Urgent.News

What's breaking now, across thousands of outlets.

AI

A small search win: Google's AI Overview cited my Figma multiplayer teardown

I write about production system design on Buildopsy. Recently, I published a teardown of Figma’s multiplayer architecture: how edits move between browsers and servers, how conflicts get resolved, and where the system’s scaling costs come from. I later searched for “how Figma’s multiplayer technology works.” Figma’s own engineering post appeared in the results. In the AI Overview below it, I…

A teardown of Figma's multiplayer architecture was published on Buildopsy by the author, who writes about production system design. When searching for "how Figma's multiplayer technology works," the AI Overview included citations from Buildopsy alongside explanations of optimistic rendering and the server's role in ordering edits. The author saved a screenshot of this result, as it was unexpected to see their article cited there.

The AI Overview mentioned two key practices in Figma's system: optimistic rendering, where a browser applies edits immediately, and the server's role in ordering changes; and a server-owned document ordering edits and broadcasting the accepted version to other clients. The conflict resolution method is more specific than "the server wins," with the server handling same-property conflicts using last-writer-wins, and tree changes undergoing an additional validation step to prevent cycles.

This design choice is specific to Figma's product needs, allowing a design canvas to tolerate some conflicts that might be problematic in a collaborative text editor.

Furthermore, Figma's recent Rust rewrite addressed the issue of a slow document stalling unrelated documents sharing a TypeScript worker. By moving document operations into a Rust child process, while Node.js continued handling the network layer, serialization became more than 10 times faster, as reported by Figma. The author wonders about which events need durable ordering and which can be dropped when they are stale, presenting a modeled scenario of 1.5 million concurrent users sending two coalesced cursor frames per second, resulting in 3 million ephemeral frames per second.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

Build a Minimal AI Rank Tracker in Python

If you've ever typed "best X for Y" into ChatGPT to see whether your product shows up, you've already done AI rank tracking, just with a sample size of one.

  • Minimal AI rank tracker built in Python
  • Sends buyer-style prompt to multiple search engines
  • Measures mention rate with confidence interval

More from Tuesday 6 October →