Sick of cloud keyboards harvesting your telemetry for AI training? Same.

While drafting a highly sensitive patent, I got extremely paranoid about my dictation data and decided to hand-roll my own solution.

Meet Aegis: A strictly 100% offline, air-gapped AI voice engine for Android.

⚙️ The Architecture:
• Dual-model synergy running entirely on local silicon.
• Pushes the Snapdragon 8 Gen 3 NPU to its absolute physical limits (requires 12GB RAM).
• Zero network code. Zero telemetry. "Faraday Cage" UI.

If you care about true local computing and absolute silence on your network ports, check out the raw airplane-mode demos here:
🔗 https://aegisvoice.com

I just launched this on Hacker News and I'm answering questions about NPU buffer optimization there. Come say hi:https://news.ycombinator.com/item?id=47263679

Boosts heavily appreciated from fellow privacy advocates! 🛡️

#LocalAI #Privacy #Infosec #OnDeviceAI #CyberSecurity #ShowHN