Local brain
Import GGUF models and run inference directly on Android with llama.cpp.
SI is a personal AI that runs locally on your Android phone. Your model, memory, permissions and working style stay under your control.
Give me a URL, codebase or technical problem. I’ll inspect, reason, implement and verify.
01 / PHILOSOPHY
SI does not need a remote server to think. The model runs on your device, memory stays local, and internet access becomes a tool you can allow, ask for or deny.
Import GGUF models and run inference directly on Android with llama.cpp.
Conversations, benchmarks and operator memory are stored locally in app-private storage.
SI is not one model. Swap a faster 4B model for a stronger 8B model without changing SI itself.
SI is built to inspect the actual system, show evidence, implement and verify — not simply generate text.
02 / OPERATING MODES
Direct, technical problem solving at senior engineering and research level.
DEFAULT MODEPlans, uses permitted tools, executes steps and stops when your approval is required.
TOOL MODETreats the task as a benchmark. Evidence, implementation and verification are mandatory.
BENCHMARK MODE03 / OPERATOR CONTROL
Every capability has its own operator policy. Android permission is only the first layer — SI also checks your decision before using a capability.
04 / HARD TEST
05 / ARCHITECTURE
ANDROID / V0.1
First release targets modern Android and foldable devices. Local model. Local memory. Your rules.
SI V0.1ANDROID APK · AVAILABLE NOW→ V0.1 build verified · direct download