{"id":537,"date":"2026-03-17T22:26:08","date_gmt":"2026-03-17T22:26:08","guid":{"rendered":"https:\/\/blog.aminalam.info\/?p=537"},"modified":"2026-03-18T10:50:37","modified_gmt":"2026-03-18T10:50:37","slug":"how-i-connected-1-million-people-to-internet","status":"publish","type":"post","link":"https:\/\/blog.aminalam.info\/?p=537","title":{"rendered":"When the internet was cut off, we helped 1 million people connect"},"content":{"rendered":"\n
\"\"<\/figure>\n\n\n\n

January 2026 was a breaking point for Iran.<\/p>\n\n\n\n

After massacring more than 35,000 protesters in the deadliest crackdown in the country\u2019s modern history, the IRGC imposed a near-total internet shutdown on roughly 90 million people. For days, almost nothing could enter or leave the country. At the same time, they sharply increased the price of cellular data. Iranians were already paying a premium for a censored internet, and then paying again for VPNs just to access the open web. When limited connectivity returned, the economics no longer worked: many people could not afford reliable circumvention tools, and the VPNs still available were often low-quality, unstable, or already detected.<\/p>\n\n\n\n

I knew this was the moment to do more.<\/p>\n\n\n\n

I could not be on the streets of Tehran. I could not physically stand beside the people risking their lives for freedom. But I could contribute with the skills I had: infrastructure, security, rapid execution, and the ability to organize technical systems under pressure.<\/p>\n\n\n\n

I launched a fundraising campaign and raised more than $1,000 in three days. With that money, I rented 30 servers across multiple regions and registered 30 domains with deliberately ordinary names\u2014domains that looked like harmless small-business websites rather than circumvention infrastructure. The goal was not just to deploy servers, but to design a network that could survive scrutiny, adapt quickly, and remain reachable under aggressive filtering conditions.<\/p>\n\n\n\n\n\n\n\n

My first attempts relied on more traditional VPN configurations. They failed.<\/p>\n\n\n\n

That failure forced a rapid rethink. Deep packet inspection and traffic analysis had become more sophisticated, and methods that worked in earlier waves of censorship were no longer enough. I needed to evaluate new protocol choices, model detection risks, compare deployment tradeoffs, and move fast. I used AI tools as part of that process\u2014not as a substitute for engineering judgment, but as a force multiplier. They helped me compress research time, pressure-test architecture decisions, and iterate faster on implementation plans while the situation on the ground was changing by the hour.<\/p>\n\n\n\n

From there, I redesigned the stack around more resilient, modern circumvention methods: VLESS routed through Cloudflare\u2019s proxy layer, Shadowsocks with REALITY-style obfuscation, and other configurations optimized for stealth, redundancy, and rapid rotation. I treated the system as an adversarial engineering problem: assume active interference, assume probing, assume protocol fingerprinting, and design for recovery.<\/p>\n\n\n\n

That meant:<\/p>\n\n\n\n