Briar

For five months in 2026, while an internet blackout cut somewhere between 85 and 92 million Iranians off from the outside world (the estimates disagree), Briar was one of the few messaging tools reported to keep working, and its Farsi manual circulated widely. That moment made Briar briefly famous, eight years after a small team led by Michael Rogers built the app with a narrow audience in mind from day one: activists and journalists whose adversary is a surveillance state, not the average person picking a chat app. Calling it a mesh is technically wrong, too. Briar only ever links two people who already know each other, one connection at a time, carried over Bluetooth, Wi-Fi, or Tor, never through a stranger’s device. That limitation is intentional, not a gap, and it’s the reason security researchers vouch for the app in the first place. It’s also why the project stayed a rounding error next to bigger messaging tools, roughly 2.6 million downloads cumulative as reported in early 2026, until a national shutdown gave it a reason to matter beyond a small circle of experts.

Security first, adoption second, by design

Every awkward edge in Briar is a security decision wearing the costume of a usability flaw, and that single fact explains both the trust and the ceiling. Look first at what the app refuses to hold: no phone number, no account bound to a person, no server permitted to watch traffic pass. Those refusals are why no metadata trail accumulates to be seized, and equally why a newcomer gets no frictionless way in. The security pedigree behind the refusals is not decoration: a formal Cure53 audit dated March 2017, an all-Tor path for internet sync that keeps every server blind, reproducible builds that let a doubter compile the source and confirm the release matches instead of taking it on faith, a public beta having come and gone between that audit and the 1.0. This is the evidence set that made “the tool experts recommend” the standing press default.

The friction is the identical list read from the far side. Getting started demands that two people already run Briar and, at first, stand in one room to trade a QR code; the briar.link exchange over Tor that shipped with version 1.2 in June 2019 eased the in-person part and left the both-installed requirement exactly where it was. iOS is absent by choice rather than neglect, the team’s own reasoning being that Apple’s background-execution limits make a dependable client impractical. Delivery stalls whenever two people aren’t online in the same moment, a gap the 2023 Briar Mailbox narrows by drafting a spare Android phone into an always-listening relay. Read the audit trail and the adoption ceiling off one page, and they turn out to be written in the same ink.

What changed, 2025 to mid-2026

Development didn’t pause for any of this. Bengali support landed in the Android app in March 2025 (version 1.5.14), a round of maintenance late that year cleared up Bluetooth reconnection bugs, and version 1.5.17 shipped March 12, 2026. The desktop client reached 0.6.5-beta on February 20, 2026, still labeled beta after roughly four years, and it’s missing something the phone app has: there’s no Bluetooth or nearby-device discovery on desktop at all, so those clients can only sync through Tor, meaning the offline capability that defines Briar on mobile isn’t there on a computer at all. Iran wasn’t Briar’s first brush with a real crisis, either: back in the days right after Russia’s February 2022 invasion, it climbed into the top five of Ukraine’s Google Play charts. This time, running from January 8 through May 26, 2026, the Iran shutdown put Briar on Hacker News’s front page and drew coverage from Cybernews. A separate claim that Briar saw meaningful use during Uganda’s 2021 election shutdown shows up repeatedly in roundup articles, but nothing traces back to a primary source, so that one stays flagged as unconfirmed. Through all of this, no independent group has re-audited the code since Cure53’s original 2017 pass, and an iOS release remains off the roadmap entirely.

No revenue, no growth pressure

There’s no venture capital behind Briar and no revenue target to hit, which changes what “success” even means for the project. Funding comes from grants: the Open Technology Fund contributed $361,100 by 2018, NLnet and NGI Assure back the desktop client specifically, and Access Now, Small Media, and individual donors round out the rest. Without shareholders demanding growth, Briar has outlasted every for-profit competitor that tried to serve the same niche, without ever needing to dominate the wider category the way a startup would.

Where it stands against this project

Briar’s eight-year run without a public security scandal, and its status as the tool researchers reach for by default, is proof that getting the fundamentals right pays off over time; matching that audit trail is the bar this project needs to clear before making similar safety claims. It’s also proof that rigor alone caps growth: 2.6 million downloads spread across eight years looks small against Bridgefy’s cumulative 12 million-plus, a widely repeated number that nobody has pinned to a specific date. The bigger structural gap is architectural. Briar only ever connects two people who are already contacts, one hop at a time; it was never built to relay through a stranger’s device the way a real multi-hop mesh does, and that’s the core distinction from what this project does. Windows relay hubs provide always-on coverage that Briar’s Mailbox only replicates one device at a time, and a dashboard turns those relays into something operators can manage, not merely leave running. There’s also a delivery gap: this project confirms message delivery, ack and relay, as a baseline even between people who haven’t connected as contacts yet, something Briar has no mechanism for at all. And platform coverage differs: this project runs cross-platform including Windows, while Briar stays confined to Android for full functionality with a Tor-only desktop beta and no iOS client whatsoever.