Local-first
summary
Actyx explains seven years of running local‑first software in factories: devices (Raspberry Pi, tablets) keep single-writer logs replicated peer‑to‑peer, merged deterministically with Lamport clocks so systems remain available despite no central DB; a TypeChecker (compiles to Wasm) and a machine runner enforce session‑style protocols; live deployments include a 10‑robot logistics fleet and devices with ~20 GB storage; problems include schema evolution, UX for connectivity, and a proposed version‑everything approach plus use of Cambria for data transformations.
AI-written summary · about
chapters (13)
- 0:00Local-first systems at Actic
- 1:27Factories need never-stop software
- 3:58Single-writer principle and event logs
- 5:45Eventual consensus with global types
- 7:14Customer feedback and mindset shift
- 8:58State machine diagrams in workflows
- 10:44Software evolution and schema changes
- 13:39Q&A on schema compatibility
- 17:54Coordination without consensus
- 20:20Eventual consistency and compensating actions
- 22:46Programming challenges and open problems
- 24:08Log size and replication
- 25:54Closing remarks
related talks
24mGoing Local-First: A Primer
41mThe UI/UX Benefits and Trade-Offs of Local-First Apps
18mRethinking Apps with Local-First Architecture
01h 04mMeshing Robots and Humans with Distributed Apps
35mLocal-First Distributed Configuration
33mProgramming Support for Local-First Software
26mLocal First Svelte
32mNative-Grade Web Apps with Local-First Data