Local First Write
Committing to local storage first and treating the server as an eventual replica, which makes the interface responsive and connectivity an optimisation.
The conventional arrangement makes the server authoritative: the client sends a request, waits, and updates its display on success. Every action therefore costs a round trip, and with no connection there is no action.
Local-first inverts it. The write commits to a local store immediately, the interface updates from local state, and synchronisation happens in the background when a connection exists. The application is fully usable offline, and it is faster when online, because nothing waits for the network.
The price is that the client and server can diverge, which introduces two problems that must be designed rather than discovered. Conflicts occur when the same record changed in both places, and require a resolution strategy chosen with an understanding of what losing a change would mean. And server-side rejection — a validation failure, a permission change, a business rule — arrives after the user has already seen their action succeed, so there must be a designed way to surface and reverse it.
The related discipline is to validate as much as possible locally, so the number of actions the server can reject is small and the cases that remain are genuinely exceptional.