Coming in June: a report on how Iranians stayed online when the country almost disappeared from the internet

In June 2025, Iran tried something it hadn't tried before. Instead of yanking the country off the internet the way it did in 2019, authorities kept Iran looking online to the outside world while quietly cutting roughly 90% of international traffic. To BGP monitors, routes stayed up. Iran responded to pings. From the outside, things looked normal. Inside Iran, the global internet had effectively disappeared.

We surveyed nearly 900 people in Iran in the weeks that followed, in collaboration with ASL19. The full report comes out in June 2026. We want to surface one finding now:

Almost all of the resilience we measured ran through people, not through tools.

Seventy-six percent of people who knew about circumvention software had helped someone else use it. Tens of millions of Iranians played some kind of helper role during the shutdown: forwarding a config, telling a relative to "try this VPN," running a Telegram channel, configuring an aunt's phone at the kitchen table. Formal documentation played almost no role. Manuals were not consulted. People were.

That is a technology finding. It is also a media finding.

Large parts of the internet freedom work has spent more than a decade organized around a one-to-many delivery model: build a tool, ship it to users, count installs. Editorial work runs on the same logic: publish a story, push it through a channel, measure reach. The model is delivery.

Our research suggests the actual infrastructure is something else: Knowledge, trust, and tools move through dense, distributed, layered networks of ordinary people embedded in everyday relationships. Family, friends, colleagues, neighborhood phone shops. Tools get blocked. Servers get taken down. New protocols get throttled. The social network that routes knowledge and access around those failures is much harder to disrupt, and reconstitutes itself after each attempt at repression.

If that is the actual infrastructure, the strategic question changes. The question is no longer "how do we deliver our content or our tool." The question becomes: who are the people already doing the delivery on our behalf, and how do we build for them?

Different helpers have different needs. An information sharer who forwards a link needs something that fits in a chat. A setup helper sitting next to someone's phone needs a one-tap install with a working config bundled inside. An infrastructure enabler running a channel needs shareable subscriptions, ready-to-use configs, safer ways to distribute resources at scale. They are not the same person. The products built for them should not be the same product.

We'll release the full report June 1, 2026. It will be available here for download alongside companion pieces on what the helper frame means for product, for editorial, and for service design more broadly.

Until then, one question to sit with: if other people are partly serving as the last mile of your distribution, what would you build for them differently from what you build today?