The case for local-first architecture in a cloud-saturated market

Long-standing cloud-native paradigms are being challenged by an architectural shift toward local-first software, a model that embeds data and compute on the user’s device while treating the cloud as a secondary synchronization layer. Adam Wiggins, co-founder of Heroku and Ink & Switch, argues that the maturation of Conflict-free Replicated Data Types (CRDTs) has moved this approach from academic theory to a viable production strategy for enterprise tools.

The strategy addresses two primary inefficiencies in modern software: high-latency network roundtrips and the lack of robust version control in non-code environments. By adopting primitives similar to Git—enabling branching, merging, and diffing for documents and spreadsheets—developers can offer users offline capability and instantaneous response times. This transition is not a total rejection of centralized infrastructure but a realignment where the device handles the majority of state changes and the cloud manages broad distribution.

Emerging AI integration patterns support this local-bias. As small, high-performance language models become more capable, the architectural consensus is shifting toward running routine tasks on local silicon. Large, cloud-based models are increasingly reserved for high-compute intensive operations, a hybrid approach that allows for greater data ownership and reduced operational costs.