ahmad parr

1. Local sovereign AI Use Ollama with: Qwen DeepSeek Llama Mistral Gemma Best for private agents, CLI tools, repo analysis, trust-deed reasoning, and no-subscription workflows. 2. Cloud “power lens” AI Use APIs only when needed: Gemini for hackathon demos, multimodal, Google ecosystem Claude for long reasoning/docs OpenAI for tool calling, structured workflows, polished assistants This keeps SnapKitty sovereign but lets you rent extra horsepower when useful. 3. ML / data stack For actual machine learning: Python + PyTorch for model experiments scikit-learn for simple classifiers, scoring, fraud/risk models XGBoost / LightGBM for credit score, vendor risk, trust scoring Polars / DuckDB for fast local data pipelines 4. SnapKitty-native intelligence This is your special sauce: Prolog for trust deeds / governance rules Lisp for symbolic agent reasoning Rust/WASM for secure execution TypeScript for UI/API orchestration WORM ledger for audit memory