We build precise, thoughtful applications. AI assistants, productivity tools, and personal utilities — crafted with care, no bloat.
We go below the UI layer. JNI, Win32, AccessibilityService, null-routing VPNs. The kind of work most teams outsource or skip.
JNI bridges to llama.cpp for local LLM inference. Win32 process killing via JNA. Android AccessibilityService for instant OS interception. No UI-only abstractions.
Soft timers are not enforcement. We use Device Admin mode, null-routing on-device VPNs, Windows Firewall rules, and SHA-256 native PINs. Bypassing our apps takes real effort.
Our AI runs on your GPU. Your relationship data stays on your phone. Our VPNs null-route locally — no packet ever reaches our servers. Privacy is the architecture, not a setting.
On-device AI agent for Android. Runs quantized LLMs locally via a JNI bridge to llama.cpp with OpenCL GPU acceleration. Custom 1,063-line AgentLoop, SAM2 vision, RL policy network, AccessibilityService screen reading. Not a cloud chatbot.
Windows focus enforcement via Kotlin + Compose Multiplatform. Kills blocked processes with Win32 GetForegroundWindow() + JNA, adds live firewall rules via PowerShell. Packages as a standalone EXE or MSI. Real blocking.
All of our work is open source. Browse, contribute, or star what you find useful. We build in public.