Local-First / Mobile

Offline-first client sync

Field / mobile clients work offline, then converge on reconnect by exchanging deltas — no server merge pass, no last-writer-wins data loss, a fraction of the bandwidth.

CONVERGEScoordination-free convergence + ~92% less data moved

Convergence (correctness)PASS

replicas reach identical state with no coordinator — byte-identical across 1/2/4/8 banks (0x17e9fe29b5dc0aad)

any node, any order, after any partition → identical merged state. The same property, on silicon.

Data moved (bandwidth)PASS

2048 B/tick full-state → 160 B/tick delta (measured) = ~92% less

at ~5% of state changed per sync — scales with change density, not a universal multiplier

Bandwidth is the representative delta tick measured on the AX7020 ATOMiK adapter: 160 B moved vs 2,048 B of full state, with exact 64-bit LOAD/ACCUM/READ round-trips on a falsification-tested harness. The convergence guarantee is the same byte-identical, order-independent merge. Convergence is the same order-independence proven on the AX7020 (2026-06-21); reproduce with abench 65537 0xDEADBEEF0BADF00D.

Run this on our silicon → sign conditional on your bar

This is you if

You build local-first / offline-capable apps (mobile, field service, retail) where clients work disconnected.
Clients accumulate counters, tallies, and inventory deltas offline and must converge on reconnect.
You use (or are weighing) a CRDT library or a server reconciliation pass, and you pay for it in bandwidth and merge complexity.

The workload

Each client holds a replica of shared state: counters, tallies, and signed inventory / usage deltas.
Clients mutate state offline for minutes to hours, then reconnect over a metered or flaky link.
On reconnect, replicas must converge to identical state regardless of which client synced first.
No client is authoritative; reconnection order is arbitrary.

Today — coordinated / full-state sync

Converging offline clients means a CRDT library (per-field metadata that grows, and you still ship state or a fat op-log) or a server-side reconciliation / last-writer-wins pass that silently drops concurrent edits. And each sync tends to ship the full record set or a growing op-log, not just what changed — expensive on metered, intermittent mobile links.

With ATOMiK — order-independent, coordination-free

Clients exchange delta-state. Because the merge is order-independent and self-consistent — byte-identical no matter which client's deltas arrive in which order — replicas converge with NO server reconciliation and NO per-field CRDT metadata, and each sync ships only the delta, not the full record set.

Coordination-free convergence on reconnect — no server merge pass, no last-writer-wins loss of concurrent edits.
Any client, any reconnection order, after any offline gap → identical state (the same byte-identical property proven on silicon).
Ships the delta, not the full record set — measured 160 B vs 2,048 B (~92% less) at ~5% change density.
No growing per-field CRDT metadata on the wire — the delta IS the merge.

Want to test a different rate or operator? Open the interactive benchmark tool · Representative archetype, not a named customer. Numbers are measured engine facts; the value is lock-elimination + the byte-identical guarantee.