The Architecture of Instant: How Linear Eliminates Latency

The speed of a web application is often defined by the time a user spends looking at a loading spinner. While a traditional CRUD application takes about 300ms to update an issue, Linear operates in a few milliseconds. This performance gap is not the result of a single silver bullet, but rather a fundamental shift in how data moves between the client and the server.

The core of this speed lies in inverting the traditional relationship between the browser and the database. In a standard web architecture, a user action triggers an HTTP request, the server queries a database, and the browser waits for the response. This cycle creates a bottleneck where the network cost is measured in hundreds of milliseconds.

Linear bypasses this bottleneck by using a database in the browser via IndexedDB. When a mutation occurs, it applies locally first. The application then asynchronously pushes these changes to the server, which broadcasts deltas back to other clients via WebSocket.

This approach changes the developer's primary objective. To build high-performance software, the goal is to hide network requests from the user entirely. By treating the network as an asynchronous background process rather than a blocking foreground requirement, you eliminate the need for the loading states that define the modern web experience.

The implications for software engineering are clear: the network is your greatest enemy. If you can design a system where the UI reads from a local source and handles synchronization in the background, you move from a reactive, latency-dependent model to a proactive, instant-feel interface.

When building your next platform, ask yourself: how many of my user's actions are currently waiting on a network round-trip?