← All tech signals
Infrastructure

Data Stays Home

A new wave of software keeps a person’s data on their own device by default, treating the cloud as a backup rather than the source of truth.

What is it?

Sync tooling has matured enough that a device, not a server, can hold the authoritative copy of a person’s data. The pitch is resilience: work survives a dropped connection or a vanished vendor. Whether it can match cloud tools on live collaboration is still open.

Why now?

Years of shutdown-prone cloud services taught developers that data living only on someone else’s server can simply disappear. Conflict-resolution libraries reached production maturity at the same moment, making device-first storage practical rather than theoretical. The scarce thing shifts from server uptime to software that survives without one.

Who’s building it?

Sync-engine startups plug conflict-resolution libraries into ordinary apps; developer conferences now teach the pattern.

Sync-engine vendorsCRDT library maintainersApp developersCloud incumbentsConference organisers
The full read
Why now, who is building it, where the market goes, and what stays unsettled.
Get weekly →
01 · The market
Who’s in it
Builders — sync-engine vendors and the CRDT libraries underneath them.
Buyers — developers who want an app that still works offline, and compliance teams.
Gatekeepers — cloud incumbents deciding whether to absorb or ignore the pattern.
What’s at stake
Cloud vendors built a decade of business on being the source of truth; local-first quietly demotes them to backup.
A category of sync-engine infrastructure companies is forming right now.
File formats must survive years of migration, not one release.
What’s unsettled
Whether local-first apps ever match cloud tools for real-time collaboration.
Whether sync engines stay a niche tool, or become default.
Whether compliance teams treat on-device storage as a real answer, or a workaround.
02 · How the demand matureda progression, not a timeline
Tooling
CRDT libraries and sync engines reach production maturity, not just research demos.
Adoption
A dedicated conference track signals developers are now building on this by default.
Trust
Years of vanished cloud services make device-first storage read as insurance, not a downgrade.
Infrastructure
Sync-engine vendors become a standing category developers reach for, not a workaround.
03 · Where the market goesthe forks, not a prediction
→Standard or workaround — decided by whether sync engines close the real-time collaboration gap that cloud tools still win on.
→Compliance teams may adopt local-first fastest, since the authoritative copy never leaves the device.
→Cloud incumbents could absorb the sync-engine vendors outright rather than build the pattern themselves from scratch.
→File-format longevity matters more than sync speed once data must survive years of migration.
Underneath it all: does keeping data on-device actually protect a person, or just move the same fragility to a device that also fails and gets lost?
04 · The Edge Shift deskadjacent markets, moving now
Hardware Before Habit
A chip built for on-PC AI.
A chip built for the PC.
AP and Reuters via Al Jazeera
Intelligence That Stays Home
Models small enough to run on a phone.
A model that fits a phone.
The Next Web
Hyperlocal AI
On-device models are moving closer to mainstream phones.
Apple is in talks with PrismML over iPhone-sized AI models.
CNBC
05 · The Wire
01
FOSDEM 2026 ran a developer room on local-first, sync engines and CRDTs
FOSDEM ↗
02
Academic keynote: local-first apps must tolerate malicious clients
Martin Kleppmann (SRDS 2026 keynote) ↗
The panel on this signal
They agreed on
–The shift is driven by trust lost to shutdown-prone cloud services, not by raw technical preference alone.
–Sync-engine tooling has crossed from research demo to production maturity.
–Compliance and data-residency questions get a cleaner answer once the device holds the authoritative copy.
The ML systems engineer disagreed on
–Conflict resolution across simultaneous edits from multiple users is a substantially harder distributed-systems problem than single-device sync; current local-first demonstrations have not yet shown this working at the concurrency levels cloud-native tools already handle.
Know more about the method →
✦How this signal was built
01 · Panel
A Delphi panel of 4 AI analysts, each reasoning independently.
→
02 · Debate
They challenge and revise each other’s reads over two rounds.
→
03 · Evidence
Every claim is tied to a cited, linked source on The Wire.
→
04 · Synthesis
A moderator writes it into one read, disagreement and all.