable developer experience for teams that don’t already structure their systems around shapes and collections. In short: TanStack DB may offer a better optimistic foundation, but it also signals a shift toward a more opinionated sync/data model than many React Query users are used to.