Ctrl.

Ctrl. reinvents the app blocker by trading willpower for "Knowledge as Currency". It intercepts users at the OS level when opening distracting apps, requiring a 50% score on a generated micro-quiz to unlock the device. The backend uses a local ChromaDB and FastAPI, parsing PDFs via PyMuPDF or physical books via a Gemma model. Quiz Generation & Breaks The RAG pipeline filters chunks strictly to prevent hallucination. A "Fact-Scratchpad Strategy" isolates facts to generate multiple-choice questions. To manage cognitive load, an Auto-Adjusted Break gives a 20% break for sessions up to 60 minutes, scaling dynamically for longer sessions. AMD Compute Highlight We initially architected the RAG backend on the AMD Developer Cloud using a dedicated AMD Instinct MI300X GPU. This hardware-accelerated ROCm infrastructure was crucial for testing the GPU-intensive all-MiniLM-L6-v2 embedding model. We generated a hardware_proof.log file and verified API lifecycles before pivoting to a serverless architecture for production deployment on Render. Professional Expansion Ctrl. offers a freemium model with vast B2C and B2B potential. Notably, professionals can snap pictures of their own work, corporate memos, industry reports, or compliance manuals. This acts as a unique test type: the system generates dynamic quizzes directly from these materials to enforce engagement with workplace tasks before unlocking distractions.