Convex: Life Without a Backend Team
summary
Convex offers a serverless database that runs deterministic JavaScript transactions in V8 on a Rust-backed MVCC engine, using timestamp-assigned snapshots and range-set conflict detection for serializability. It ships client-bound function source into the DB, provides automatic subscriptions over websockets (tokens encode read sets) and consistent caching, and exposes incremental schema inference (union types) with just-in-time migrations. Storage targets object stores with Parquet exports. Testing uses QuickCheck-style fuzzing and a virtualized runtime for deterministic interleaving; an early beta is rolling out.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 3:25Archetypes of Engineers
- 6:40Serverless Computing
- 10:08Primitives
- 12:33Convex
- 16:42Running transactions
- 18:17Transaction model
- 20:20Read sets
- 21:35Subscriptions
- 22:21Automatic Subscriptions
- 24:48Partial Results
- 26:43Caching
- 28:12Complications
- 29:20Composable abstractions
- 33:47Incrementality
- 36:12Schema Migration
- 38:05Type Understanding
- 42:12Escape Hatch
- 43:05Column Store
- 47:14Testing
- 51:03Integration Testing
- 53:25Production Validation
- 55:41Questions
related talks
32mExtending reactivity to the server with Convex
08mWhen Worlds Collide: Frontend vs Database
51mBuilding a Realtime "Backend in a Box" with PostgreSQL and Vue
34mEnabling data at massive scale with Next.js
32mReactive Data on the Full Stack
29mHow to develop your APIs without a backend
40mScalable Apps without the Hassle with Next.js Vercel & FaunaDB
26mTransactional Serverless with DBOS