SE Radio 716: Martin Kleppmann Local-First Software

In this episode of Software Engineering Radio, host Adi Nadayan interviews Martin Kleppmann, associate professor at the University of Cambridge and author of the seminal book 'Designing Data-Intensive Applications'. The conversation centers on 'Local-First Software', a paradigm that combines the offline resilience and data ownership of traditional local software with the real-time collaboration features of modern cloud apps. Kleppmann explains how Local-First software stores data locally on users' devices while enabling seamless sync across devices via sync engines like AutoMerge, which uses Conflict-Free Replicated Data Types (CRDTs) to automatically resolve conflicts. He emphasizes the benefits for both users—such as instant responsiveness, offline functionality, and data sovereignty—and developers, who can simplify their stack by focusing only on the frontend, as the sync and backend logic are abstracted away. The discussion also covers real-world applications, challenges like conflict resolution and search functionality, and the growing ecosystem of tools and open-source projects like AutoMerge and Keyhive. Kleppmann highlights that Local-First is not suitable for all software—especially systems requiring a single source of truth like banking—but excels in creative, collaborative tools like note-taking apps, spreadsheets, and design software. He concludes by noting that the movement is gaining traction, driven by concerns over cloud dependency, geopolitical risks, and the rise of AI, which can be integrated into the Local-First model as another collaborator. The episode ends with practical guidance for developers wanting to get started, including resources like the LocalFirst.fm landscape and AutoMerge’s documentation. Key takeaways include: 1) Local-First software enables real-time collaboration with full offline support by keeping data on the user’s device; 2) Sync engines like AutoMerge abstract away complex distributed systems problems, simplifying development; 3) CRDTs allow automatic, conflict-free merging of edits across users; 4) The model is ideal for user-generated content apps but less suitable for systems requiring centralized control like financial software; 5) Local-First enhances privacy, resilience, and developer productivity; 6) It aligns well with AI collaboration, treating AI agents as co-editors with version control; 7) Tools like Keyhive enable decentralized access control and end-to-end encryption; 8) Developers should consider starting with Local-First by default, only moving data to the cloud when necessary.