Local-First Conf 2026 argued that owning your data is not the finish line

The local-first crowd met in Berlin last month, and the most pointed conversation of the event was aimed squarely at itself.

Local-First Conf 2026 ran from 12 to 14 July at Festsaal Kreuzberg in Berlin. Two days of talks, then a third day billed as Lab Day, hosted by Ink & Switch, the research lab that gave the movement its founding essay back in 2019. Recordings from the first two days are now public.

The organisers set this year's theme as, in their words, "user empowerment in an age of fluid software", and widened the call for talks beyond the usual sync-engine internals. That widening turned out to matter.

The panel that picked a fight with the premise

The session that got the most attention afterwards was a panel titled "Data Ownership Beyond Local-First". Its argument, roughly: keeping a copy of your bytes on your own laptop is necessary and nowhere near sufficient. Real ownership, the panellists contended, has to mean practical user power, interoperable infrastructure, and governance models that go beyond one person controlling one account.

That is a genuinely uncomfortable thing to say at a conference named after the technique it's criticising, which is the main reason it's worth repeating. A movement that can only produce applause lines about the cloud being bad is a marketing campaign. One that can stand on its own stage and say "this is the ceiling of what we've built" is a research programme.

The critique lands. You can hold every byte of your data locally and still be stuck, because the format is undocumented, or because the app that reads it is abandoned, or because the three other people you work with are all on something else. Ownership without interoperability is a filing cabinet with no key.

The plumbing quietly got good

The reason this argument can happen at all is that the underlying technology stopped being the bottleneck. Automerge, the CRDT library out of Ink & Switch, shipped a 3.0 release with dramatic reductions in memory use and document size, which turned "local-first is elegant in theory" into "local-first runs on a laptop". Alongside it there is now a real field of SQLite-based sync engines and purpose-built frameworks, plus growing platform-level support from the operating system vendors.

None of that is news to anyone who follows the space closely. It is news, in the sense that matters, to everyone else. Five years ago building an app that worked offline, synced properly, and never held your data hostage meant writing your own merge algorithm and regretting it. Now it means picking a library.

We covered the underlying case for this approach in Digital Sovereignty, and the practical argument for working without a connection in Offline by Design. The short version has not changed. Software that keeps the primary copy of your work on your device is faster, works on a train, and cannot be switched off by a pricing decision.

Where we think the movement is still soft

Two things, and we say this as people building in this space rather than watching it.

First, collaboration remains the hard part and everyone knows it. CRDTs solve merging text and structured documents well. They do not solve the human question of who is allowed to see what, which is a permissions problem, and permissions want a server. Most local-first products end up with a server anyway, at which point the honest framing is "local-first with sync" rather than "no cloud", and a lot of marketing in this category is not that honest. We tried to be straight about the tradeoff in Sync Without the Sacrifice.

Second, the movement's rhetoric is much better developed than its answer to abandonment. If a local-first app dies, you have your files, which is a real improvement over having nothing. But you still need something that can read them. Documented, boring formats matter more than clever ones, and there's less glamour in writing a format spec than in writing a merge algorithm.

Why any of this matters if you just want a task board

Because the alternative has a track record. Six weeks from now Microsoft is switching off Project Online, taking a decade of hosted project data with it on a schedule the customers did not set. That is not a scandal. It's just how hosted software ends, every time, and the only variable is how much notice you get.

A tool that keeps your board on your machine and exports to a format you can open in a text editor makes that whole category of event boring. Your work does not depend on our funding round. It depends on your hard drive, which is a dependency you can actually inspect.

Axtio's free tier stores everything in your browser and never sends it anywhere. Pro adds sync for people who want it, and the export is plain JSON either way. That's not a philosophical position so much as a refusal to build a hostage situation and then charge rent on it.

The Berlin panel was right, though. Holding the file is the floor, not the ceiling. The next argument, the one about interoperability and governance and what happens when a format outlives its author, is the interesting one, and the field has barely started it.

Sources

Local-First Conf 2026, Berlin, 12 to 14 July 2026, including the "Data Ownership Beyond Local-First" panel. Automerge 3.0 release notes, automerge.org. Founding essay: Kleppmann et al., "Local-First Software: You Own Your Data, in spite of the Cloud", Ink & Switch.