Heroku co-founder and Ink & Switch founder Adam Wiggins makes the case for local-first software architecture as a necessary correction to the all-cloud paradigm. He argues that CRDTs and sync engines have matured enough for production use, citing Linear as a prime example of local-first done right. Wiggins envisions extending Git-like version control primitives beyond code to documents and spreadsheets, and predicts a hybrid AI future where small local models handle 80% of routine tasks while large cloud LLMs handle compute-intensive work. He also discusses the growing Local-First Conference community in Berlin and the importance of balancing user agency, data ownership, and offline capability with the collaboration benefits of the cloud.
Table of contents
TranscriptSoftware for More Productive Humans [ 03:22 ]Why Local First Matters [ 06:42 ]The Balance of Cloud and Local Collaboration [ 18:27 ]How Will The AI Agents' Space Evolve? [ 22:04 ]The Local First Community is Growing [ 25:56 ]About the AuthorQuestions this post answers
What is local-first software architecture and how does it differ from cloud-native?
Local-first software prioritizes storing and processing data on the user's own device while still supporting cloud-based collaboration, rather than routing every operation through centralized servers. It relies on CRDTs (Conflict-free Replicated Data Types) and sync engines so that local copies of data update instantly and sync with the server in the background, giving users offline capability, lower latency, and greater data ownership without abandoning cloud collaboration entirely. For teams weighing cloud-native versus local-first trade-offs, daily.dev surfaces architecture discussions like this one.
How does the ticket tracking tool Linear achieve such fast performance in the browser?
Linear stores copies of tickets locally, likely in IndexedDB, so that clicking or updating a ticket writes instantly to local storage rather than waiting on a server round-trip. A background thread continuously syncs that local data with the server, using CRDT-based conflict resolution techniques developed over roughly 15 years of research, giving the interface a genuinely fast, non-optimistic UI experience. Developers researching sync-engine patterns like Linear's can track this space through daily.dev.
What role will local AI models play alongside large cloud-based LLMs in the future?
Small, high-performance local models are expected to handle around 80% of routine productivity tasks, with large cloud-based LLMs reserved for high-compute needs, mirroring the local-first pattern already emerging in data sync. This would let developers keep working with AI-assisted coding even with restricted connectivity, similar to how local-first apps degrade gracefully offline rather than losing all functionality. Engineers planning around hybrid local and cloud AI workflows can follow this evolving trend on daily.dev.
Share this post