The Tiny Tools philosophy nails it. When inference runs on your hardware, there's no data processing agreement to negotiate because there's no vendor. That gap between compliance-by-policy and compliance-by-architecture is what most people miss. Been testing the newer Qwen 3.5 small models for exactly this kind of local-first work and the 4B is surprisingly capable. Wrote up the full picture here: https://reading.sh/your-laptop-is-an-ai-server-now-370bad238461?sk=1cf7a4391e614720ecbd6e9bc3f076a2
The Tiny Tools philosophy nails it. When inference runs on your hardware, there's no data processing agreement to negotiate because there's no vendor. That gap between compliance-by-policy and compliance-by-architecture is what most people miss. Been testing the newer Qwen 3.5 small models for exactly this kind of local-first work and the 4B is surprisingly capable. Wrote up the full picture here: https://reading.sh/your-laptop-is-an-ai-server-now-370bad238461?sk=1cf7a4391e614720ecbd6e9bc3f076a2