Cyren

Inspiration

We've all seen the images from Hurricane Katrina: a multi-billion dollar infrastructure crumbling in hours. Rescue teams were left deaf, blind, and unable to coordinate simple evacuations. When the grid goes down, our modern world evaporates. No Wi-Fi. No signal.

First responders are forced back to 1940s technology: the analog walkie-talkie. Today, they're still manually relaying messages using the phonetic alphabet ("Alpha, Bravo, Charlie..."). This is slow, error-prone, and dangerously inefficient when seconds count. Misheard coordinates send rescue teams to the wrong house, and unverified distress calls lead helicopters into traps.

We asked ourselves: In the age of AI, why are we still screaming into the walkies? Our motivation was simple: replace the voice in the walkie-talkie with intelligence. We wanted to build a system that guarantees trust and clarity even when the entire internet is blacked out.

What it does

Meet Cyren: an AI-Native, offline communication mesh that transforms a standard smartphone into an intelligent, cryptographic tactical radio.

It uses Google Gemini Nano for on-device semantic triage. It literally listens to the chaos of a disaster scene, filters out the noise, and extracts critical data into a tiny, bulletproof 24-byte semantic code.

This hyper-compressed code is then blasted over standard radio waves or even speaker-to-microphone using a custom FSK acoustic modem, requiring no Wi-Fi or cellular signal.

For security, every single packet is cryptographically signed and encrypted offline using Solana's Ed25519 keys. This guarantees that only trusted responders can decode the message, while looters hear only static. Once connectivity is restored, verified logs are synced to the Solana Devnet, creating an immutable, censorship-resistant history of every SOS call.

How we built it

We built Cyren as a lean Turbo/PNPM monorepo, blending a sleek, high-contrast React/Vite frontend with a powerful backend.

Our "AI Brain" uses Google Gemini to semantically compress voice data into a strict JSON format. This compressed data is transmitted via a custom Web Audio API modem using the Goertzel Algorithm for precise frequency detection.

To ensure offline trust, we used @noble/ed25519 for cryptographic signing and verified message logs on the Solana Devnet using the SPL Memo program.Challenges we ran into

Our biggest headache was the "No-Wifi" logic gap. We quickly realized that relying on a Cloud API completely contradicted our core disaster scenario. This forced a critical architectural pivot to Edge AI (Gemini Nano)—the intelligence had to live on the silicon, not the server.

We also battled the laws of acoustic physics! Background noise, echoes, and audio bleed caused frequent data corruption. We had to implement a robust CRC-16 checksum to guarantee 100% accuracy for any arriving message. It was a fight against noise, but we won!

Accomplishments that we're proud of

We are incredibly proud of successfully turning static into signal. We celebrated the moment we transmitted a complex medical alert from one laptop to another using nothing but sound waves. It felt like magic!

Beyond the physics, we proved the viability of "Offline Trust." We demonstrated that a blockchain like Solana can verify identity without an internet connection simply by using cryptographic keys locally.What we learned

This project taught us that bandwidth is a luxury. It forced us to become ruthless with data because every single byte counts when transmitting over sound.

We also expanded our view of artificial intelligence, moving beyond simple "chat" interfaces to utilize AI as a core infrastructure component capable of complex compression and routing. Ultimately, we realized the immense potential of DePIN (Decentralized Physical Infrastructure Networks), understanding that the mesh networks we are building represent the future of true disaster resilience.

What's next for Cyren

Looking forward, we plan to expand beyond audio transmission by integrating LoRa (Long Range) radio modules to achieve multi-kilometer range for our mesh network. We also aim to port our software onto dedicated, ruggedized hardware devices, such as a Rabbit R1 or a custom Raspberry Pi build, to prepare for actual field deployment. Finally, we intend to move our audit logs from the Solana Devnet to Mainnet Beta, ensuring permanent, immutable record-keeping for all emergency communications.