When we started building the platform, the industry default was a handful of large regions and a CDN bolted on top. It worked — until your users were in Jakarta and your database was in Virginia.
We made a bet early: every workload should run as close to the user as physics allows, and the platform — not the developer — should be responsible for making that true.
The parts nobody warns you about
Edge-first sounds like a deployment detail. It isn't. It changes how you think about state, sessions, and consistency. A request served in Tokyo and a request served in Frankfurt can't both assume the database is next door.
Our answer was a replicated read layer with regional write-forwarding. Reads are local everywhere; writes converge on a primary and fan back out in under 300 milliseconds globally.
What it bought us
Median global latency dropped from 210ms to 34ms. But the real win was simpler mental models: developers stopped writing region-aware code, because there were no regions to be aware of.
If you're weighing the same move, our advice is simple — do it for the developer experience, and let the latency numbers be the bonus.